protect bpool from buffer pollution by invalid buffers (#20342)

Currently, the bucket events and replication targets are only reloaded
with buckets that failed to load during the first cluster startup,
which is wrong because if one bucket change was done in one node but
that node was not able to notify other nodes; the other nodes will
reload the bucket metadata config but fails to set the events and bucket
targets in the memory.

.github/workflows/vulncheck.yml

@@ -21,8 +21,7 @@ jobs:
       - name: Set up Go
         uses: actions/setup-go@v5
         with:
-          go-version: 1.22.4
-          check-latest: true
+          go-version: 1.22.5
       - name: Get official govulncheck
         run: go install golang.org/x/vuln/cmd/govulncheck@latest
         shell: bash

.typos.toml

@@ -2,6 +2,9 @@
 extend-exclude = [
     ".git/",
     "docs/",
+    "CREDITS",
+    "go.mod",
+    "go.sum",
 ]
 ignore-hidden = false
 
@@ -16,6 +19,7 @@ extend-ignore-re = [
     "MIIDBTCCAe2gAwIBAgIQWHw7h.*",
     'http\.Header\{"X-Amz-Server-Side-Encryptio":',
     "ZoEoZdLlzVbOlT9rbhD7ZN7TLyiYXSAlB79uGEge",
+    "ERRO:",
 ]
 
 [default.extend-words]

CONTRIBUTING.md

@@ -12,8 +12,9 @@ Fork [MinIO upstream](https://github.com/minio/minio/fork) source repository to
 
 ```sh
 git clone https://github.com/minio/minio
+cd minio
 go install -v
-ls /go/bin/minio
+ls $(go env GOPATH)/bin/minio
 ```
 
 ### Set up git remote as ``upstream``

Dockerfile.hotfix

@@ -1,24 +1,26 @@
-FROM golang:1.21-alpine as build
+FROM golang:1.22-alpine as build
 
 ARG TARGETARCH
 ARG RELEASE
 
-ENV GOPATH /go
-ENV CGO_ENABLED 0
+ENV GOPATH=/go
+ENV CGO_ENABLED=0
 
 # Install curl and minisign
 RUN apk add -U --no-cache ca-certificates && \
     apk add -U --no-cache curl && \
     go install aead.dev/minisign/cmd/minisign@v0.2.1
 
-# Download minio binary and signature file
+# Download minio binary and signature files
 RUN curl -s -q https://dl.min.io/server/minio/hotfixes/linux-${TARGETARCH}/archive/minio.${RELEASE} -o /go/bin/minio && \
     curl -s -q https://dl.min.io/server/minio/hotfixes/linux-${TARGETARCH}/archive/minio.${RELEASE}.minisig -o /go/bin/minio.minisig && \
+    curl -s -q https://dl.min.io/server/minio/hotfixes/linux-${TARGETARCH}/archive/minio.${RELEASE}.sha256sum -o /go/bin/minio.sha256sum && \
     chmod +x /go/bin/minio
 
-# Download mc binary and signature file
+# Download mc binary and signature files
 RUN curl -s -q https://dl.min.io/client/mc/release/linux-${TARGETARCH}/mc -o /go/bin/mc && \
     curl -s -q https://dl.min.io/client/mc/release/linux-${TARGETARCH}/mc.minisig -o /go/bin/mc.minisig && \
+    curl -s -q https://dl.min.io/client/mc/release/linux-${TARGETARCH}/mc.sha256sum -o /go/bin/mc.sha256sum && \
     chmod +x /go/bin/mc
 
 RUN if [ "$TARGETARCH" = "amd64" ]; then \
@@ -52,8 +54,8 @@ ENV MINIO_ACCESS_KEY_FILE=access_key \
     MC_CONFIG_DIR=/tmp/.mc
 
 COPY --from=build /etc/ssl/certs/ca-certificates.crt /etc/ssl/certs/
-COPY --from=build /go/bin/minio /usr/bin/minio
-COPY --from=build /go/bin/mc /usr/bin/mc
+COPY --from=build /go/bin/minio* /usr/bin/
+COPY --from=build /go/bin/mc* /usr/bin/
 COPY --from=build /go/bin/cur* /usr/bin/
 
 COPY CREDITS /licenses/CREDITS

Dockerfile.release

@@ -1,24 +1,26 @@
-FROM golang:1.21-alpine as build
+FROM golang:1.22-alpine AS build
 
 ARG TARGETARCH
 ARG RELEASE
 
-ENV GOPATH /go
-ENV CGO_ENABLED 0
+ENV GOPATH=/go
+ENV CGO_ENABLED=0
 
 # Install curl and minisign
 RUN apk add -U --no-cache ca-certificates && \
     apk add -U --no-cache curl && \
     go install aead.dev/minisign/cmd/minisign@v0.2.1
 
-# Download minio binary and signature file
+# Download minio binary and signature files
 RUN curl -s -q https://dl.min.io/server/minio/release/linux-${TARGETARCH}/archive/minio.${RELEASE} -o /go/bin/minio && \
     curl -s -q https://dl.min.io/server/minio/release/linux-${TARGETARCH}/archive/minio.${RELEASE}.minisig -o /go/bin/minio.minisig && \
+    curl -s -q https://dl.min.io/server/minio/release/linux-${TARGETARCH}/archive/minio.${RELEASE}.sha256sum -o /go/bin/minio.sha256sum && \
     chmod +x /go/bin/minio
 
-# Download mc binary and signature file
+# Download mc binary and signature files
 RUN curl -s -q https://dl.min.io/client/mc/release/linux-${TARGETARCH}/mc -o /go/bin/mc && \
     curl -s -q https://dl.min.io/client/mc/release/linux-${TARGETARCH}/mc.minisig -o /go/bin/mc.minisig && \
+    curl -s -q https://dl.min.io/client/mc/release/linux-${TARGETARCH}/mc.sha256sum -o /go/bin/mc.sha256sum && \
     chmod +x /go/bin/mc
 
 RUN if [ "$TARGETARCH" = "amd64" ]; then \
@@ -52,8 +54,8 @@ ENV MINIO_ACCESS_KEY_FILE=access_key \
     MC_CONFIG_DIR=/tmp/.mc
 
 COPY --from=build /etc/ssl/certs/ca-certificates.crt /etc/ssl/certs/
-COPY --from=build /go/bin/minio /usr/bin/minio
-COPY --from=build /go/bin/mc /usr/bin/mc
+COPY --from=build /go/bin/minio* /usr/bin/
+COPY --from=build /go/bin/mc* /usr/bin/
 COPY --from=build /go/bin/cur* /usr/bin/
 
 COPY CREDITS /licenses/CREDITS

Dockerfile.release.fips

@@ -1,21 +1,28 @@
-FROM golang:1.21-alpine as build
+FROM golang:1.22-alpine AS build
 
 ARG TARGETARCH
 ARG RELEASE
 
-ENV GOPATH /go
-ENV CGO_ENABLED 0
+ENV GOPATH=/go
+ENV CGO_ENABLED=0
 
 # Install curl and minisign
 RUN apk add -U --no-cache ca-certificates && \
     apk add -U --no-cache curl && \
     go install aead.dev/minisign/cmd/minisign@v0.2.1
 
-# Download minio binary and signature file
+# Download minio binary and signature files
 RUN curl -s -q https://dl.min.io/server/minio/release/linux-${TARGETARCH}/archive/minio.${RELEASE}.fips -o /go/bin/minio && \
     curl -s -q https://dl.min.io/server/minio/release/linux-${TARGETARCH}/archive/minio.${RELEASE}.fips.minisig -o /go/bin/minio.minisig && \
+    curl -s -q https://dl.min.io/server/minio/release/linux-${TARGETARCH}/archive/minio.${RELEASE}.fips.sha256sum -o /go/bin/minio.sha256sum && \
     chmod +x /go/bin/minio
 
+# Download mc binary and signature files
+RUN curl -s -q https://dl.min.io/client/mc/release/linux-${TARGETARCH}/mc.fips -o /go/bin/mc && \
+    curl -s -q https://dl.min.io/client/mc/release/linux-${TARGETARCH}/mc.fips.minisig -o /go/bin/mc.minisig && \
+    curl -s -q https://dl.min.io/client/mc/release/linux-${TARGETARCH}/mc.fips.sha256sum -o /go/bin/mc.sha256sum && \
+    chmod +x /go/bin/mc
+
 RUN if [ "$TARGETARCH" = "amd64" ]; then \
        curl -L -s -q https://github.com/moparisthebest/static-curl/releases/latest/download/curl-${TARGETARCH} -o /go/bin/curl; \
        chmod +x /go/bin/curl; \
@@ -45,7 +52,8 @@ ENV MINIO_ACCESS_KEY_FILE=access_key \
     MINIO_CONFIG_ENV_FILE=config.env
 
 COPY --from=build /etc/ssl/certs/ca-certificates.crt /etc/ssl/certs/
-COPY --from=build /go/bin/minio /usr/bin/minio
+COPY --from=build /go/bin/minio* /usr/bin/
+COPY --from=build /go/bin/mc* /usr/bin/
 COPY --from=build /go/bin/cur* /usr/bin/
 
 COPY CREDITS /licenses/CREDITS

Dockerfile.release.old_cpu

@@ -1,24 +1,26 @@
-FROM golang:1.21-alpine as build
+FROM golang:1.22-alpine AS build
 
 ARG TARGETARCH
 ARG RELEASE
 
-ENV GOPATH /go
-ENV CGO_ENABLED 0
+ENV GOPATH=/go
+ENV CGO_ENABLED=0
 
 # Install curl and minisign
 RUN apk add -U --no-cache ca-certificates && \
     apk add -U --no-cache curl && \
     go install aead.dev/minisign/cmd/minisign@v0.2.1
 
-# Download minio binary and signature file
+# Download minio binary and signature files
 RUN curl -s -q https://dl.min.io/server/minio/release/linux-${TARGETARCH}/archive/minio.${RELEASE} -o /go/bin/minio && \
     curl -s -q https://dl.min.io/server/minio/release/linux-${TARGETARCH}/archive/minio.${RELEASE}.minisig -o /go/bin/minio.minisig && \
+    curl -s -q https://dl.min.io/server/minio/release/linux-${TARGETARCH}/archive/minio.${RELEASE}.sha256sum -o /go/bin/minio.sha256sum && \
     chmod +x /go/bin/minio
 
-# Download mc binary and signature file
+# Download mc binary and signature files
 RUN curl -s -q https://dl.min.io/client/mc/release/linux-${TARGETARCH}/mc -o /go/bin/mc && \
     curl -s -q https://dl.min.io/client/mc/release/linux-${TARGETARCH}/mc.minisig -o /go/bin/mc.minisig && \
+    curl -s -q https://dl.min.io/client/mc/release/linux-${TARGETARCH}/mc.sha256sum -o /go/bin/mc.sha256sum && \
     chmod +x /go/bin/mc
 
 RUN if [ "$TARGETARCH" = "amd64" ]; then \
@@ -52,8 +54,8 @@ ENV MINIO_ACCESS_KEY_FILE=access_key \
     MC_CONFIG_DIR=/tmp/.mc
 
 COPY --from=build /etc/ssl/certs/ca-certificates.crt /etc/ssl/certs/
-COPY --from=build /go/bin/minio /usr/bin/minio
-COPY --from=build /go/bin/mc /usr/bin/mc
+COPY --from=build /go/bin/minio* /usr/bin/
+COPY --from=build /go/bin/mc* /usr/bin/
 COPY --from=build /go/bin/cur* /usr/bin/
 
 COPY CREDITS /licenses/CREDITS

Makefile

@@ -34,11 +34,14 @@ verifiers: lint check-gen
 
 check-gen: ## check for updated autogenerated files
 	@go generate ./... >/dev/null
+	@go mod tidy -compat=1.21
 	@(! git diff --name-only | grep '_gen.go$$') || (echo "Non-committed changes in auto-generated code is detected, please commit them to proceed." && false)
+	@(! git diff --name-only | grep 'go.sum') || (echo "Non-committed changes in auto-generated go.sum is detected, please commit them to proceed." && false)
 
 lint: getdeps ## runs golangci-lint suite of linters
 	@echo "Running $@ check"
 	@$(GOLANGCI) run --build-tags kqueue --timeout=10m --config ./.golangci.yml
+	@command typos && typos ./ || echo "typos binary is not found.. skipping.."
 
 lint-fix: getdeps ## runs golangci-lint suite of linters with automatic fixes
 	@echo "Running $@ check"
@@ -86,9 +89,9 @@ test-race: verifiers build ## builds minio, runs linters, tests (race)
 
 test-iam: install-race ## verify IAM (external IDP, etcd backends)
 	@echo "Running tests for IAM (external IDP, etcd backends)"
-	@MINIO_API_REQUESTS_MAX=10000 CGO_ENABLED=0 go test -tags kqueue,dev -v -run TestIAM* ./cmd
+	@MINIO_API_REQUESTS_MAX=10000 CGO_ENABLED=0 go test -timeout 15m -tags kqueue,dev -v -run TestIAM* ./cmd
 	@echo "Running tests for IAM (external IDP, etcd backends) with -race"
-	@MINIO_API_REQUESTS_MAX=10000 GORACE=history_size=7 CGO_ENABLED=1 go test -race -tags kqueue,dev -v -run TestIAM* ./cmd
+	@MINIO_API_REQUESTS_MAX=10000 GORACE=history_size=7 CGO_ENABLED=1 go test -timeout 15m -race -tags kqueue,dev -v -run TestIAM* ./cmd
 
 test-iam-ldap-upgrade-import: install-race ## verify IAM (external LDAP IDP)
 	@echo "Running upgrade tests for IAM (LDAP backend)"
@@ -165,9 +168,9 @@ hotfix-vars:
 	$(eval VERSION := $(shell git describe --tags --abbrev=0).hotfix.$(shell git rev-parse --short HEAD))
 
 hotfix: hotfix-vars clean install ## builds minio binary with hotfix tags
-	@wget -q -c https://github.com/minio/pkger/releases/download/v2.2.9/pkger_2.2.9_linux_amd64.deb
+	@wget -q -c https://github.com/minio/pkger/releases/download/v2.3.1/pkger_2.3.1_linux_amd64.deb
 	@wget -q -c https://raw.githubusercontent.com/minio/minio-service/v1.0.1/linux-systemd/distributed/minio.service
-	@sudo apt install ./pkger_2.2.9_linux_amd64.deb --yes
+	@sudo apt install ./pkger_2.3.1_linux_amd64.deb --yes
 	@mkdir -p minio-release/$(GOOS)-$(GOARCH)/archive
 	@cp -af ./minio minio-release/$(GOOS)-$(GOARCH)/minio
 	@cp -af ./minio minio-release/$(GOOS)-$(GOARCH)/minio.$(VERSION)

buildscripts/verify-healing-empty-erasure-set.sh

@@ -39,7 +39,7 @@ function start_minio_3_node() {
 
 	export MC_HOST_myminio="http://minio:minio123@127.0.0.1:$((start_port + 1))"
 
-	/tmp/mc ready myminio
+	timeout 15m /tmp/mc ready myminio || fail
 
 	# Wait for all drives to be online and formatted
 	while [ $(/tmp/mc admin info --json myminio | jq '.info.servers[].drives[].state | select(. != "ok")' | wc -l) -gt 0 ]; do sleep 1; done
@@ -62,48 +62,23 @@ function start_minio_3_node() {
 		fi
 
 		# Failure
-		for i in $(seq 1 3); do
-			echo "server$i log:"
-			cat "${WORK_DIR}/dist-minio-server$i.log"
-		done
-		pkill -9 minio
-		echo "FAILED"
-		purge "$WORK_DIR"
-		exit 1
+		fail
 	done
 
 	if ! ps -p $pid1 1>&2 >/dev/null; then
-		echo "server1 log:"
-		cat "${WORK_DIR}/dist-minio-server1.log"
-		echo "FAILED"
-		purge "$WORK_DIR"
-		exit 1
+		echo "minio-server-1 is not running." && fail
 	fi
 
 	if ! ps -p $pid2 1>&2 >/dev/null; then
-		echo "server2 log:"
-		cat "${WORK_DIR}/dist-minio-server2.log"
-		echo "FAILED"
-		purge "$WORK_DIR"
-		exit 1
+		echo "minio-server-2 is not running." && fail
 	fi
 
 	if ! ps -p $pid3 1>&2 >/dev/null; then
-		echo "server3 log:"
-		cat "${WORK_DIR}/dist-minio-server3.log"
-		echo "FAILED"
-		purge "$WORK_DIR"
-		exit 1
+		echo "minio-server-3 is not running." && fail
 	fi
 
 	if ! pkill minio; then
-		for i in $(seq 1 3); do
-			echo "server$i log:"
-			cat "${WORK_DIR}/dist-minio-server$i.log"
-		done
-		echo "FAILED"
-		purge "$WORK_DIR"
-		exit 1
+		fail
 	fi
 
 	sleep 1
@@ -115,8 +90,18 @@ function start_minio_3_node() {
 	fi
 }
 
+function fail() {
+	for i in $(seq 1 3); do
+		echo "server$i log:"
+		cat "${WORK_DIR}/dist-minio-server$i.log"
+	done
+	echo "FAILED"
+	purge "$WORK_DIR"
+	exit 1
+}
+
 function check_online() {
-	if ! grep -q 'Status:' ${WORK_DIR}/dist-minio-*.log; then
+	if ! grep -q 'API:' ${WORK_DIR}/dist-minio-*.log; then
 		echo "1"
 	fi
 }

buildscripts/verify-healing.sh

@@ -47,43 +47,25 @@ function start_minio_3_node() {
 	disown $pid3
 
 	export MC_HOST_myminio="http://minio:minio123@127.0.0.1:$((start_port + 1))"
-	/tmp/mc ready myminio
+	timeout 15m /tmp/mc ready myminio || fail
 
 	[ ${first_time} -eq 0 ] && upload_objects
 	[ ${first_time} -ne 0 ] && sleep 120
 
 	if ! ps -p $pid1 1>&2 >/dev/null; then
-		echo "server1 log:"
-		cat "${WORK_DIR}/dist-minio-server1.log"
-		echo "FAILED"
-		purge "$WORK_DIR"
-		exit 1
+		echo "minio server 1 is not running" && fail
 	fi
 
 	if ! ps -p $pid2 1>&2 >/dev/null; then
-		echo "server2 log:"
-		cat "${WORK_DIR}/dist-minio-server2.log"
-		echo "FAILED"
-		purge "$WORK_DIR"
-		exit 1
+		echo "minio server 2 is not running" && fail
 	fi
 
 	if ! ps -p $pid3 1>&2 >/dev/null; then
-		echo "server3 log:"
-		cat "${WORK_DIR}/dist-minio-server3.log"
-		echo "FAILED"
-		purge "$WORK_DIR"
-		exit 1
+		echo "minio server 3 is not running" && fail
 	fi
 
 	if ! pkill minio; then
-		for i in $(seq 1 3); do
-			echo "server$i log:"
-			cat "${WORK_DIR}/dist-minio-server$i.log"
-		done
-		echo "FAILED"
-		purge "$WORK_DIR"
-		exit 1
+		fail
 	fi
 
 	sleep 1
@@ -96,7 +78,7 @@ function start_minio_3_node() {
 }
 
 function check_heal() {
-	if ! grep -q 'Status:' ${WORK_DIR}/dist-minio-*.log; then
+	if ! grep -q 'API:' ${WORK_DIR}/dist-minio-*.log; then
 		return 1
 	fi
 
@@ -118,6 +100,17 @@ function purge() {
 	rm -rf "$1"
 }
 
+function fail() {
+	for i in $(seq 1 3); do
+		echo "server$i log:"
+		cat "${WORK_DIR}/dist-minio-server$i.log"
+	done
+	pkill -9 minio
+	echo "FAILED"
+	purge "$WORK_DIR"
+	exit 1
+}
+
 function __init__() {
 	echo "Initializing environment"
 	mkdir -p "$WORK_DIR"
@@ -155,14 +148,7 @@ function perform_test() {
 	check_heal ${1}
 	rv=$?
 	if [ "$rv" == "1" ]; then
-		for i in $(seq 1 3); do
-			echo "server$i log:"
-			cat "${WORK_DIR}/dist-minio-server$i.log"
-		done
-		pkill -9 minio
-		echo "FAILED"
-		purge "$WORK_DIR"
-		exit 1
+		fail
 	fi
 }
 

cmd/admin-bucket-handlers.go

@@ -427,6 +427,21 @@ func (a adminAPIHandlers) ExportBucketMetadataHandler(w http.ResponseWriter, r *
 			cfgPath := pathJoin(bi.Name, cfgFile)
 			bucket := bi.Name
 			switch cfgFile {
+			case bucketPolicyConfig:
+				config, _, err := globalBucketMetadataSys.GetBucketPolicy(bucket)
+				if err != nil {
+					if errors.Is(err, BucketPolicyNotFound{Bucket: bucket}) {
+						continue
+					}
+					writeErrorResponse(ctx, w, exportError(ctx, err, cfgFile, bucket), r.URL)
+					return
+				}
+				configData, err := json.Marshal(config)
+				if err != nil {
+					writeErrorResponse(ctx, w, exportError(ctx, err, cfgFile, bucket), r.URL)
+					return
+				}
+				rawDataFn(bytes.NewReader(configData), cfgPath, len(configData))
 			case bucketNotificationConfig:
 				config, err := globalBucketMetadataSys.GetNotificationConfig(bucket)
 				if err != nil {
@@ -796,11 +811,12 @@ func (a adminAPIHandlers) ImportBucketMetadataHandler(w http.ResponseWriter, r *
 			}
 
 			bucketMap[bucket].NotificationConfigXML = configData
+			bucketMap[bucket].NotificationConfigUpdatedAt = updatedAt
 			rpt.SetStatus(bucket, fileName, nil)
 		case bucketPolicyConfig:
 			// Error out if Content-Length is beyond allowed size.
 			if sz > maxBucketPolicySize {
-				rpt.SetStatus(bucket, fileName, fmt.Errorf(ErrPolicyTooLarge.String()))
+				rpt.SetStatus(bucket, fileName, errors.New(ErrPolicyTooLarge.String()))
 				continue
 			}
 
@@ -818,7 +834,7 @@ func (a adminAPIHandlers) ImportBucketMetadataHandler(w http.ResponseWriter, r *
 
 			// Version in policy must not be empty
 			if bucketPolicy.Version == "" {
-				rpt.SetStatus(bucket, fileName, fmt.Errorf(ErrPolicyInvalidVersion.String()))
+				rpt.SetStatus(bucket, fileName, errors.New(ErrPolicyInvalidVersion.String()))
 				continue
 			}
 

cmd/admin-handler-utils.go

@@ -216,6 +216,12 @@ func toAdminAPIErr(ctx context.Context, err error) APIError {
 				Description:    err.Error(),
 				HTTPStatusCode: http.StatusBadRequest,
 			}
+		case errors.Is(err, errTierInvalidConfig):
+			apiErr = APIError{
+				Code:           "XMinioAdminTierInvalidConfig",
+				Description:    err.Error(),
+				HTTPStatusCode: http.StatusBadRequest,
+			}
 		default:
 			apiErr = errorCodes.ToAPIErrWithErr(toAdminAPIErrCode(ctx, err), err)
 		}

cmd/admin-handlers-idp-ldap.go

@@ -479,3 +479,180 @@ func (a adminAPIHandlers) ListAccessKeysLDAP(w http.ResponseWriter, r *http.Requ
 
 	writeSuccessResponseJSON(w, encryptedData)
 }
+
+// ListAccessKeysLDAPBulk - GET /minio/admin/v3/idp/ldap/list-access-keys-bulk
+func (a adminAPIHandlers) ListAccessKeysLDAPBulk(w http.ResponseWriter, r *http.Request) {
+	ctx := r.Context()
+
+	// Get current object layer instance.
+	objectAPI := newObjectLayerFn()
+	if objectAPI == nil || globalNotificationSys == nil {
+		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrServerNotInitialized), r.URL)
+		return
+	}
+
+	cred, owner, s3Err := validateAdminSignature(ctx, r, "")
+	if s3Err != ErrNone {
+		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(s3Err), r.URL)
+		return
+	}
+
+	dnList := r.Form["userDNs"]
+	isAll := r.Form.Get("all") == "true"
+	onlySelf := !isAll && len(dnList) == 0
+
+	if isAll && len(dnList) > 0 {
+		// This should be checked on client side, so return generic error
+		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrInvalidRequest), r.URL)
+		return
+	}
+
+	// Empty DN list and not self, list access keys for all users
+	if isAll {
+		if !globalIAMSys.IsAllowed(policy.Args{
+			AccountName:     cred.AccessKey,
+			Groups:          cred.Groups,
+			Action:          policy.ListUsersAdminAction,
+			ConditionValues: getConditionValues(r, "", cred),
+			IsOwner:         owner,
+			Claims:          cred.Claims,
+		}) {
+			writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrAccessDenied), r.URL)
+			return
+		}
+	} else if len(dnList) == 1 {
+		var dn string
+		foundResult, err := globalIAMSys.LDAPConfig.GetValidatedDNForUsername(dnList[0])
+		if err == nil {
+			dn = foundResult.NormDN
+		}
+		if dn == cred.ParentUser || dnList[0] == cred.ParentUser {
+			onlySelf = true
+		}
+	}
+
+	if !globalIAMSys.IsAllowed(policy.Args{
+		AccountName:     cred.AccessKey,
+		Groups:          cred.Groups,
+		Action:          policy.ListServiceAccountsAdminAction,
+		ConditionValues: getConditionValues(r, "", cred),
+		IsOwner:         owner,
+		Claims:          cred.Claims,
+		DenyOnly:        onlySelf,
+	}) {
+		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrAccessDenied), r.URL)
+		return
+	}
+
+	if onlySelf && len(dnList) == 0 {
+		selfDN := cred.AccessKey
+		if cred.ParentUser != "" {
+			selfDN = cred.ParentUser
+		}
+		dnList = append(dnList, selfDN)
+	}
+
+	var ldapUserList []string
+	if isAll {
+		ldapUsers, err := globalIAMSys.ListLDAPUsers(ctx)
+		if err != nil {
+			writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
+			return
+		}
+		for user := range ldapUsers {
+			ldapUserList = append(ldapUserList, user)
+		}
+	} else {
+		for _, userDN := range dnList {
+			// Validate the userDN
+			foundResult, err := globalIAMSys.LDAPConfig.GetValidatedDNForUsername(userDN)
+			if err != nil {
+				writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
+				return
+			}
+			if foundResult == nil {
+				continue
+			}
+			ldapUserList = append(ldapUserList, foundResult.NormDN)
+		}
+	}
+
+	listType := r.Form.Get("listType")
+	var listSTSKeys, listServiceAccounts bool
+	switch listType {
+	case madmin.AccessKeyListUsersOnly:
+		listSTSKeys = false
+		listServiceAccounts = false
+	case madmin.AccessKeyListSTSOnly:
+		listSTSKeys = true
+		listServiceAccounts = false
+	case madmin.AccessKeyListSvcaccOnly:
+		listSTSKeys = false
+		listServiceAccounts = true
+	case madmin.AccessKeyListAll:
+		listSTSKeys = true
+		listServiceAccounts = true
+	default:
+		err := errors.New("invalid list type")
+		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErrWithErr(ErrInvalidRequest, err), r.URL)
+		return
+	}
+
+	accessKeyMap := make(map[string]madmin.ListAccessKeysLDAPResp)
+	for _, internalDN := range ldapUserList {
+		externalDN := globalIAMSys.LDAPConfig.DecodeDN(internalDN)
+		accessKeys := madmin.ListAccessKeysLDAPResp{}
+		if listSTSKeys {
+			stsKeys, err := globalIAMSys.ListSTSAccounts(ctx, internalDN)
+			if err != nil {
+				writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
+				return
+			}
+			for _, sts := range stsKeys {
+				expiryTime := sts.Expiration
+				accessKeys.STSKeys = append(accessKeys.STSKeys, madmin.ServiceAccountInfo{
+					AccessKey:  sts.AccessKey,
+					Expiration: &expiryTime,
+				})
+			}
+			// if only STS keys, skip if user has no STS keys
+			if !listServiceAccounts && len(stsKeys) == 0 {
+				continue
+			}
+		}
+
+		if listServiceAccounts {
+			serviceAccounts, err := globalIAMSys.ListServiceAccounts(ctx, internalDN)
+			if err != nil {
+				writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
+				return
+			}
+			for _, svc := range serviceAccounts {
+				expiryTime := svc.Expiration
+				accessKeys.ServiceAccounts = append(accessKeys.ServiceAccounts, madmin.ServiceAccountInfo{
+					AccessKey:  svc.AccessKey,
+					Expiration: &expiryTime,
+				})
+			}
+			// if only service accounts, skip if user has no service accounts
+			if !listSTSKeys && len(serviceAccounts) == 0 {
+				continue
+			}
+		}
+		accessKeyMap[externalDN] = accessKeys
+	}
+
+	data, err := json.Marshal(accessKeyMap)
+	if err != nil {
+		writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
+		return
+	}
+
+	encryptedData, err := madmin.EncryptData(cred.SecretKey, data)
+	if err != nil {
+		writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
+		return
+	}
+
+	writeSuccessResponseJSON(w, encryptedData)
+}

cmd/admin-handlers-pools.go

@@ -374,6 +374,7 @@ func (a adminAPIHandlers) RebalanceStop(w http.ResponseWriter, r *http.Request)
 	globalNotificationSys.StopRebalance(r.Context())
 	writeSuccessResponseHeadersOnly(w)
 	adminLogIf(ctx, pools.saveRebalanceStats(GlobalContext, 0, rebalSaveStoppedAt))
+	globalNotificationSys.LoadRebalanceMeta(ctx, false)
 }
 
 func proxyDecommissionRequest(ctx context.Context, defaultEndPoint Endpoint, w http.ResponseWriter, r *http.Request) (proxy bool) {

cmd/admin-handlers-users-race_test.go

@@ -120,9 +120,12 @@ func (s *TestSuiteIAM) TestDeleteUserRace(c *check) {
 			c.Fatalf("Unable to set user: %v", err)
 		}
 
-		err = s.adm.SetPolicy(ctx, policy, accessKey, false)
-		if err != nil {
-			c.Fatalf("Unable to set policy: %v", err)
+		userReq := madmin.PolicyAssociationReq{
+			Policies: []string{policy},
+			User:     accessKey,
+		}
+		if _, err := s.adm.AttachPolicy(ctx, userReq); err != nil {
+			c.Fatalf("Unable to attach policy: %v", err)
 		}
 
 		accessKeys[i] = accessKey

cmd/admin-handlers-users.go

@@ -34,7 +34,6 @@ import (
 	"github.com/klauspost/compress/zip"
 	"github.com/minio/madmin-go/v3"
 	"github.com/minio/minio/internal/auth"
-	"github.com/minio/minio/internal/cachevalue"
 	"github.com/minio/minio/internal/config/dns"
 	"github.com/minio/mux"
 	xldap "github.com/minio/pkg/v3/ldap"
@@ -1235,18 +1234,6 @@ func (a adminAPIHandlers) AccountInfoHandler(w http.ResponseWriter, r *http.Requ
 		return rd, wr
 	}
 
-	bucketStorageCache.InitOnce(10*time.Second,
-		cachevalue.Opts{ReturnLastGood: true},
-		func(ctx context.Context) (DataUsageInfo, error) {
-			ctx, done := context.WithTimeout(ctx, 2*time.Second)
-			defer done()
-
-			return loadDataUsageFromBackend(ctx, objectAPI)
-		},
-	)
-
-	dataUsageInfo, _ := bucketStorageCache.Get()
-
 	// If etcd, dns federation configured list buckets from etcd.
 	var err error
 	var buckets []BucketInfo
@@ -1333,15 +1320,12 @@ func (a adminAPIHandlers) AccountInfoHandler(w http.ResponseWriter, r *http.Requ
 		rd, wr := isAllowedAccess(bucket.Name)
 		if rd || wr {
 			// Fetch the data usage of the current bucket
-			var size uint64
-			var objectsCount uint64
-			var objectsHist, versionsHist map[string]uint64
-			if !dataUsageInfo.LastUpdate.IsZero() {
-				size = dataUsageInfo.BucketsUsage[bucket.Name].Size
-				objectsCount = dataUsageInfo.BucketsUsage[bucket.Name].ObjectsCount
-				objectsHist = dataUsageInfo.BucketsUsage[bucket.Name].ObjectSizesHistogram
-				versionsHist = dataUsageInfo.BucketsUsage[bucket.Name].ObjectVersionsHistogram
-			}
+			bui := globalBucketQuotaSys.GetBucketUsageInfo(ctx, bucket.Name)
+			size := bui.Size
+			objectsCount := bui.ObjectsCount
+			objectsHist := bui.ObjectSizesHistogram
+			versionsHist := bui.ObjectVersionsHistogram
+
 			// Fetch the prefix usage of the current bucket
 			var prefixUsage map[string]uint64
 			if enablePrefixUsage {
@@ -2308,7 +2292,7 @@ func (a adminAPIHandlers) ImportIAM(w http.ResponseWriter, r *http.Request) {
 					// clean import.
 					err := globalIAMSys.DeleteServiceAccount(ctx, svcAcctReq.AccessKey, true)
 					if err != nil {
-						delErr := fmt.Errorf("failed to delete existing service account(%s) before importing it: %w", svcAcctReq.AccessKey, err)
+						delErr := fmt.Errorf("failed to delete existing service account (%s) before importing it: %w", svcAcctReq.AccessKey, err)
 						writeErrorResponseJSON(ctx, w, importError(ctx, delErr, allSvcAcctsFile, user), r.URL)
 						return
 					}

cmd/admin-handlers-users_test.go

@@ -239,9 +239,12 @@ func (s *TestSuiteIAM) TestUserCreate(c *check) {
 	c.Assert(v.Status, madmin.AccountEnabled)
 
 	// 3. Associate policy and check that user can access
-	err = s.adm.SetPolicy(ctx, "readwrite", accessKey, false)
+	_, err = s.adm.AttachPolicy(ctx, madmin.PolicyAssociationReq{
+		Policies: []string{"readwrite"},
+		User:     accessKey,
+	})
 	if err != nil {
-		c.Fatalf("unable to set policy: %v", err)
+		c.Fatalf("unable to attach policy: %v", err)
 	}
 
 	client := s.getUserClient(c, accessKey, secretKey, "")
@@ -348,9 +351,12 @@ func (s *TestSuiteIAM) TestUserPolicyEscalationBug(c *check) {
 	if err != nil {
 		c.Fatalf("policy add error: %v", err)
 	}
-	err = s.adm.SetPolicy(ctx, policy, accessKey, false)
+	_, err = s.adm.AttachPolicy(ctx, madmin.PolicyAssociationReq{
+		Policies: []string{policy},
+		User:     accessKey,
+	})
 	if err != nil {
-		c.Fatalf("Unable to set policy: %v", err)
+		c.Fatalf("unable to attach policy: %v", err)
 	}
 	// 2.3 check user has access to bucket
 	c.mustListObjects(ctx, uClient, bucket)
@@ -470,9 +476,12 @@ func (s *TestSuiteIAM) TestAddServiceAccountPerms(c *check) {
 	c.mustNotListObjects(ctx, uClient, "testbucket")
 
 	// 3.2 associate policy to user
-	err = s.adm.SetPolicy(ctx, policy1, accessKey, false)
+	_, err = s.adm.AttachPolicy(ctx, madmin.PolicyAssociationReq{
+		Policies: []string{policy1},
+		User:     accessKey,
+	})
 	if err != nil {
-		c.Fatalf("Unable to set policy: %v", err)
+		c.Fatalf("unable to attach policy: %v", err)
 	}
 
 	admClnt := s.getAdminClient(c, accessKey, secretKey, "")
@@ -490,10 +499,22 @@ func (s *TestSuiteIAM) TestAddServiceAccountPerms(c *check) {
 		c.Fatalf("policy was missing!")
 	}
 
-	// 3.2 associate policy to user
-	err = s.adm.SetPolicy(ctx, policy2, accessKey, false)
+	// Detach policy1 to set up for policy2
+	_, err = s.adm.DetachPolicy(ctx, madmin.PolicyAssociationReq{
+		Policies: []string{policy1},
+		User:     accessKey,
+	})
 	if err != nil {
-		c.Fatalf("Unable to set policy: %v", err)
+		c.Fatalf("unable to detach policy: %v", err)
+	}
+
+	// 3.2 associate policy to user
+	_, err = s.adm.AttachPolicy(ctx, madmin.PolicyAssociationReq{
+		Policies: []string{policy2},
+		User:     accessKey,
+	})
+	if err != nil {
+		c.Fatalf("unable to attach policy: %v", err)
 	}
 
 	// 3.3 check user can create service account implicitly.
@@ -571,9 +592,12 @@ func (s *TestSuiteIAM) TestPolicyCreate(c *check) {
 	c.mustNotListObjects(ctx, uClient, bucket)
 
 	// 3.2 associate policy to user
-	err = s.adm.SetPolicy(ctx, policy, accessKey, false)
+	_, err = s.adm.AttachPolicy(ctx, madmin.PolicyAssociationReq{
+		Policies: []string{policy},
+		User:     accessKey,
+	})
 	if err != nil {
-		c.Fatalf("Unable to set policy: %v", err)
+		c.Fatalf("unable to attach policy: %v", err)
 	}
 	// 3.3 check user has access to bucket
 	c.mustListObjects(ctx, uClient, bucket)
@@ -726,9 +750,12 @@ func (s *TestSuiteIAM) TestGroupAddRemove(c *check) {
 	c.mustNotListObjects(ctx, uClient, bucket)
 
 	// 3. Associate policy to group and check user got access.
-	err = s.adm.SetPolicy(ctx, policy, group, true)
+	_, err = s.adm.AttachPolicy(ctx, madmin.PolicyAssociationReq{
+		Policies: []string{policy},
+		Group:    group,
+	})
 	if err != nil {
-		c.Fatalf("Unable to set policy: %v", err)
+		c.Fatalf("unable to attach policy: %v", err)
 	}
 	// 3.1 check user has access to bucket
 	c.mustListObjects(ctx, uClient, bucket)
@@ -871,9 +898,12 @@ func (s *TestSuiteIAM) TestServiceAccountOpsByUser(c *check) {
 		c.Fatalf("Unable to set user: %v", err)
 	}
 
-	err = s.adm.SetPolicy(ctx, policy, accessKey, false)
+	_, err = s.adm.AttachPolicy(ctx, madmin.PolicyAssociationReq{
+		Policies: []string{policy},
+		User:     accessKey,
+	})
 	if err != nil {
-		c.Fatalf("Unable to set policy: %v", err)
+		c.Fatalf("unable to attach policy: %v", err)
 	}
 
 	// Create an madmin client with user creds
@@ -952,9 +982,12 @@ func (s *TestSuiteIAM) TestServiceAccountDurationSecondsCondition(c *check) {
 		c.Fatalf("Unable to set user: %v", err)
 	}
 
-	err = s.adm.SetPolicy(ctx, policy, accessKey, false)
+	_, err = s.adm.AttachPolicy(ctx, madmin.PolicyAssociationReq{
+		Policies: []string{policy},
+		User:     accessKey,
+	})
 	if err != nil {
-		c.Fatalf("Unable to set policy: %v", err)
+		c.Fatalf("unable to attach policy: %v", err)
 	}
 
 	// Create an madmin client with user creds
@@ -1031,9 +1064,12 @@ func (s *TestSuiteIAM) TestServiceAccountOpsByAdmin(c *check) {
 		c.Fatalf("Unable to set user: %v", err)
 	}
 
-	err = s.adm.SetPolicy(ctx, policy, accessKey, false)
+	_, err = s.adm.AttachPolicy(ctx, madmin.PolicyAssociationReq{
+		Policies: []string{policy},
+		User:     accessKey,
+	})
 	if err != nil {
-		c.Fatalf("Unable to set policy: %v", err)
+		c.Fatalf("unable to attach policy: %v", err)
 	}
 
 	// 1. Create a service account for the user

cmd/admin-handlers.go

@@ -237,6 +237,7 @@ func (a adminAPIHandlers) ServerUpdateV2Handler(w http.ResponseWriter, r *http.R
 
 	if globalIsDistErasure {
 		// Notify all other MinIO peers signal service.
+		startTime := time.Now().Add(restartUpdateDelay)
 		ng := WithNPeers(len(globalNotificationSys.peerClients))
 		for idx, client := range globalNotificationSys.peerClients {
 			_, ok := failedClients[idx]
@@ -247,7 +248,7 @@ func (a adminAPIHandlers) ServerUpdateV2Handler(w http.ResponseWriter, r *http.R
 			ng.Go(ctx, func() error {
 				prs, ok := peerResults[client.String()]
 				if ok && prs.CurrentVersion != prs.UpdatedVersion && prs.UpdatedVersion != "" {
-					return client.SignalService(serviceRestart, "", dryRun)
+					return client.SignalService(serviceRestart, "", dryRun, &startTime)
 				}
 				return nil
 			}, idx, *client.host)
@@ -542,9 +543,12 @@ func (a adminAPIHandlers) ServiceV2Handler(w http.ResponseWriter, r *http.Reques
 	}
 
 	var objectAPI ObjectLayer
+	var execAt *time.Time
 	switch serviceSig {
 	case serviceRestart:
 		objectAPI, _ = validateAdminReq(ctx, w, r, policy.ServiceRestartAdminAction)
+		t := time.Now().Add(restartUpdateDelay)
+		execAt = &t
 	case serviceStop:
 		objectAPI, _ = validateAdminReq(ctx, w, r, policy.ServiceStopAdminAction)
 	case serviceFreeze, serviceUnFreeze:
@@ -571,7 +575,7 @@ func (a adminAPIHandlers) ServiceV2Handler(w http.ResponseWriter, r *http.Reques
 	}
 
 	if globalIsDistErasure {
-		for _, nerr := range globalNotificationSys.SignalServiceV2(serviceSig, dryRun) {
+		for _, nerr := range globalNotificationSys.SignalServiceV2(serviceSig, dryRun, execAt) {
 			if nerr.Err != nil && process {
 				waitingDrives := map[string]madmin.DiskMetrics{}
 				jerr := json.Unmarshal([]byte(nerr.Err.Error()), &waitingDrives)
@@ -844,9 +848,10 @@ func (a adminAPIHandlers) DataUsageInfoHandler(w http.ResponseWriter, r *http.Re
 }
 
 func lriToLockEntry(l lockRequesterInfo, now time.Time, resource, server string) *madmin.LockEntry {
+	t := time.Unix(0, l.Timestamp)
 	entry := &madmin.LockEntry{
-		Timestamp:  l.Timestamp,
-		Elapsed:    now.Sub(l.Timestamp),
+		Timestamp:  t,
+		Elapsed:    now.Sub(t),
 		Resource:   resource,
 		ServerList: []string{server},
 		Source:     l.Source,
@@ -1031,7 +1036,7 @@ func (a adminAPIHandlers) StartProfilingHandler(w http.ResponseWriter, r *http.R
 	// Start profiling on remote servers.
 	var hostErrs []NotificationPeerErr
 	for _, profiler := range profiles {
-		hostErrs = append(hostErrs, globalNotificationSys.StartProfiling(profiler)...)
+		hostErrs = append(hostErrs, globalNotificationSys.StartProfiling(ctx, profiler)...)
 
 		// Start profiling locally as well.
 		prof, err := startProfiler(profiler)
@@ -1112,7 +1117,11 @@ func (a adminAPIHandlers) ProfileHandler(w http.ResponseWriter, r *http.Request)
 
 	// Start profiling on remote servers.
 	for _, profiler := range profiles {
-		globalNotificationSys.StartProfiling(profiler)
+		// Limit start time to max 10s.
+		ctx, cancel := context.WithTimeout(ctx, 10*time.Second)
+		globalNotificationSys.StartProfiling(ctx, profiler)
+		// StartProfiling blocks, so we can cancel now.
+		cancel()
 
 		// Start profiling locally as well.
 		prof, err := startProfiler(profiler)
@@ -1127,6 +1136,10 @@ func (a adminAPIHandlers) ProfileHandler(w http.ResponseWriter, r *http.Request)
 	for {
 		select {
 		case <-ctx.Done():
+			// Stop remote profiles
+			go globalNotificationSys.DownloadProfilingData(GlobalContext, io.Discard)
+
+			// Stop local
 			globalProfilerMu.Lock()
 			defer globalProfilerMu.Unlock()
 			for k, v := range globalProfiler {
@@ -2182,7 +2195,7 @@ func (a adminAPIHandlers) KMSCreateKeyHandler(w http.ResponseWriter, r *http.Req
 	writeSuccessResponseHeadersOnly(w)
 }
 
-// KMSKeyStatusHandler - GET /minio/admin/v3/kms/status
+// KMSStatusHandler - GET /minio/admin/v3/kms/status
 func (a adminAPIHandlers) KMSStatusHandler(w http.ResponseWriter, r *http.Request) {
 	ctx := r.Context()
 
@@ -2326,6 +2339,7 @@ func getPoolsInfo(ctx context.Context, allDisks []madmin.Disk) (map[int]map[int]
 }
 
 func getServerInfo(ctx context.Context, pools, metrics bool, r *http.Request) madmin.InfoMessage {
+	const operationTimeout = 10 * time.Second
 	ldap := madmin.LDAP{}
 	if globalIAMSys.LDAPConfig.Enabled() {
 		ldapConn, err := globalIAMSys.LDAPConfig.LDAP.Connect()
@@ -2366,7 +2380,9 @@ func getServerInfo(ctx context.Context, pools, metrics bool, r *http.Request) ma
 		mode = madmin.ItemOnline
 
 		// Load data usage
-		dataUsageInfo, err := loadDataUsageFromBackend(ctx, objectAPI)
+		ctx2, cancel := context.WithTimeout(ctx, operationTimeout)
+		dataUsageInfo, err := loadDataUsageFromBackend(ctx2, objectAPI)
+		cancel()
 		if err == nil {
 			buckets = madmin.Buckets{Count: dataUsageInfo.BucketsCount}
 			objects = madmin.Objects{Count: dataUsageInfo.ObjectsTotalCount}
@@ -2400,24 +2416,30 @@ func getServerInfo(ctx context.Context, pools, metrics bool, r *http.Request) ma
 		}
 
 		if pools {
-			poolsInfo, _ = getPoolsInfo(ctx, allDisks)
+			ctx2, cancel := context.WithTimeout(ctx, operationTimeout)
+			poolsInfo, _ = getPoolsInfo(ctx2, allDisks)
+			cancel()
 		}
 	}
 
 	domain := globalDomainNames
 	services := madmin.Services{
-		KMSStatus:     fetchKMSStatus(ctx),
 		LDAP:          ldap,
 		Logger:        log,
 		Audit:         audit,
 		Notifications: notifyTarget,
 	}
+	{
+		ctx2, cancel := context.WithTimeout(ctx, operationTimeout)
+		services.KMSStatus = fetchKMSStatus(ctx2)
+		cancel()
+	}
 
 	return madmin.InfoMessage{
 		Mode:          string(mode),
 		Domain:        domain,
 		Region:        globalSite.Region(),
-		SQSARN:        globalEventNotifier.GetARNList(false),
+		SQSARN:        globalEventNotifier.GetARNList(),
 		DeploymentID:  globalDeploymentID(),
 		Buckets:       buckets,
 		Objects:       objects,
@@ -3045,7 +3067,7 @@ func targetStatus(ctx context.Context, h logger.Target) madmin.Status {
 	return madmin.Status{Status: string(madmin.ItemOffline)}
 }
 
-// fetchLoggerDetails return log info
+// fetchLoggerInfo return log info
 func fetchLoggerInfo(ctx context.Context) ([]madmin.Logger, []madmin.Audit) {
 	var loggerInfo []madmin.Logger
 	var auditloggerInfo []madmin.Audit

cmd/admin-handlers_test.go

@@ -463,6 +463,7 @@ func TestTopLockEntries(t *testing.T) {
 			Owner:      lri.Owner,
 			ID:         lri.UID,
 			Quorum:     lri.Quorum,
+			Timestamp:  time.Unix(0, lri.Timestamp),
 		})
 	}
 

cmd/admin-heal-ops.go

@@ -761,6 +761,15 @@ func (h *healSequence) queueHealTask(source healSource, healType madmin.HealItem
 		return nil
 	}
 
+	countOKDrives := func(drives []madmin.HealDriveInfo) (count int) {
+		for _, drive := range drives {
+			if drive.State == madmin.DriveStateOk {
+				count++
+			}
+		}
+		return count
+	}
+
 	// task queued, now wait for the response.
 	select {
 	case res := <-task.respCh:
@@ -781,6 +790,11 @@ func (h *healSequence) queueHealTask(source healSource, healType madmin.HealItem
 		if res.err != nil {
 			res.result.Detail = res.err.Error()
 		}
+		if res.result.ParityBlocks > 0 && res.result.DataBlocks > 0 && res.result.DataBlocks > res.result.ParityBlocks {
+			if got := countOKDrives(res.result.After.Drives); got < res.result.ParityBlocks {
+				res.result.Detail = fmt.Sprintf("quorum loss - expected %d minimum, got drive states in OK %d", res.result.ParityBlocks, got)
+			}
+		}
 		return h.pushHealResultItem(res.result)
 	case <-h.ctx.Done():
 		return nil

cmd/admin-router.go

@@ -301,8 +301,9 @@ func registerAdminRouter(router *mux.Router, enableConfigOps bool) {
 		// LDAP specific service accounts ops
 		adminRouter.Methods(http.MethodPut).Path(adminVersion + "/idp/ldap/add-service-account").HandlerFunc(adminMiddleware(adminAPI.AddServiceAccountLDAP))
 		adminRouter.Methods(http.MethodGet).Path(adminVersion+"/idp/ldap/list-access-keys").
-			HandlerFunc(adminMiddleware(adminAPI.ListAccessKeysLDAP)).
-			Queries("userDN", "{userDN:.*}", "listType", "{listType:.*}")
+			HandlerFunc(adminMiddleware(adminAPI.ListAccessKeysLDAP)).Queries("userDN", "{userDN:.*}", "listType", "{listType:.*}")
+		adminRouter.Methods(http.MethodGet).Path(adminVersion+"/idp/ldap/list-access-keys-bulk").
+			HandlerFunc(adminMiddleware(adminAPI.ListAccessKeysLDAPBulk)).Queries("listType", "{listType:.*}")
 
 		// LDAP IAM operations
 		adminRouter.Methods(http.MethodGet).Path(adminVersion + "/idp/ldap/policy-entities").HandlerFunc(adminMiddleware(adminAPI.ListLDAPPolicyMappingEntities))
@@ -340,6 +341,9 @@ func registerAdminRouter(router *mux.Router, enableConfigOps bool) {
 		adminRouter.Methods(http.MethodGet).Path(adminVersion + "/list-jobs").HandlerFunc(
 			adminMiddleware(adminAPI.ListBatchJobs))
 
+		adminRouter.Methods(http.MethodGet).Path(adminVersion + "/status-job").HandlerFunc(
+			adminMiddleware(adminAPI.BatchJobStatus))
+
 		adminRouter.Methods(http.MethodGet).Path(adminVersion + "/describe-job").HandlerFunc(
 			adminMiddleware(adminAPI.DescribeBatchJob))
 		adminRouter.Methods(http.MethodDelete).Path(adminVersion + "/cancel-job").HandlerFunc(

cmd/api-errors.go

@@ -28,7 +28,7 @@ import (
 	"strconv"
 	"strings"
 
-	"github.com/Azure/azure-storage-blob-go/azblob"
+	"github.com/Azure/azure-sdk-for-go/sdk/azcore"
 	"github.com/minio/minio/internal/ioutil"
 	"google.golang.org/api/googleapi"
 
@@ -1486,7 +1486,7 @@ var errorCodes = errorCodeMap{
 	},
 	ErrTooManyRequests: {
 		Code:           "TooManyRequests",
-		Description:    "Deadline exceeded while waiting in incoming queue, please reduce your request rate",
+		Description:    "Please reduce your request rate",
 		HTTPStatusCode: http.StatusTooManyRequests,
 	},
 	ErrUnsupportedMetadata: {
@@ -2549,11 +2549,11 @@ func toAPIError(ctx context.Context, err error) APIError {
 			if len(e.Errors) >= 1 {
 				apiErr.Code = e.Errors[0].Reason
 			}
-		case azblob.StorageError:
+		case *azcore.ResponseError:
 			apiErr = APIError{
-				Code:           string(e.ServiceCode()),
+				Code:           e.ErrorCode,
 				Description:    e.Error(),
-				HTTPStatusCode: e.Response().StatusCode,
+				HTTPStatusCode: e.StatusCode,
 			}
 			// Add more other SDK related errors here if any in future.
 		default:
@@ -2592,7 +2592,7 @@ func getAPIError(code APIErrorCode) APIError {
 	return errorCodes.ToAPIErr(ErrInternalError)
 }
 
-// getErrorResponse gets in standard error and resource value and
+// getAPIErrorResponse gets in standard error and resource value and
 // provides a encodable populated response values
 func getAPIErrorResponse(ctx context.Context, err APIError, resource, requestID, hostID string) APIErrorResponse {
 	reqInfo := logger.GetReqInfo(ctx)

cmd/api-response.go

@@ -946,10 +946,11 @@ func writeSuccessResponseHeadersOnly(w http.ResponseWriter) {
 
 // writeErrorResponse writes error headers
 func writeErrorResponse(ctx context.Context, w http.ResponseWriter, err APIError, reqURL *url.URL) {
-	if err.HTTPStatusCode == http.StatusServiceUnavailable {
-		// Set retry-after header to indicate user-agents to retry request after 120secs.
+	switch err.HTTPStatusCode {
+	case http.StatusServiceUnavailable, http.StatusTooManyRequests:
+		// Set retry-after header to indicate user-agents to retry request after 60 seconds.
 		// https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Retry-After
-		w.Header().Set(xhttp.RetryAfter, "120")
+		w.Header().Set(xhttp.RetryAfter, "60")
 	}
 
 	switch err.Code {

cmd/api-router.go

@@ -436,7 +436,7 @@ func registerAPIRouter(router *mux.Router) {
 			Queries("notification", "")
 		// ListenNotification
 		router.Methods(http.MethodGet).
-			HandlerFunc(s3APIMiddleware(api.ListenNotificationHandler, noThrottleS3HFlag)).
+			HandlerFunc(s3APIMiddleware(api.ListenNotificationHandler, noThrottleS3HFlag, traceHdrsS3HFlag)).
 			Queries("events", "{events:.*}")
 		// ResetBucketReplicationStatus - MinIO extension API
 		router.Methods(http.MethodGet).
@@ -456,6 +456,14 @@ func registerAPIRouter(router *mux.Router) {
 		router.Methods(http.MethodGet).
 			HandlerFunc(s3APIMiddleware(api.GetBucketCorsHandler)).
 			Queries("cors", "")
+		// PutBucketCors - this is a dummy call.
+		router.Methods(http.MethodPut).
+			HandlerFunc(s3APIMiddleware(api.PutBucketCorsHandler)).
+			Queries("cors", "")
+		// DeleteBucketCors - this is a dummy call.
+		router.Methods(http.MethodDelete).
+			HandlerFunc(s3APIMiddleware(api.DeleteBucketCorsHandler)).
+			Queries("cors", "")
 		// GetBucketWebsiteHandler - this is a dummy call.
 		router.Methods(http.MethodGet).
 			HandlerFunc(s3APIMiddleware(api.GetBucketWebsiteHandler)).
@@ -472,6 +480,7 @@ func registerAPIRouter(router *mux.Router) {
 		router.Methods(http.MethodGet).
 			HandlerFunc(s3APIMiddleware(api.GetBucketLoggingHandler)).
 			Queries("logging", "")
+
 		// GetBucketTaggingHandler
 		router.Methods(http.MethodGet).
 			HandlerFunc(s3APIMiddleware(api.GetBucketTaggingHandler)).
@@ -615,7 +624,7 @@ func registerAPIRouter(router *mux.Router) {
 
 	// ListenNotification
 	apiRouter.Methods(http.MethodGet).Path(SlashSeparator).
-		HandlerFunc(s3APIMiddleware(api.ListenNotificationHandler, noThrottleS3HFlag)).
+		HandlerFunc(s3APIMiddleware(api.ListenNotificationHandler, noThrottleS3HFlag, traceHdrsS3HFlag)).
 		Queries("events", "{events:.*}")
 
 	// ListBuckets

cmd/background-newdisks-heal-ops.go

@@ -72,10 +72,12 @@ type healingTracker struct {
 
 	// Numbers when current bucket started healing,
 	// for resuming with correct numbers.
-	ResumeItemsHealed uint64 `json:"-"`
-	ResumeItemsFailed uint64 `json:"-"`
-	ResumeBytesDone   uint64 `json:"-"`
-	ResumeBytesFailed uint64 `json:"-"`
+	ResumeItemsHealed  uint64 `json:"-"`
+	ResumeItemsFailed  uint64 `json:"-"`
+	ResumeItemsSkipped uint64 `json:"-"`
+	ResumeBytesDone    uint64 `json:"-"`
+	ResumeBytesFailed  uint64 `json:"-"`
+	ResumeBytesSkipped uint64 `json:"-"`
 
 	// Filled on startup/restarts.
 	QueuedBuckets []string
@@ -88,6 +90,11 @@ type healingTracker struct {
 
 	ItemsSkipped uint64
 	BytesSkipped uint64
+
+	RetryAttempts uint64
+
+	Finished bool // finished healing, whether with errors or not
+
 	// Add future tracking capabilities
 	// Be sure that they are included in toHealingDisk
 }
@@ -141,6 +148,26 @@ func initHealingTracker(disk StorageAPI, healID string) *healingTracker {
 	return h
 }
 
+func (h *healingTracker) resetHealing() {
+	h.mu.Lock()
+	defer h.mu.Unlock()
+
+	h.ItemsHealed = 0
+	h.ItemsFailed = 0
+	h.BytesDone = 0
+	h.BytesFailed = 0
+	h.ResumeItemsHealed = 0
+	h.ResumeItemsFailed = 0
+	h.ResumeBytesDone = 0
+	h.ResumeBytesFailed = 0
+	h.ItemsSkipped = 0
+	h.BytesSkipped = 0
+
+	h.HealedBuckets = nil
+	h.Object = ""
+	h.Bucket = ""
+}
+
 func (h *healingTracker) getLastUpdate() time.Time {
 	h.mu.RLock()
 	defer h.mu.RUnlock()
@@ -260,8 +287,10 @@ func (h *healingTracker) resume() {
 
 	h.ItemsHealed = h.ResumeItemsHealed
 	h.ItemsFailed = h.ResumeItemsFailed
+	h.ItemsSkipped = h.ResumeItemsSkipped
 	h.BytesDone = h.ResumeBytesDone
 	h.BytesFailed = h.ResumeBytesFailed
+	h.BytesSkipped = h.ResumeBytesSkipped
 }
 
 // bucketDone should be called when a bucket is done healing.
@@ -272,8 +301,10 @@ func (h *healingTracker) bucketDone(bucket string) {
 
 	h.ResumeItemsHealed = h.ItemsHealed
 	h.ResumeItemsFailed = h.ItemsFailed
+	h.ResumeItemsSkipped = h.ItemsSkipped
 	h.ResumeBytesDone = h.BytesDone
 	h.ResumeBytesFailed = h.BytesFailed
+	h.ResumeBytesSkipped = h.BytesSkipped
 	h.HealedBuckets = append(h.HealedBuckets, bucket)
 	for i, b := range h.QueuedBuckets {
 		if b == bucket {
@@ -322,6 +353,7 @@ func (h *healingTracker) toHealingDisk() madmin.HealingDisk {
 		PoolIndex:         h.PoolIndex,
 		SetIndex:          h.SetIndex,
 		DiskIndex:         h.DiskIndex,
+		Finished:          h.Finished,
 		Path:              h.Path,
 		Started:           h.Started.UTC(),
 		LastUpdate:        h.LastUpdate.UTC(),
@@ -337,6 +369,7 @@ func (h *healingTracker) toHealingDisk() madmin.HealingDisk {
 		Object:            h.Object,
 		QueuedBuckets:     h.QueuedBuckets,
 		HealedBuckets:     h.HealedBuckets,
+		RetryAttempts:     h.RetryAttempts,
 
 		ObjectsHealed: h.ItemsHealed, // Deprecated July 2021
 		ObjectsFailed: h.ItemsFailed, // Deprecated July 2021
@@ -351,24 +384,26 @@ func initAutoHeal(ctx context.Context, objAPI ObjectLayer) {
 	}
 
 	initBackgroundHealing(ctx, objAPI) // start quick background healing
-
-	if env.Get("_MINIO_AUTO_DRIVE_HEALING", config.EnableOn) == config.EnableOn || env.Get("_MINIO_AUTO_DISK_HEALING", config.EnableOn) == config.EnableOn {
+	if env.Get("_MINIO_AUTO_DRIVE_HEALING", config.EnableOn) == config.EnableOn {
 		globalBackgroundHealState.pushHealLocalDisks(getLocalDisksToHeal()...)
 		go monitorLocalDisksAndHeal(ctx, z)
 	}
+
+	go globalMRFState.startMRFPersistence()
+	go globalMRFState.healRoutine(z)
 }
 
 func getLocalDisksToHeal() (disksToHeal Endpoints) {
 	globalLocalDrivesMu.RLock()
-	localDrives := cloneDrives(globalLocalDrives)
+	localDrives := cloneDrives(globalLocalDrivesMap)
 	globalLocalDrivesMu.RUnlock()
 	for _, disk := range localDrives {
-		_, err := disk.GetDiskID()
+		_, err := disk.DiskInfo(context.Background(), DiskInfoOptions{})
 		if errors.Is(err, errUnformattedDisk) {
 			disksToHeal = append(disksToHeal, disk.Endpoint())
 			continue
 		}
-		if disk.Healing() != nil {
+		if h := disk.Healing(); h != nil && !h.Finished {
 			disksToHeal = append(disksToHeal, disk.Endpoint())
 		}
 	}
@@ -382,6 +417,8 @@ func getLocalDisksToHeal() (disksToHeal Endpoints) {
 
 var newDiskHealingTimeout = newDynamicTimeout(30*time.Second, 10*time.Second)
 
+var errRetryHealing = errors.New("some items failed to heal, we will retry healing this drive again")
+
 func healFreshDisk(ctx context.Context, z *erasureServerPools, endpoint Endpoint) error {
 	poolIdx, setIdx := endpoint.PoolIdx, endpoint.SetIdx
 	disk := getStorageViaEndpoint(endpoint)
@@ -389,6 +426,17 @@ func healFreshDisk(ctx context.Context, z *erasureServerPools, endpoint Endpoint
 		return fmt.Errorf("Unexpected error disk must be initialized by now after formatting: %s", endpoint)
 	}
 
+	_, err := disk.DiskInfo(ctx, DiskInfoOptions{})
+	if err != nil {
+		if errors.Is(err, errDriveIsRoot) {
+			// This is a root drive, ignore and move on
+			return nil
+		}
+		if !errors.Is(err, errUnformattedDisk) {
+			return err
+		}
+	}
+
 	// Prevent parallel erasure set healing
 	locker := z.NewNSLock(minioMetaBucket, fmt.Sprintf("new-drive-healing/%d/%d", poolIdx, setIdx))
 	lkctx, err := locker.GetLock(ctx, newDiskHealingTimeout)
@@ -451,8 +499,30 @@ func healFreshDisk(ctx context.Context, z *erasureServerPools, endpoint Endpoint
 		return err
 	}
 
-	healingLogEvent(ctx, "Healing of drive '%s' is finished (healed: %d, skipped: %d, failed: %d).", disk, tracker.ItemsHealed, tracker.ItemsSkipped, tracker.ItemsFailed)
+	// if objects have failed healing, we attempt a retry to heal the drive upto 3 times before giving up.
+	if tracker.ItemsFailed > 0 && tracker.RetryAttempts < 4 {
+		tracker.RetryAttempts++
 
+		healingLogEvent(ctx, "Healing of drive '%s' is incomplete, retrying %s time (healed: %d, skipped: %d, failed: %d).", disk,
+			humanize.Ordinal(int(tracker.RetryAttempts)), tracker.ItemsHealed, tracker.ItemsSkipped, tracker.ItemsFailed)
+
+		tracker.resetHealing()
+		bugLogIf(ctx, tracker.update(ctx))
+
+		return errRetryHealing
+	}
+
+	if tracker.ItemsFailed > 0 {
+		healingLogEvent(ctx, "Healing of drive '%s' is incomplete, retried %d times (healed: %d, skipped: %d, failed: %d).", disk,
+			tracker.RetryAttempts, tracker.ItemsHealed, tracker.ItemsSkipped, tracker.ItemsFailed)
+	} else {
+		if tracker.RetryAttempts > 0 {
+			healingLogEvent(ctx, "Healing of drive '%s' is complete, retried %d times (healed: %d, skipped: %d).", disk,
+				tracker.RetryAttempts-1, tracker.ItemsHealed, tracker.ItemsSkipped)
+		} else {
+			healingLogEvent(ctx, "Healing of drive '%s' is finished (healed: %d, skipped: %d).", disk, tracker.ItemsHealed, tracker.ItemsSkipped)
+		}
+	}
 	if serverDebugLog {
 		tracker.printTo(os.Stdout)
 		fmt.Printf("\n")
@@ -482,7 +552,8 @@ func healFreshDisk(ctx context.Context, z *erasureServerPools, endpoint Endpoint
 			continue
 		}
 		if t.HealID == tracker.HealID {
-			t.delete(ctx)
+			t.Finished = true
+			t.update(ctx)
 		}
 	}
 
@@ -524,7 +595,7 @@ func monitorLocalDisksAndHeal(ctx context.Context, z *erasureServerPools) {
 					if err := healFreshDisk(ctx, z, disk); err != nil {
 						globalBackgroundHealState.setDiskHealingStatus(disk, false)
 						timedout := OperationTimedOut{}
-						if !errors.Is(err, context.Canceled) && !errors.As(err, &timedout) {
+						if !errors.Is(err, context.Canceled) && !errors.As(err, &timedout) && !errors.Is(err, errRetryHealing) {
 							printEndpointError(disk, err, false)
 						}
 						return

cmd/background-newdisks-heal-ops_gen.go

@@ -132,6 +132,12 @@ func (z *healingTracker) DecodeMsg(dc *msgp.Reader) (err error) {
 				err = msgp.WrapError(err, "ResumeItemsFailed")
 				return
 			}
+		case "ResumeItemsSkipped":
+			z.ResumeItemsSkipped, err = dc.ReadUint64()
+			if err != nil {
+				err = msgp.WrapError(err, "ResumeItemsSkipped")
+				return
+			}
 		case "ResumeBytesDone":
 			z.ResumeBytesDone, err = dc.ReadUint64()
 			if err != nil {
@@ -144,6 +150,12 @@ func (z *healingTracker) DecodeMsg(dc *msgp.Reader) (err error) {
 				err = msgp.WrapError(err, "ResumeBytesFailed")
 				return
 			}
+		case "ResumeBytesSkipped":
+			z.ResumeBytesSkipped, err = dc.ReadUint64()
+			if err != nil {
+				err = msgp.WrapError(err, "ResumeBytesSkipped")
+				return
+			}
 		case "QueuedBuckets":
 			var zb0002 uint32
 			zb0002, err = dc.ReadArrayHeader()
@@ -200,6 +212,18 @@ func (z *healingTracker) DecodeMsg(dc *msgp.Reader) (err error) {
 				err = msgp.WrapError(err, "BytesSkipped")
 				return
 			}
+		case "RetryAttempts":
+			z.RetryAttempts, err = dc.ReadUint64()
+			if err != nil {
+				err = msgp.WrapError(err, "RetryAttempts")
+				return
+			}
+		case "Finished":
+			z.Finished, err = dc.ReadBool()
+			if err != nil {
+				err = msgp.WrapError(err, "Finished")
+				return
+			}
 		default:
 			err = dc.Skip()
 			if err != nil {
@@ -213,9 +237,9 @@ func (z *healingTracker) DecodeMsg(dc *msgp.Reader) (err error) {
 
 // EncodeMsg implements msgp.Encodable
 func (z *healingTracker) EncodeMsg(en *msgp.Writer) (err error) {
-	// map header, size 25
+	// map header, size 29
 	// write "ID"
-	err = en.Append(0xde, 0x0, 0x19, 0xa2, 0x49, 0x44)
+	err = en.Append(0xde, 0x0, 0x1d, 0xa2, 0x49, 0x44)
 	if err != nil {
 		return
 	}
@@ -394,6 +418,16 @@ func (z *healingTracker) EncodeMsg(en *msgp.Writer) (err error) {
 		err = msgp.WrapError(err, "ResumeItemsFailed")
 		return
 	}
+	// write "ResumeItemsSkipped"
+	err = en.Append(0xb2, 0x52, 0x65, 0x73, 0x75, 0x6d, 0x65, 0x49, 0x74, 0x65, 0x6d, 0x73, 0x53, 0x6b, 0x69, 0x70, 0x70, 0x65, 0x64)
+	if err != nil {
+		return
+	}
+	err = en.WriteUint64(z.ResumeItemsSkipped)
+	if err != nil {
+		err = msgp.WrapError(err, "ResumeItemsSkipped")
+		return
+	}
 	// write "ResumeBytesDone"
 	err = en.Append(0xaf, 0x52, 0x65, 0x73, 0x75, 0x6d, 0x65, 0x42, 0x79, 0x74, 0x65, 0x73, 0x44, 0x6f, 0x6e, 0x65)
 	if err != nil {
@@ -414,6 +448,16 @@ func (z *healingTracker) EncodeMsg(en *msgp.Writer) (err error) {
 		err = msgp.WrapError(err, "ResumeBytesFailed")
 		return
 	}
+	// write "ResumeBytesSkipped"
+	err = en.Append(0xb2, 0x52, 0x65, 0x73, 0x75, 0x6d, 0x65, 0x42, 0x79, 0x74, 0x65, 0x73, 0x53, 0x6b, 0x69, 0x70, 0x70, 0x65, 0x64)
+	if err != nil {
+		return
+	}
+	err = en.WriteUint64(z.ResumeBytesSkipped)
+	if err != nil {
+		err = msgp.WrapError(err, "ResumeBytesSkipped")
+		return
+	}
 	// write "QueuedBuckets"
 	err = en.Append(0xad, 0x51, 0x75, 0x65, 0x75, 0x65, 0x64, 0x42, 0x75, 0x63, 0x6b, 0x65, 0x74, 0x73)
 	if err != nil {
@@ -478,15 +522,35 @@ func (z *healingTracker) EncodeMsg(en *msgp.Writer) (err error) {
 		err = msgp.WrapError(err, "BytesSkipped")
 		return
 	}
+	// write "RetryAttempts"
+	err = en.Append(0xad, 0x52, 0x65, 0x74, 0x72, 0x79, 0x41, 0x74, 0x74, 0x65, 0x6d, 0x70, 0x74, 0x73)
+	if err != nil {
+		return
+	}
+	err = en.WriteUint64(z.RetryAttempts)
+	if err != nil {
+		err = msgp.WrapError(err, "RetryAttempts")
+		return
+	}
+	// write "Finished"
+	err = en.Append(0xa8, 0x46, 0x69, 0x6e, 0x69, 0x73, 0x68, 0x65, 0x64)
+	if err != nil {
+		return
+	}
+	err = en.WriteBool(z.Finished)
+	if err != nil {
+		err = msgp.WrapError(err, "Finished")
+		return
+	}
 	return
 }
 
 // MarshalMsg implements msgp.Marshaler
 func (z *healingTracker) MarshalMsg(b []byte) (o []byte, err error) {
 	o = msgp.Require(b, z.Msgsize())
-	// map header, size 25
+	// map header, size 29
 	// string "ID"
-	o = append(o, 0xde, 0x0, 0x19, 0xa2, 0x49, 0x44)
+	o = append(o, 0xde, 0x0, 0x1d, 0xa2, 0x49, 0x44)
 	o = msgp.AppendString(o, z.ID)
 	// string "PoolIndex"
 	o = append(o, 0xa9, 0x50, 0x6f, 0x6f, 0x6c, 0x49, 0x6e, 0x64, 0x65, 0x78)
@@ -539,12 +603,18 @@ func (z *healingTracker) MarshalMsg(b []byte) (o []byte, err error) {
 	// string "ResumeItemsFailed"
 	o = append(o, 0xb1, 0x52, 0x65, 0x73, 0x75, 0x6d, 0x65, 0x49, 0x74, 0x65, 0x6d, 0x73, 0x46, 0x61, 0x69, 0x6c, 0x65, 0x64)
 	o = msgp.AppendUint64(o, z.ResumeItemsFailed)
+	// string "ResumeItemsSkipped"
+	o = append(o, 0xb2, 0x52, 0x65, 0x73, 0x75, 0x6d, 0x65, 0x49, 0x74, 0x65, 0x6d, 0x73, 0x53, 0x6b, 0x69, 0x70, 0x70, 0x65, 0x64)
+	o = msgp.AppendUint64(o, z.ResumeItemsSkipped)
 	// string "ResumeBytesDone"
 	o = append(o, 0xaf, 0x52, 0x65, 0x73, 0x75, 0x6d, 0x65, 0x42, 0x79, 0x74, 0x65, 0x73, 0x44, 0x6f, 0x6e, 0x65)
 	o = msgp.AppendUint64(o, z.ResumeBytesDone)
 	// string "ResumeBytesFailed"
 	o = append(o, 0xb1, 0x52, 0x65, 0x73, 0x75, 0x6d, 0x65, 0x42, 0x79, 0x74, 0x65, 0x73, 0x46, 0x61, 0x69, 0x6c, 0x65, 0x64)
 	o = msgp.AppendUint64(o, z.ResumeBytesFailed)
+	// string "ResumeBytesSkipped"
+	o = append(o, 0xb2, 0x52, 0x65, 0x73, 0x75, 0x6d, 0x65, 0x42, 0x79, 0x74, 0x65, 0x73, 0x53, 0x6b, 0x69, 0x70, 0x70, 0x65, 0x64)
+	o = msgp.AppendUint64(o, z.ResumeBytesSkipped)
 	// string "QueuedBuckets"
 	o = append(o, 0xad, 0x51, 0x75, 0x65, 0x75, 0x65, 0x64, 0x42, 0x75, 0x63, 0x6b, 0x65, 0x74, 0x73)
 	o = msgp.AppendArrayHeader(o, uint32(len(z.QueuedBuckets)))
@@ -566,6 +636,12 @@ func (z *healingTracker) MarshalMsg(b []byte) (o []byte, err error) {
 	// string "BytesSkipped"
 	o = append(o, 0xac, 0x42, 0x79, 0x74, 0x65, 0x73, 0x53, 0x6b, 0x69, 0x70, 0x70, 0x65, 0x64)
 	o = msgp.AppendUint64(o, z.BytesSkipped)
+	// string "RetryAttempts"
+	o = append(o, 0xad, 0x52, 0x65, 0x74, 0x72, 0x79, 0x41, 0x74, 0x74, 0x65, 0x6d, 0x70, 0x74, 0x73)
+	o = msgp.AppendUint64(o, z.RetryAttempts)
+	// string "Finished"
+	o = append(o, 0xa8, 0x46, 0x69, 0x6e, 0x69, 0x73, 0x68, 0x65, 0x64)
+	o = msgp.AppendBool(o, z.Finished)
 	return
 }
 
@@ -695,6 +771,12 @@ func (z *healingTracker) UnmarshalMsg(bts []byte) (o []byte, err error) {
 				err = msgp.WrapError(err, "ResumeItemsFailed")
 				return
 			}
+		case "ResumeItemsSkipped":
+			z.ResumeItemsSkipped, bts, err = msgp.ReadUint64Bytes(bts)
+			if err != nil {
+				err = msgp.WrapError(err, "ResumeItemsSkipped")
+				return
+			}
 		case "ResumeBytesDone":
 			z.ResumeBytesDone, bts, err = msgp.ReadUint64Bytes(bts)
 			if err != nil {
@@ -707,6 +789,12 @@ func (z *healingTracker) UnmarshalMsg(bts []byte) (o []byte, err error) {
 				err = msgp.WrapError(err, "ResumeBytesFailed")
 				return
 			}
+		case "ResumeBytesSkipped":
+			z.ResumeBytesSkipped, bts, err = msgp.ReadUint64Bytes(bts)
+			if err != nil {
+				err = msgp.WrapError(err, "ResumeBytesSkipped")
+				return
+			}
 		case "QueuedBuckets":
 			var zb0002 uint32
 			zb0002, bts, err = msgp.ReadArrayHeaderBytes(bts)
@@ -763,6 +851,18 @@ func (z *healingTracker) UnmarshalMsg(bts []byte) (o []byte, err error) {
 				err = msgp.WrapError(err, "BytesSkipped")
 				return
 			}
+		case "RetryAttempts":
+			z.RetryAttempts, bts, err = msgp.ReadUint64Bytes(bts)
+			if err != nil {
+				err = msgp.WrapError(err, "RetryAttempts")
+				return
+			}
+		case "Finished":
+			z.Finished, bts, err = msgp.ReadBoolBytes(bts)
+			if err != nil {
+				err = msgp.WrapError(err, "Finished")
+				return
+			}
 		default:
 			bts, err = msgp.Skip(bts)
 			if err != nil {
@@ -777,7 +877,7 @@ func (z *healingTracker) UnmarshalMsg(bts []byte) (o []byte, err error) {
 
 // Msgsize returns an upper bound estimate of the number of bytes occupied by the serialized message
 func (z *healingTracker) Msgsize() (s int) {
-	s = 3 + 3 + msgp.StringPrefixSize + len(z.ID) + 10 + msgp.IntSize + 9 + msgp.IntSize + 10 + msgp.IntSize + 5 + msgp.StringPrefixSize + len(z.Path) + 9 + msgp.StringPrefixSize + len(z.Endpoint) + 8 + msgp.TimeSize + 11 + msgp.TimeSize + 18 + msgp.Uint64Size + 17 + msgp.Uint64Size + 12 + msgp.Uint64Size + 12 + msgp.Uint64Size + 10 + msgp.Uint64Size + 12 + msgp.Uint64Size + 7 + msgp.StringPrefixSize + len(z.Bucket) + 7 + msgp.StringPrefixSize + len(z.Object) + 18 + msgp.Uint64Size + 18 + msgp.Uint64Size + 16 + msgp.Uint64Size + 18 + msgp.Uint64Size + 14 + msgp.ArrayHeaderSize
+	s = 3 + 3 + msgp.StringPrefixSize + len(z.ID) + 10 + msgp.IntSize + 9 + msgp.IntSize + 10 + msgp.IntSize + 5 + msgp.StringPrefixSize + len(z.Path) + 9 + msgp.StringPrefixSize + len(z.Endpoint) + 8 + msgp.TimeSize + 11 + msgp.TimeSize + 18 + msgp.Uint64Size + 17 + msgp.Uint64Size + 12 + msgp.Uint64Size + 12 + msgp.Uint64Size + 10 + msgp.Uint64Size + 12 + msgp.Uint64Size + 7 + msgp.StringPrefixSize + len(z.Bucket) + 7 + msgp.StringPrefixSize + len(z.Object) + 18 + msgp.Uint64Size + 18 + msgp.Uint64Size + 19 + msgp.Uint64Size + 16 + msgp.Uint64Size + 18 + msgp.Uint64Size + 19 + msgp.Uint64Size + 14 + msgp.ArrayHeaderSize
 	for za0001 := range z.QueuedBuckets {
 		s += msgp.StringPrefixSize + len(z.QueuedBuckets[za0001])
 	}
@@ -785,6 +885,6 @@ func (z *healingTracker) Msgsize() (s int) {
 	for za0002 := range z.HealedBuckets {
 		s += msgp.StringPrefixSize + len(z.HealedBuckets[za0002])
 	}
-	s += 7 + msgp.StringPrefixSize + len(z.HealID) + 13 + msgp.Uint64Size + 13 + msgp.Uint64Size
+	s += 7 + msgp.StringPrefixSize + len(z.HealID) + 13 + msgp.Uint64Size + 13 + msgp.Uint64Size + 14 + msgp.Uint64Size + 9 + msgp.BoolSize
 	return
 }

cmd/batch-expire.go

@@ -27,6 +27,7 @@ import (
 	"net/http"
 	"runtime"
 	"strconv"
+	"strings"
 	"time"
 
 	"github.com/minio/minio-go/v7/pkg/tags"
@@ -36,6 +37,7 @@ import (
 	"github.com/minio/pkg/v3/env"
 	"github.com/minio/pkg/v3/wildcard"
 	"github.com/minio/pkg/v3/workers"
+	"github.com/minio/pkg/v3/xtime"
 	"gopkg.in/yaml.v3"
 )
 
@@ -116,7 +118,7 @@ func (p BatchJobExpirePurge) Validate() error {
 // BatchJobExpireFilter holds all the filters currently supported for batch replication
 type BatchJobExpireFilter struct {
 	line, col     int
-	OlderThan     time.Duration       `yaml:"olderThan,omitempty" json:"olderThan"`
+	OlderThan     xtime.Duration      `yaml:"olderThan,omitempty" json:"olderThan"`
 	CreatedBefore *time.Time          `yaml:"createdBefore,omitempty" json:"createdBefore"`
 	Tags          []BatchJobKV        `yaml:"tags,omitempty" json:"tags"`
 	Metadata      []BatchJobKV        `yaml:"metadata,omitempty" json:"metadata"`
@@ -162,7 +164,7 @@ func (ef BatchJobExpireFilter) Matches(obj ObjectInfo, now time.Time) bool {
 	if len(ef.Name) > 0 && !wildcard.Match(ef.Name, obj.Name) {
 		return false
 	}
-	if ef.OlderThan > 0 && now.Sub(obj.ModTime) <= ef.OlderThan {
+	if ef.OlderThan > 0 && now.Sub(obj.ModTime) <= ef.OlderThan.D() {
 		return false
 	}
 
@@ -280,7 +282,7 @@ type BatchJobExpire struct {
 	line, col       int
 	APIVersion      string                 `yaml:"apiVersion" json:"apiVersion"`
 	Bucket          string                 `yaml:"bucket" json:"bucket"`
-	Prefix          string                 `yaml:"prefix" json:"prefix"`
+	Prefix          BatchJobPrefix         `yaml:"prefix" json:"prefix"`
 	NotificationCfg BatchJobNotification   `yaml:"notify" json:"notify"`
 	Retry           BatchJobRetry          `yaml:"retry" json:"retry"`
 	Rules           []BatchJobExpireFilter `yaml:"rules" json:"rules"`
@@ -340,8 +342,24 @@ func (r *BatchJobExpire) Expire(ctx context.Context, api ObjectLayer, vc *versio
 		PrefixEnabledFn:  vc.PrefixEnabled,
 		VersionSuspended: vc.Suspended(),
 	}
-	_, errs := api.DeleteObjects(ctx, r.Bucket, objsToDel, opts)
-	return errs
+
+	allErrs := make([]error, 0, len(objsToDel))
+
+	for {
+		count := len(objsToDel)
+		if count == 0 {
+			break
+		}
+		if count > maxDeleteList {
+			count = maxDeleteList
+		}
+		_, errs := api.DeleteObjects(ctx, r.Bucket, objsToDel[:count], opts)
+		allErrs = append(allErrs, errs...)
+		// Next batch of deletion
+		objsToDel = objsToDel[count:]
+	}
+
+	return allErrs
 }
 
 const (
@@ -514,7 +532,7 @@ func (r *BatchJobExpire) Start(ctx context.Context, api ObjectLayer, job BatchJo
 		JobType:   string(job.Type()),
 		StartTime: job.Started,
 	}
-	if err := ri.load(ctx, api, job); err != nil {
+	if err := ri.loadOrInit(ctx, api, job); err != nil {
 		return err
 	}
 
@@ -534,40 +552,54 @@ func (r *BatchJobExpire) Start(ctx context.Context, api ObjectLayer, job BatchJo
 		return err
 	}
 
-	ctx, cancel := context.WithCancel(ctx)
-	defer cancel()
+	ctx, cancelCause := context.WithCancelCause(ctx)
+	defer cancelCause(nil)
 
 	results := make(chan itemOrErr[ObjectInfo], workerSize)
-	if err := api.Walk(ctx, r.Bucket, r.Prefix, results, WalkOptions{
-		Marker:       lastObject,
-		LatestOnly:   false, // we need to visit all versions of the object to implement purge: retainVersions
-		VersionsSort: WalkVersionsSortDesc,
-	}); err != nil {
-		// Do not need to retry if we can't list objects on source.
-		return err
-	}
+	go func() {
+		for _, prefix := range r.Prefix.F() {
+			prefixResultCh := make(chan itemOrErr[ObjectInfo], workerSize)
+			err := api.Walk(ctx, r.Bucket, strings.TrimSpace(prefix), prefixResultCh, WalkOptions{
+				Marker:       lastObject,
+				LatestOnly:   false, // we need to visit all versions of the object to implement purge: retainVersions
+				VersionsSort: WalkVersionsSortDesc,
+			})
+			if err != nil {
+				cancelCause(err)
+				xioutil.SafeClose(results)
+				return
+			}
+			for result := range prefixResultCh {
+				results <- result
+			}
+		}
+		xioutil.SafeClose(results)
+	}()
 
 	// Goroutine to periodically save batch-expire job's in-memory state
 	saverQuitCh := make(chan struct{})
 	go func() {
 		saveTicker := time.NewTicker(10 * time.Second)
 		defer saveTicker.Stop()
-		for {
+		quit := false
+		after := time.Minute
+		for !quit {
 			select {
 			case <-saveTicker.C:
-				// persist in-memory state to disk after every 10secs.
-				batchLogIf(ctx, ri.updateAfter(ctx, api, 10*time.Second, job))
-
 			case <-ctx.Done():
-				// persist in-memory state immediately before exiting due to context cancellation.
-				batchLogIf(ctx, ri.updateAfter(ctx, api, 0, job))
-				return
-
+				quit = true
 			case <-saverQuitCh:
-				// persist in-memory state immediately to disk.
-				batchLogIf(ctx, ri.updateAfter(ctx, api, 0, job))
-				return
+				quit = true
 			}
+
+			if quit {
+				// save immediately if we are quitting
+				after = 0
+			}
+
+			ctx, cancel := context.WithTimeout(GlobalContext, 30*time.Second) // independent context
+			batchLogIf(ctx, ri.updateAfter(ctx, api, after, job))
+			cancel()
 		}
 	}()
 
@@ -584,7 +616,7 @@ func (r *BatchJobExpire) Start(ctx context.Context, api ObjectLayer, job BatchJo
 		versionsCount int
 		toDel         []expireObjInfo
 	)
-	failed := true
+	failed := false
 	for result := range results {
 		if result.Err != nil {
 			failed = true
@@ -653,6 +685,10 @@ func (r *BatchJobExpire) Start(ctx context.Context, api ObjectLayer, job BatchJo
 			ObjectInfo: result.Item,
 		})
 	}
+	if context.Cause(ctx) != nil {
+		xioutil.SafeClose(expireCh)
+		return context.Cause(ctx)
+	}
 	// Send any remaining objects downstream
 	if len(toDel) > 0 {
 		select {

cmd/batch-expire_gen.go

@@ -39,7 +39,7 @@ func (z *BatchJobExpire) DecodeMsg(dc *msgp.Reader) (err error) {
 				return
 			}
 		case "Prefix":
-			z.Prefix, err = dc.ReadString()
+			err = z.Prefix.DecodeMsg(dc)
 			if err != nil {
 				err = msgp.WrapError(err, "Prefix")
 				return
@@ -114,7 +114,7 @@ func (z *BatchJobExpire) EncodeMsg(en *msgp.Writer) (err error) {
 	if err != nil {
 		return
 	}
-	err = en.WriteString(z.Prefix)
+	err = z.Prefix.EncodeMsg(en)
 	if err != nil {
 		err = msgp.WrapError(err, "Prefix")
 		return
@@ -171,7 +171,11 @@ func (z *BatchJobExpire) MarshalMsg(b []byte) (o []byte, err error) {
 	o = msgp.AppendString(o, z.Bucket)
 	// string "Prefix"
 	o = append(o, 0xa6, 0x50, 0x72, 0x65, 0x66, 0x69, 0x78)
-	o = msgp.AppendString(o, z.Prefix)
+	o, err = z.Prefix.MarshalMsg(o)
+	if err != nil {
+		err = msgp.WrapError(err, "Prefix")
+		return
+	}
 	// string "NotificationCfg"
 	o = append(o, 0xaf, 0x4e, 0x6f, 0x74, 0x69, 0x66, 0x69, 0x63, 0x61, 0x74, 0x69, 0x6f, 0x6e, 0x43, 0x66, 0x67)
 	o, err = z.NotificationCfg.MarshalMsg(o)
@@ -230,7 +234,7 @@ func (z *BatchJobExpire) UnmarshalMsg(bts []byte) (o []byte, err error) {
 				return
 			}
 		case "Prefix":
-			z.Prefix, bts, err = msgp.ReadStringBytes(bts)
+			bts, err = z.Prefix.UnmarshalMsg(bts)
 			if err != nil {
 				err = msgp.WrapError(err, "Prefix")
 				return
@@ -280,7 +284,7 @@ func (z *BatchJobExpire) UnmarshalMsg(bts []byte) (o []byte, err error) {
 
 // Msgsize returns an upper bound estimate of the number of bytes occupied by the serialized message
 func (z *BatchJobExpire) Msgsize() (s int) {
-	s = 1 + 11 + msgp.StringPrefixSize + len(z.APIVersion) + 7 + msgp.StringPrefixSize + len(z.Bucket) + 7 + msgp.StringPrefixSize + len(z.Prefix) + 16 + z.NotificationCfg.Msgsize() + 6 + z.Retry.Msgsize() + 6 + msgp.ArrayHeaderSize
+	s = 1 + 11 + msgp.StringPrefixSize + len(z.APIVersion) + 7 + msgp.StringPrefixSize + len(z.Bucket) + 7 + z.Prefix.Msgsize() + 16 + z.NotificationCfg.Msgsize() + 6 + z.Retry.Msgsize() + 6 + msgp.ArrayHeaderSize
 	for za0001 := range z.Rules {
 		s += z.Rules[za0001].Msgsize()
 	}
@@ -306,7 +310,7 @@ func (z *BatchJobExpireFilter) DecodeMsg(dc *msgp.Reader) (err error) {
 		}
 		switch msgp.UnsafeString(field) {
 		case "OlderThan":
-			z.OlderThan, err = dc.ReadDuration()
+			err = z.OlderThan.DecodeMsg(dc)
 			if err != nil {
 				err = msgp.WrapError(err, "OlderThan")
 				return
@@ -433,7 +437,7 @@ func (z *BatchJobExpireFilter) EncodeMsg(en *msgp.Writer) (err error) {
 	if err != nil {
 		return
 	}
-	err = en.WriteDuration(z.OlderThan)
+	err = z.OlderThan.EncodeMsg(en)
 	if err != nil {
 		err = msgp.WrapError(err, "OlderThan")
 		return
@@ -544,7 +548,11 @@ func (z *BatchJobExpireFilter) MarshalMsg(b []byte) (o []byte, err error) {
 	// map header, size 8
 	// string "OlderThan"
 	o = append(o, 0x88, 0xa9, 0x4f, 0x6c, 0x64, 0x65, 0x72, 0x54, 0x68, 0x61, 0x6e)
-	o = msgp.AppendDuration(o, z.OlderThan)
+	o, err = z.OlderThan.MarshalMsg(o)
+	if err != nil {
+		err = msgp.WrapError(err, "OlderThan")
+		return
+	}
 	// string "CreatedBefore"
 	o = append(o, 0xad, 0x43, 0x72, 0x65, 0x61, 0x74, 0x65, 0x64, 0x42, 0x65, 0x66, 0x6f, 0x72, 0x65)
 	if z.CreatedBefore == nil {
@@ -613,7 +621,7 @@ func (z *BatchJobExpireFilter) UnmarshalMsg(bts []byte) (o []byte, err error) {
 		}
 		switch msgp.UnsafeString(field) {
 		case "OlderThan":
-			z.OlderThan, bts, err = msgp.ReadDurationBytes(bts)
+			bts, err = z.OlderThan.UnmarshalMsg(bts)
 			if err != nil {
 				err = msgp.WrapError(err, "OlderThan")
 				return
@@ -734,7 +742,7 @@ func (z *BatchJobExpireFilter) UnmarshalMsg(bts []byte) (o []byte, err error) {
 
 // Msgsize returns an upper bound estimate of the number of bytes occupied by the serialized message
 func (z *BatchJobExpireFilter) Msgsize() (s int) {
-	s = 1 + 10 + msgp.DurationSize + 14
+	s = 1 + 10 + z.OlderThan.Msgsize() + 14
 	if z.CreatedBefore == nil {
 		s += msgp.NilSize
 	} else {

cmd/batch-expire_test.go

@@ -18,9 +18,10 @@
 package cmd
 
 import (
+	"slices"
 	"testing"
 
-	"gopkg.in/yaml.v2"
+	"gopkg.in/yaml.v3"
 )
 
 func TestParseBatchJobExpire(t *testing.T) {
@@ -32,7 +33,7 @@ expire: # Expire objects that match a condition
   rules:
     - type: object  # regular objects with zero or more older versions
       name: NAME # match object names that satisfy the wildcard expression.
-      olderThan: 70h # match objects older than this value
+      olderThan: 7d10h # match objects older than this value
       createdBefore: "2006-01-02T15:04:05.00Z" # match objects created before "date"
       tags:
         - key: name
@@ -64,8 +65,61 @@ expire: # Expire objects that match a condition
     delay: 500ms # least amount of delay between each retry
 `
 	var job BatchJobRequest
-	err := yaml.UnmarshalStrict([]byte(expireYaml), &job)
+	err := yaml.Unmarshal([]byte(expireYaml), &job)
 	if err != nil {
 		t.Fatal("Failed to parse batch-job-expire yaml", err)
 	}
+	if !slices.Equal(job.Expire.Prefix.F(), []string{"myprefix"}) {
+		t.Fatal("Failed to parse batch-job-expire yaml")
+	}
+
+	multiPrefixExpireYaml := `
+expire: # Expire objects that match a condition
+  apiVersion: v1
+  bucket: mybucket # Bucket where this batch job will expire matching objects from
+  prefix: # (Optional) Prefix under which this job will expire objects matching the rules below.
+    - myprefix
+    - myprefix1
+  rules:
+    - type: object  # regular objects with zero or more older versions
+      name: NAME # match object names that satisfy the wildcard expression.
+      olderThan: 7d10h # match objects older than this value
+      createdBefore: "2006-01-02T15:04:05.00Z" # match objects created before "date"
+      tags:
+        - key: name
+          value: pick* # match objects with tag 'name', all values starting with 'pick'
+      metadata:
+        - key: content-type
+          value: image/* # match objects with 'content-type', all values starting with 'image/'
+      size:
+        lessThan: "10MiB" # match objects with size less than this value (e.g. 10MiB)
+        greaterThan: 1MiB # match objects with size greater than this value (e.g. 1MiB)
+      purge:
+          # retainVersions: 0 # (default) delete all versions of the object. This option is the fastest.
+          # retainVersions: 5 # keep the latest 5 versions of the object.
+  
+    - type: deleted # objects with delete marker as their latest version
+      name: NAME # match object names that satisfy the wildcard expression.
+      olderThan: 10h # match objects older than this value (e.g. 7d10h31s)
+      createdBefore: "2006-01-02T15:04:05.00Z" # match objects created before "date"
+      purge:
+          # retainVersions: 0 # (default) delete all versions of the object. This option is the fastest.
+          # retainVersions: 5 # keep the latest 5 versions of the object including delete markers.
+
+  notify:
+    endpoint: https://notify.endpoint # notification endpoint to receive job completion status
+    token: Bearer xxxxx # optional authentication token for the notification endpoint
+  
+  retry:
+    attempts: 10 # number of retries for the job before giving up
+    delay: 500ms # least amount of delay between each retry
+`
+	var multiPrefixJob BatchJobRequest
+	err = yaml.Unmarshal([]byte(multiPrefixExpireYaml), &multiPrefixJob)
+	if err != nil {
+		t.Fatal("Failed to parse batch-job-expire yaml", err)
+	}
+	if !slices.Equal(multiPrefixJob.Expire.Prefix.F(), []string{"myprefix", "myprefix1"}) {
+		t.Fatal("Failed to parse batch-job-expire yaml")
+	}
 }

cmd/batch-handlers.go

@@ -28,6 +28,7 @@ import (
 	"math/rand"
 	"net/http"
 	"net/url"
+	"path/filepath"
 	"runtime"
 	"strconv"
 	"strings"
@@ -57,6 +58,11 @@ import (
 
 var globalBatchConfig batch.Config
 
+const (
+	// Keep the completed/failed job stats 3 days before removing it
+	oldJobsExpiration = 3 * 24 * time.Hour
+)
+
 // BatchJobRequest this is an internal data structure not for external consumption.
 type BatchJobRequest struct {
 	ID        string               `yaml:"-" json:"name"`
@@ -262,7 +268,7 @@ func (r *BatchJobReplicateV1) StartFromSource(ctx context.Context, api ObjectLay
 		JobType:   string(job.Type()),
 		StartTime: job.Started,
 	}
-	if err := ri.load(ctx, api, job); err != nil {
+	if err := ri.loadOrInit(ctx, api, job); err != nil {
 		return err
 	}
 	if ri.Complete {
@@ -281,12 +287,12 @@ func (r *BatchJobReplicateV1) StartFromSource(ctx context.Context, api ObjectLay
 	isStorageClassOnly := len(r.Flags.Filter.Metadata) == 1 && strings.EqualFold(r.Flags.Filter.Metadata[0].Key, xhttp.AmzStorageClass)
 
 	skip := func(oi ObjectInfo) (ok bool) {
-		if r.Flags.Filter.OlderThan > 0 && time.Since(oi.ModTime) < r.Flags.Filter.OlderThan {
+		if r.Flags.Filter.OlderThan > 0 && time.Since(oi.ModTime) < r.Flags.Filter.OlderThan.D() {
 			// skip all objects that are newer than specified older duration
 			return true
 		}
 
-		if r.Flags.Filter.NewerThan > 0 && time.Since(oi.ModTime) >= r.Flags.Filter.NewerThan {
+		if r.Flags.Filter.NewerThan > 0 && time.Since(oi.ModTime) >= r.Flags.Filter.NewerThan.D() {
 			// skip all objects that are older than specified newer duration
 			return true
 		}
@@ -379,12 +385,23 @@ func (r *BatchJobReplicateV1) StartFromSource(ctx context.Context, api ObjectLay
 		s3Type := r.Target.Type == BatchJobReplicateResourceS3 || r.Source.Type == BatchJobReplicateResourceS3
 		minioSrc := r.Source.Type == BatchJobReplicateResourceMinIO
 		ctx, cancel := context.WithCancel(ctx)
-		objInfoCh := c.ListObjects(ctx, r.Source.Bucket, miniogo.ListObjectsOptions{
-			Prefix:       r.Source.Prefix,
-			WithVersions: minioSrc,
-			Recursive:    true,
-			WithMetadata: true,
-		})
+
+		objInfoCh := make(chan miniogo.ObjectInfo, 1)
+		go func() {
+			for _, prefix := range r.Source.Prefix.F() {
+				prefixObjInfoCh := c.ListObjects(ctx, r.Source.Bucket, miniogo.ListObjectsOptions{
+					Prefix:       strings.TrimSpace(prefix),
+					WithVersions: minioSrc,
+					Recursive:    true,
+					WithMetadata: true,
+				})
+				for obj := range prefixObjInfoCh {
+					objInfoCh <- obj
+				}
+			}
+			xioutil.SafeClose(objInfoCh)
+		}()
+
 		prevObj := ""
 		skipReplicate := false
 
@@ -534,7 +551,7 @@ func toObjectInfo(bucket, object string, objInfo miniogo.ObjectInfo) ObjectInfo
 	return oi
 }
 
-func (r BatchJobReplicateV1) writeAsArchive(ctx context.Context, objAPI ObjectLayer, remoteClnt *minio.Client, entries []ObjectInfo) error {
+func (r BatchJobReplicateV1) writeAsArchive(ctx context.Context, objAPI ObjectLayer, remoteClnt *minio.Client, entries []ObjectInfo, prefix string) error {
 	input := make(chan minio.SnowballObject, 1)
 	opts := minio.SnowballOptions{
 		Opts:     minio.PutObjectOptions{},
@@ -556,6 +573,10 @@ func (r BatchJobReplicateV1) writeAsArchive(ctx context.Context, objAPI ObjectLa
 				continue
 			}
 
+			if prefix != "" {
+				entry.Name = pathJoin(prefix, entry.Name)
+			}
+
 			snowballObj := minio.SnowballObject{
 				// Create path to store objects within the bucket.
 				Key:       entry.Name,
@@ -719,63 +740,85 @@ const (
 
 	batchReplJobAPIVersion        = "v1"
 	batchReplJobDefaultRetries    = 3
-	batchReplJobDefaultRetryDelay = 250 * time.Millisecond
+	batchReplJobDefaultRetryDelay = time.Second
 )
 
-func getJobReportPath(job BatchJobRequest) string {
-	var fileName string
-	switch {
-	case job.Replicate != nil:
-		fileName = batchReplName
-	case job.KeyRotate != nil:
-		fileName = batchKeyRotationName
-	case job.Expire != nil:
-		fileName = batchExpireName
-	}
-	return pathJoin(batchJobReportsPrefix, job.ID, fileName)
-}
-
 func getJobPath(job BatchJobRequest) string {
 	return pathJoin(batchJobPrefix, job.ID)
 }
 
+func (ri *batchJobInfo) getJobReportPath() (string, error) {
+	var fileName string
+	switch madmin.BatchJobType(ri.JobType) {
+	case madmin.BatchJobReplicate:
+		fileName = batchReplName
+	case madmin.BatchJobKeyRotate:
+		fileName = batchKeyRotationName
+	case madmin.BatchJobExpire:
+		fileName = batchExpireName
+	default:
+		return "", fmt.Errorf("unknown job type: %v", ri.JobType)
+	}
+	return pathJoin(batchJobReportsPrefix, ri.JobID, fileName), nil
+}
+
+func (ri *batchJobInfo) loadOrInit(ctx context.Context, api ObjectLayer, job BatchJobRequest) error {
+	err := ri.load(ctx, api, job)
+	if errors.Is(err, errNoSuchJob) {
+		switch {
+		case job.Replicate != nil:
+			ri.Version = batchReplVersionV1
+			ri.RetryAttempts = batchReplJobDefaultRetries
+			if job.Replicate.Flags.Retry.Attempts > 0 {
+				ri.RetryAttempts = job.Replicate.Flags.Retry.Attempts
+			}
+		case job.KeyRotate != nil:
+			ri.Version = batchKeyRotateVersionV1
+			ri.RetryAttempts = batchKeyRotateJobDefaultRetries
+			if job.KeyRotate.Flags.Retry.Attempts > 0 {
+				ri.RetryAttempts = job.KeyRotate.Flags.Retry.Attempts
+			}
+		case job.Expire != nil:
+			ri.Version = batchExpireVersionV1
+			ri.RetryAttempts = batchExpireJobDefaultRetries
+			if job.Expire.Retry.Attempts > 0 {
+				ri.RetryAttempts = job.Expire.Retry.Attempts
+			}
+		}
+		return nil
+	}
+	return err
+}
+
 func (ri *batchJobInfo) load(ctx context.Context, api ObjectLayer, job BatchJobRequest) error {
+	path, err := job.getJobReportPath()
+	if err != nil {
+		batchLogIf(ctx, err)
+		return err
+	}
+	return ri.loadByPath(ctx, api, path)
+}
+
+func (ri *batchJobInfo) loadByPath(ctx context.Context, api ObjectLayer, path string) error {
 	var format, version uint16
-	switch {
-	case job.Replicate != nil:
+	switch filepath.Base(path) {
+	case batchReplName:
 		version = batchReplVersionV1
 		format = batchReplFormat
-	case job.KeyRotate != nil:
+	case batchKeyRotationName:
 		version = batchKeyRotateVersionV1
 		format = batchKeyRotationFormat
-	case job.Expire != nil:
+	case batchExpireName:
 		version = batchExpireVersionV1
 		format = batchExpireFormat
 	default:
 		return errors.New("no supported batch job request specified")
 	}
-	data, err := readConfig(ctx, api, getJobReportPath(job))
+
+	data, err := readConfig(ctx, api, path)
 	if err != nil {
 		if errors.Is(err, errConfigNotFound) || isErrObjectNotFound(err) {
-			ri.Version = int(version)
-			switch {
-			case job.Replicate != nil:
-				ri.RetryAttempts = batchReplJobDefaultRetries
-				if job.Replicate.Flags.Retry.Attempts > 0 {
-					ri.RetryAttempts = job.Replicate.Flags.Retry.Attempts
-				}
-			case job.KeyRotate != nil:
-				ri.RetryAttempts = batchKeyRotateJobDefaultRetries
-				if job.KeyRotate.Flags.Retry.Attempts > 0 {
-					ri.RetryAttempts = job.KeyRotate.Flags.Retry.Attempts
-				}
-			case job.Expire != nil:
-				ri.RetryAttempts = batchExpireJobDefaultRetries
-				if job.Expire.Retry.Attempts > 0 {
-					ri.RetryAttempts = job.Expire.Retry.Attempts
-				}
-			}
-			return nil
+			return errNoSuchJob
 		}
 		return err
 	}
@@ -919,7 +962,12 @@ func (ri *batchJobInfo) updateAfter(ctx context.Context, api ObjectLayer, durati
 		if err != nil {
 			return err
 		}
-		return saveConfig(ctx, api, getJobReportPath(job), buf)
+		path, err := ri.getJobReportPath()
+		if err != nil {
+			batchLogIf(ctx, err)
+			return err
+		}
+		return saveConfig(ctx, api, path, buf)
 	}
 	ri.mu.Unlock()
 	return nil
@@ -944,8 +992,10 @@ func (ri *batchJobInfo) trackCurrentBucketObject(bucket string, info ObjectInfo,
 	ri.mu.Lock()
 	defer ri.mu.Unlock()
 
-	ri.Bucket = bucket
-	ri.Object = info.Name
+	if success {
+		ri.Bucket = bucket
+		ri.Object = info.Name
+	}
 	ri.countItem(info.Size, info.DeleteMarker, success, attempt)
 }
 
@@ -971,7 +1021,7 @@ func (r *BatchJobReplicateV1) Start(ctx context.Context, api ObjectLayer, job Ba
 		JobType:   string(job.Type()),
 		StartTime: job.Started,
 	}
-	if err := ri.load(ctx, api, job); err != nil {
+	if err := ri.loadOrInit(ctx, api, job); err != nil {
 		return err
 	}
 	if ri.Complete {
@@ -981,18 +1031,18 @@ func (r *BatchJobReplicateV1) Start(ctx context.Context, api ObjectLayer, job Ba
 	lastObject := ri.Object
 
 	delay := job.Replicate.Flags.Retry.Delay
-	if delay == 0 {
+	if delay < time.Second {
 		delay = batchReplJobDefaultRetryDelay
 	}
 	rnd := rand.New(rand.NewSource(time.Now().UnixNano()))
 
 	selectObj := func(info FileInfo) (ok bool) {
-		if r.Flags.Filter.OlderThan > 0 && time.Since(info.ModTime) < r.Flags.Filter.OlderThan {
+		if r.Flags.Filter.OlderThan > 0 && time.Since(info.ModTime) < r.Flags.Filter.OlderThan.D() {
 			// skip all objects that are newer than specified older duration
 			return false
 		}
 
-		if r.Flags.Filter.NewerThan > 0 && time.Since(info.ModTime) >= r.Flags.Filter.NewerThan {
+		if r.Flags.Filter.NewerThan > 0 && time.Since(info.ModTime) >= r.Flags.Filter.NewerThan.D() {
 			// skip all objects that are older than specified newer duration
 			return false
 		}
@@ -1071,99 +1121,106 @@ func (r *BatchJobReplicateV1) Start(ctx context.Context, api ObjectLayer, job Ba
 
 	c.SetAppInfo("minio-"+batchJobPrefix, r.APIVersion+" "+job.ID)
 
-	var (
-		walkCh = make(chan itemOrErr[ObjectInfo], 100)
-		slowCh = make(chan itemOrErr[ObjectInfo], 100)
-	)
-
-	if !*r.Source.Snowball.Disable && r.Source.Type.isMinio() && r.Target.Type.isMinio() {
-		go func() {
-			// Snowball currently needs the high level minio-go Client, not the Core one
-			cl, err := miniogo.New(u.Host, &miniogo.Options{
-				Creds:        credentials.NewStaticV4(cred.AccessKey, cred.SecretKey, cred.SessionToken),
-				Secure:       u.Scheme == "https",
-				Transport:    getRemoteInstanceTransport(),
-				BucketLookup: lookupStyle(r.Target.Path),
-			})
-			if err != nil {
-				batchLogIf(ctx, err)
-				return
-			}
-
-			// Already validated before arriving here
-			smallerThan, _ := humanize.ParseBytes(*r.Source.Snowball.SmallerThan)
-
-			batch := make([]ObjectInfo, 0, *r.Source.Snowball.Batch)
-			writeFn := func(batch []ObjectInfo) {
-				if len(batch) > 0 {
-					if err := r.writeAsArchive(ctx, api, cl, batch); err != nil {
-						batchLogIf(ctx, err)
-						for _, b := range batch {
-							slowCh <- itemOrErr[ObjectInfo]{Item: b}
-						}
-					} else {
-						ri.trackCurrentBucketBatch(r.Source.Bucket, batch)
-						globalBatchJobsMetrics.save(job.ID, ri)
-						// persist in-memory state to disk after every 10secs.
-						batchLogIf(ctx, ri.updateAfter(ctx, api, 10*time.Second, job))
-					}
-				}
-			}
-			for obj := range walkCh {
-				if obj.Item.DeleteMarker || !obj.Item.VersionPurgeStatus.Empty() || obj.Item.Size >= int64(smallerThan) {
-					slowCh <- obj
-					continue
-				}
-
-				batch = append(batch, obj.Item)
-
-				if len(batch) < *r.Source.Snowball.Batch {
-					continue
-				}
-				writeFn(batch)
-				batch = batch[:0]
-			}
-			writeFn(batch)
-			xioutil.SafeClose(slowCh)
-		}()
-	} else {
-		slowCh = walkCh
-	}
-
-	workerSize, err := strconv.Atoi(env.Get("_MINIO_BATCH_REPLICATION_WORKERS", strconv.Itoa(runtime.GOMAXPROCS(0)/2)))
-	if err != nil {
-		return err
-	}
-
-	wk, err := workers.New(workerSize)
-	if err != nil {
-		// invalid worker size.
-		return err
-	}
-
-	walkQuorum := env.Get("_MINIO_BATCH_REPLICATION_WALK_QUORUM", "strict")
-	if walkQuorum == "" {
-		walkQuorum = "strict"
-	}
-
 	retryAttempts := ri.RetryAttempts
 	retry := false
 	for attempts := 1; attempts <= retryAttempts; attempts++ {
 		attempts := attempts
+		var (
+			walkCh = make(chan itemOrErr[ObjectInfo], 100)
+			slowCh = make(chan itemOrErr[ObjectInfo], 100)
+		)
 
-		ctx, cancel := context.WithCancel(ctx)
+		if r.Source.Snowball.Disable != nil && !*r.Source.Snowball.Disable && r.Source.Type.isMinio() && r.Target.Type.isMinio() {
+			go func() {
+				// Snowball currently needs the high level minio-go Client, not the Core one
+				cl, err := miniogo.New(u.Host, &miniogo.Options{
+					Creds:        credentials.NewStaticV4(cred.AccessKey, cred.SecretKey, cred.SessionToken),
+					Secure:       u.Scheme == "https",
+					Transport:    getRemoteInstanceTransport(),
+					BucketLookup: lookupStyle(r.Target.Path),
+				})
+				if err != nil {
+					batchLogOnceIf(ctx, err, job.ID+"miniogo.New")
+					return
+				}
+
+				// Already validated before arriving here
+				smallerThan, _ := humanize.ParseBytes(*r.Source.Snowball.SmallerThan)
+
+				batch := make([]ObjectInfo, 0, *r.Source.Snowball.Batch)
+				writeFn := func(batch []ObjectInfo) {
+					if len(batch) > 0 {
+						if err := r.writeAsArchive(ctx, api, cl, batch, r.Target.Prefix); err != nil {
+							batchLogOnceIf(ctx, err, job.ID+"writeAsArchive")
+							for _, b := range batch {
+								slowCh <- itemOrErr[ObjectInfo]{Item: b}
+							}
+						} else {
+							ri.trackCurrentBucketBatch(r.Source.Bucket, batch)
+							globalBatchJobsMetrics.save(job.ID, ri)
+							// persist in-memory state to disk after every 10secs.
+							batchLogOnceIf(ctx, ri.updateAfter(ctx, api, 10*time.Second, job), job.ID+"updateAfter")
+						}
+					}
+				}
+				for obj := range walkCh {
+					if obj.Item.DeleteMarker || !obj.Item.VersionPurgeStatus.Empty() || obj.Item.Size >= int64(smallerThan) {
+						slowCh <- obj
+						continue
+					}
+
+					batch = append(batch, obj.Item)
+
+					if len(batch) < *r.Source.Snowball.Batch {
+						continue
+					}
+					writeFn(batch)
+					batch = batch[:0]
+				}
+				writeFn(batch)
+				xioutil.SafeClose(slowCh)
+			}()
+		} else {
+			slowCh = walkCh
+		}
+
+		workerSize, err := strconv.Atoi(env.Get("_MINIO_BATCH_REPLICATION_WORKERS", strconv.Itoa(runtime.GOMAXPROCS(0)/2)))
+		if err != nil {
+			return err
+		}
+
+		wk, err := workers.New(workerSize)
+		if err != nil {
+			// invalid worker size.
+			return err
+		}
+
+		walkQuorum := env.Get("_MINIO_BATCH_REPLICATION_WALK_QUORUM", "strict")
+		if walkQuorum == "" {
+			walkQuorum = "strict"
+		}
+		ctx, cancelCause := context.WithCancelCause(ctx)
 		// one of source/target is s3, skip delete marker and all versions under the same object name.
 		s3Type := r.Target.Type == BatchJobReplicateResourceS3 || r.Source.Type == BatchJobReplicateResourceS3
 
-		if err := api.Walk(ctx, r.Source.Bucket, r.Source.Prefix, walkCh, WalkOptions{
-			Marker:   lastObject,
-			Filter:   selectObj,
-			AskDisks: walkQuorum,
-		}); err != nil {
-			cancel()
-			// Do not need to retry if we can't list objects on source.
-			return err
-		}
+		go func() {
+			for _, prefix := range r.Source.Prefix.F() {
+				prefixWalkCh := make(chan itemOrErr[ObjectInfo], 100)
+				if err := api.Walk(ctx, r.Source.Bucket, strings.TrimSpace(prefix), prefixWalkCh, WalkOptions{
+					Marker:   lastObject,
+					Filter:   selectObj,
+					AskDisks: walkQuorum,
+				}); err != nil {
+					cancelCause(err)
+					xioutil.SafeClose(walkCh)
+					return
+				}
+				for obj := range prefixWalkCh {
+					walkCh <- obj
+				}
+			}
+			xioutil.SafeClose(walkCh)
+		}()
 
 		prevObj := ""
 
@@ -1171,7 +1228,7 @@ func (r *BatchJobReplicateV1) Start(ctx context.Context, api ObjectLayer, job Ba
 		for res := range slowCh {
 			if res.Err != nil {
 				ri.Failed = true
-				batchLogIf(ctx, res.Err)
+				batchLogOnceIf(ctx, res.Err, job.ID+"res.Err")
 				continue
 			}
 			result := res.Item
@@ -1198,7 +1255,7 @@ func (r *BatchJobReplicateV1) Start(ctx context.Context, api ObjectLayer, job Ba
 						return
 					}
 					stopFn(result, err)
-					batchLogIf(ctx, err)
+					batchLogOnceIf(ctx, err, job.ID+"ReplicateToTarget")
 					success = false
 				} else {
 					stopFn(result, nil)
@@ -1206,7 +1263,7 @@ func (r *BatchJobReplicateV1) Start(ctx context.Context, api ObjectLayer, job Ba
 				ri.trackCurrentBucketObject(r.Source.Bucket, result, success, attempts)
 				globalBatchJobsMetrics.save(job.ID, ri)
 				// persist in-memory state to disk after every 10secs.
-				batchLogIf(ctx, ri.updateAfter(ctx, api, 10*time.Second, job))
+				batchLogOnceIf(ctx, ri.updateAfter(ctx, api, 10*time.Second, job), job.ID+"updateAfter2")
 
 				if wait := globalBatchConfig.ReplicationWait(); wait > 0 {
 					time.Sleep(wait)
@@ -1214,20 +1271,23 @@ func (r *BatchJobReplicateV1) Start(ctx context.Context, api ObjectLayer, job Ba
 			}()
 		}
 		wk.Wait()
-
+		// Do not need to retry if we can't list objects on source.
+		if context.Cause(ctx) != nil {
+			return context.Cause(ctx)
+		}
 		ri.RetryAttempts = attempts
 		ri.Complete = ri.ObjectsFailed == 0
 		ri.Failed = ri.ObjectsFailed > 0
 
 		globalBatchJobsMetrics.save(job.ID, ri)
 		// persist in-memory state to disk.
-		batchLogIf(ctx, ri.updateAfter(ctx, api, 0, job))
+		batchLogOnceIf(ctx, ri.updateAfter(ctx, api, 0, job), job.ID+"updateAfter3")
 
 		if err := r.Notify(ctx, ri); err != nil {
-			batchLogIf(ctx, fmt.Errorf("unable to notify %v", err))
+			batchLogOnceIf(ctx, fmt.Errorf("unable to notify %v", err), job.ID+"notify")
 		}
 
-		cancel()
+		cancelCause(nil)
 		if ri.Failed {
 			ri.ObjectsFailed = 0
 			ri.Bucket = ""
@@ -1436,10 +1496,24 @@ func (j BatchJobRequest) Validate(ctx context.Context, o ObjectLayer) error {
 }
 
 func (j BatchJobRequest) delete(ctx context.Context, api ObjectLayer) {
-	deleteConfig(ctx, api, getJobReportPath(j))
 	deleteConfig(ctx, api, getJobPath(j))
 }
 
+func (j BatchJobRequest) getJobReportPath() (string, error) {
+	var fileName string
+	switch {
+	case j.Replicate != nil:
+		fileName = batchReplName
+	case j.KeyRotate != nil:
+		fileName = batchKeyRotationName
+	case j.Expire != nil:
+		fileName = batchExpireName
+	default:
+		return "", errors.New("unknown job type")
+	}
+	return pathJoin(batchJobReportsPrefix, j.ID, fileName), nil
+}
+
 func (j *BatchJobRequest) save(ctx context.Context, api ObjectLayer) error {
 	if j.Replicate == nil && j.KeyRotate == nil && j.Expire == nil {
 		return errInvalidArgument
@@ -1520,6 +1594,9 @@ func (a adminAPIHandlers) ListBatchJobs(w http.ResponseWriter, r *http.Request)
 			writeErrorResponseJSON(ctx, w, toAPIError(ctx, result.Err), r.URL)
 			return
 		}
+		if strings.HasPrefix(result.Item.Name, batchJobReportsPrefix+slashSeparator) {
+			continue
+		}
 		req := &BatchJobRequest{}
 		if err := req.load(ctx, objectAPI, result.Item.Name); err != nil {
 			if !errors.Is(err, errNoSuchJob) {
@@ -1542,6 +1619,55 @@ func (a adminAPIHandlers) ListBatchJobs(w http.ResponseWriter, r *http.Request)
 	batchLogIf(ctx, json.NewEncoder(w).Encode(&listResult))
 }
 
+// BatchJobStatus - returns the status of a batch job saved in the disk
+func (a adminAPIHandlers) BatchJobStatus(w http.ResponseWriter, r *http.Request) {
+	ctx := r.Context()
+
+	objectAPI, _ := validateAdminReq(ctx, w, r, policy.ListBatchJobsAction)
+	if objectAPI == nil {
+		return
+	}
+
+	jobID := r.Form.Get("jobId")
+	if jobID == "" {
+		writeErrorResponseJSON(ctx, w, toAPIError(ctx, errInvalidArgument), r.URL)
+		return
+	}
+
+	req := BatchJobRequest{ID: jobID}
+	if i := strings.Index(jobID, "-"); i > 0 {
+		switch madmin.BatchJobType(jobID[:i]) {
+		case madmin.BatchJobReplicate:
+			req.Replicate = &BatchJobReplicateV1{}
+		case madmin.BatchJobKeyRotate:
+			req.KeyRotate = &BatchJobKeyRotateV1{}
+		case madmin.BatchJobExpire:
+			req.Expire = &BatchJobExpire{}
+		default:
+			writeErrorResponseJSON(ctx, w, toAPIError(ctx, errors.New("job ID format unrecognized")), r.URL)
+			return
+		}
+	}
+
+	ri := &batchJobInfo{}
+	if err := ri.load(ctx, objectAPI, req); err != nil {
+		if !errors.Is(err, errNoSuchJob) {
+			batchLogIf(ctx, err)
+		}
+		writeErrorResponseJSON(ctx, w, toAPIError(ctx, err), r.URL)
+		return
+	}
+
+	buf, err := json.Marshal(madmin.BatchJobStatus{LastMetric: ri.metric()})
+	if err != nil {
+		batchLogIf(ctx, err)
+		writeErrorResponseJSON(ctx, w, toAPIError(ctx, err), r.URL)
+		return
+	}
+
+	w.Write(buf)
+}
+
 var errNoSuchJob = errors.New("no such job")
 
 // DescribeBatchJob returns the currently active batch job definition
@@ -1579,7 +1705,7 @@ func (a adminAPIHandlers) DescribeBatchJob(w http.ResponseWriter, r *http.Reques
 	w.Write(buf)
 }
 
-// StarBatchJob queue a new job for execution
+// StartBatchJob queue a new job for execution
 func (a adminAPIHandlers) StartBatchJob(w http.ResponseWriter, r *http.Request) {
 	ctx := r.Context()
 
@@ -1633,7 +1759,7 @@ func (a adminAPIHandlers) StartBatchJob(w http.ResponseWriter, r *http.Request)
 		return
 	}
 
-	job.ID = fmt.Sprintf("%s%s%d", shortuuid.New(), getKeySeparator(), GetProxyEndpointLocalIndex(globalProxyEndpoints))
+	job.ID = fmt.Sprintf("%s-%s%s%d", job.Type(), shortuuid.New(), getKeySeparator(), GetProxyEndpointLocalIndex(globalProxyEndpoints))
 	job.User = user
 	job.Started = time.Now()
 
@@ -1721,11 +1847,60 @@ func newBatchJobPool(ctx context.Context, o ObjectLayer, workers int) *BatchJobP
 		jobCancelers: make(map[string]context.CancelFunc),
 	}
 	jpool.ResizeWorkers(workers)
-	jpool.resume()
+
+	randomWait := func() time.Duration {
+		// randomWait depends on the number of nodes to avoid triggering resume and cleanups at the same time.
+		return time.Duration(rand.Float64() * float64(time.Duration(globalEndpoints.NEndpoints())*time.Hour))
+	}
+
+	go func() {
+		jpool.resume(randomWait)
+		jpool.cleanupReports(randomWait)
+	}()
+
 	return jpool
 }
 
-func (j *BatchJobPool) resume() {
+func (j *BatchJobPool) cleanupReports(randomWait func() time.Duration) {
+	t := time.NewTimer(randomWait())
+	defer t.Stop()
+
+	for {
+		select {
+		case <-GlobalContext.Done():
+			return
+		case <-t.C:
+			results := make(chan itemOrErr[ObjectInfo], 100)
+			ctx, cancel := context.WithCancel(j.ctx)
+			defer cancel()
+			if err := j.objLayer.Walk(ctx, minioMetaBucket, batchJobReportsPrefix, results, WalkOptions{}); err != nil {
+				batchLogIf(j.ctx, err)
+				t.Reset(randomWait())
+				continue
+			}
+			for result := range results {
+				if result.Err != nil {
+					batchLogIf(j.ctx, result.Err)
+					continue
+				}
+				ri := &batchJobInfo{}
+				if err := ri.loadByPath(ctx, j.objLayer, result.Item.Name); err != nil {
+					batchLogIf(ctx, err)
+					continue
+				}
+				if (ri.Complete || ri.Failed) && time.Since(ri.LastUpdate) > oldJobsExpiration {
+					deleteConfig(ctx, j.objLayer, result.Item.Name)
+				}
+			}
+
+			t.Reset(randomWait())
+		}
+	}
+}
+
+func (j *BatchJobPool) resume(randomWait func() time.Duration) {
+	time.Sleep(randomWait())
+
 	results := make(chan itemOrErr[ObjectInfo], 100)
 	ctx, cancel := context.WithCancel(j.ctx)
 	defer cancel()
@@ -1738,6 +1913,9 @@ func (j *BatchJobPool) resume() {
 			batchLogIf(j.ctx, result.Err)
 			continue
 		}
+		if strings.HasPrefix(result.Item.Name, batchJobReportsPrefix+slashSeparator) {
+			continue
+		}
 		// ignore batch-replicate.bin and batch-rotate.bin entries
 		if strings.HasSuffix(result.Item.Name, slashSeparator) {
 			continue
@@ -1988,7 +2166,7 @@ func (m *batchJobMetrics) purgeJobMetrics() {
 			var toDeleteJobMetrics []string
 			m.RLock()
 			for id, metrics := range m.metrics {
-				if time.Since(metrics.LastUpdate) > 24*time.Hour && (metrics.Complete || metrics.Failed) {
+				if time.Since(metrics.LastUpdate) > oldJobsExpiration && (metrics.Complete || metrics.Failed) {
 					toDeleteJobMetrics = append(toDeleteJobMetrics, id)
 				}
 			}
@@ -2077,3 +2255,30 @@ func lookupStyle(s string) miniogo.BucketLookupType {
 	}
 	return lookup
 }
+
+// BatchJobPrefix - to support prefix field yaml unmarshalling with string or slice of strings
+type BatchJobPrefix []string
+
+var _ yaml.Unmarshaler = &BatchJobPrefix{}
+
+// UnmarshalYAML - to support prefix field yaml unmarshalling with string or slice of strings
+func (b *BatchJobPrefix) UnmarshalYAML(value *yaml.Node) error {
+	// try slice first
+	tmpSlice := []string{}
+	if err := value.Decode(&tmpSlice); err == nil {
+		*b = tmpSlice
+		return nil
+	}
+	// try string
+	tmpStr := ""
+	if err := value.Decode(&tmpStr); err == nil {
+		*b = []string{tmpStr}
+		return nil
+	}
+	return fmt.Errorf("unable to decode %s", value.Value)
+}
+
+// F - return prefix(es) as slice
+func (b *BatchJobPrefix) F() []string {
+	return *b
+}

cmd/batch-handlers_gen.go

@@ -6,6 +6,89 @@ import (
 	"github.com/tinylib/msgp/msgp"
 )
 
+// DecodeMsg implements msgp.Decodable
+func (z *BatchJobPrefix) DecodeMsg(dc *msgp.Reader) (err error) {
+	var zb0002 uint32
+	zb0002, err = dc.ReadArrayHeader()
+	if err != nil {
+		err = msgp.WrapError(err)
+		return
+	}
+	if cap((*z)) >= int(zb0002) {
+		(*z) = (*z)[:zb0002]
+	} else {
+		(*z) = make(BatchJobPrefix, zb0002)
+	}
+	for zb0001 := range *z {
+		(*z)[zb0001], err = dc.ReadString()
+		if err != nil {
+			err = msgp.WrapError(err, zb0001)
+			return
+		}
+	}
+	return
+}
+
+// EncodeMsg implements msgp.Encodable
+func (z BatchJobPrefix) EncodeMsg(en *msgp.Writer) (err error) {
+	err = en.WriteArrayHeader(uint32(len(z)))
+	if err != nil {
+		err = msgp.WrapError(err)
+		return
+	}
+	for zb0003 := range z {
+		err = en.WriteString(z[zb0003])
+		if err != nil {
+			err = msgp.WrapError(err, zb0003)
+			return
+		}
+	}
+	return
+}
+
+// MarshalMsg implements msgp.Marshaler
+func (z BatchJobPrefix) MarshalMsg(b []byte) (o []byte, err error) {
+	o = msgp.Require(b, z.Msgsize())
+	o = msgp.AppendArrayHeader(o, uint32(len(z)))
+	for zb0003 := range z {
+		o = msgp.AppendString(o, z[zb0003])
+	}
+	return
+}
+
+// UnmarshalMsg implements msgp.Unmarshaler
+func (z *BatchJobPrefix) UnmarshalMsg(bts []byte) (o []byte, err error) {
+	var zb0002 uint32
+	zb0002, bts, err = msgp.ReadArrayHeaderBytes(bts)
+	if err != nil {
+		err = msgp.WrapError(err)
+		return
+	}
+	if cap((*z)) >= int(zb0002) {
+		(*z) = (*z)[:zb0002]
+	} else {
+		(*z) = make(BatchJobPrefix, zb0002)
+	}
+	for zb0001 := range *z {
+		(*z)[zb0001], bts, err = msgp.ReadStringBytes(bts)
+		if err != nil {
+			err = msgp.WrapError(err, zb0001)
+			return
+		}
+	}
+	o = bts
+	return
+}
+
+// Msgsize returns an upper bound estimate of the number of bytes occupied by the serialized message
+func (z BatchJobPrefix) Msgsize() (s int) {
+	s = msgp.ArrayHeaderSize
+	for zb0003 := range z {
+		s += msgp.StringPrefixSize + len(z[zb0003])
+	}
+	return
+}
+
 // DecodeMsg implements msgp.Decodable
 func (z *BatchJobRequest) DecodeMsg(dc *msgp.Reader) (err error) {
 	var field []byte

cmd/batch-handlers_gen_test.go

@@ -9,6 +9,119 @@ import (
 	"github.com/tinylib/msgp/msgp"
 )
 
+func TestMarshalUnmarshalBatchJobPrefix(t *testing.T) {
+	v := BatchJobPrefix{}
+	bts, err := v.MarshalMsg(nil)
+	if err != nil {
+		t.Fatal(err)
+	}
+	left, err := v.UnmarshalMsg(bts)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(left) > 0 {
+		t.Errorf("%d bytes left over after UnmarshalMsg(): %q", len(left), left)
+	}
+
+	left, err = msgp.Skip(bts)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(left) > 0 {
+		t.Errorf("%d bytes left over after Skip(): %q", len(left), left)
+	}
+}
+
+func BenchmarkMarshalMsgBatchJobPrefix(b *testing.B) {
+	v := BatchJobPrefix{}
+	b.ReportAllocs()
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		v.MarshalMsg(nil)
+	}
+}
+
+func BenchmarkAppendMsgBatchJobPrefix(b *testing.B) {
+	v := BatchJobPrefix{}
+	bts := make([]byte, 0, v.Msgsize())
+	bts, _ = v.MarshalMsg(bts[0:0])
+	b.SetBytes(int64(len(bts)))
+	b.ReportAllocs()
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		bts, _ = v.MarshalMsg(bts[0:0])
+	}
+}
+
+func BenchmarkUnmarshalBatchJobPrefix(b *testing.B) {
+	v := BatchJobPrefix{}
+	bts, _ := v.MarshalMsg(nil)
+	b.ReportAllocs()
+	b.SetBytes(int64(len(bts)))
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		_, err := v.UnmarshalMsg(bts)
+		if err != nil {
+			b.Fatal(err)
+		}
+	}
+}
+
+func TestEncodeDecodeBatchJobPrefix(t *testing.T) {
+	v := BatchJobPrefix{}
+	var buf bytes.Buffer
+	msgp.Encode(&buf, &v)
+
+	m := v.Msgsize()
+	if buf.Len() > m {
+		t.Log("WARNING: TestEncodeDecodeBatchJobPrefix Msgsize() is inaccurate")
+	}
+
+	vn := BatchJobPrefix{}
+	err := msgp.Decode(&buf, &vn)
+	if err != nil {
+		t.Error(err)
+	}
+
+	buf.Reset()
+	msgp.Encode(&buf, &v)
+	err = msgp.NewReader(&buf).Skip()
+	if err != nil {
+		t.Error(err)
+	}
+}
+
+func BenchmarkEncodeBatchJobPrefix(b *testing.B) {
+	v := BatchJobPrefix{}
+	var buf bytes.Buffer
+	msgp.Encode(&buf, &v)
+	b.SetBytes(int64(buf.Len()))
+	en := msgp.NewWriter(msgp.Nowhere)
+	b.ReportAllocs()
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		v.EncodeMsg(en)
+	}
+	en.Flush()
+}
+
+func BenchmarkDecodeBatchJobPrefix(b *testing.B) {
+	v := BatchJobPrefix{}
+	var buf bytes.Buffer
+	msgp.Encode(&buf, &v)
+	b.SetBytes(int64(buf.Len()))
+	rd := msgp.NewEndlessReader(buf.Bytes(), b)
+	dc := msgp.NewReader(rd)
+	b.ReportAllocs()
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		err := v.DecodeMsg(dc)
+		if err != nil {
+			b.Fatal(err)
+		}
+	}
+}
+
 func TestMarshalUnmarshalBatchJobRequest(t *testing.T) {
 	v := BatchJobRequest{}
 	bts, err := v.MarshalMsg(nil)

cmd/batch-handlers_test.go

@@ -0,0 +1,75 @@
+// Copyright (c) 2015-2024 MinIO, Inc.
+//
+// This file is part of MinIO Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"slices"
+	"testing"
+
+	"gopkg.in/yaml.v3"
+)
+
+func TestBatchJobPrefix_UnmarshalYAML(t *testing.T) {
+	type args struct {
+		yamlStr string
+	}
+	type PrefixTemp struct {
+		Prefix BatchJobPrefix `yaml:"prefix"`
+	}
+	tests := []struct {
+		name    string
+		b       PrefixTemp
+		args    args
+		want    []string
+		wantErr bool
+	}{
+		{
+			name: "test1",
+			b:    PrefixTemp{},
+			args: args{
+				yamlStr: `
+prefix: "foo"
+`,
+			},
+			want:    []string{"foo"},
+			wantErr: false,
+		},
+		{
+			name: "test2",
+			b:    PrefixTemp{},
+			args: args{
+				yamlStr: `
+prefix:
+  - "foo"
+  - "bar"
+`,
+			},
+			want: []string{"foo", "bar"},
+		},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			if err := yaml.Unmarshal([]byte(tt.args.yamlStr), &tt.b); (err != nil) != tt.wantErr {
+				t.Errorf("UnmarshalYAML() error = %v, wantErr %v", err, tt.wantErr)
+			}
+			if !slices.Equal(tt.b.Prefix.F(), tt.want) {
+				t.Errorf("UnmarshalYAML() = %v, want %v", tt.b.Prefix.F(), tt.want)
+			}
+		})
+	}
+}

cmd/batch-replicate.go

@@ -21,8 +21,8 @@ import (
 	"time"
 
 	miniogo "github.com/minio/minio-go/v7"
-
 	"github.com/minio/minio/internal/auth"
+	"github.com/minio/pkg/v3/xtime"
 )
 
 //go:generate msgp -file $GOFILE
@@ -65,12 +65,12 @@ import (
 
 // BatchReplicateFilter holds all the filters currently supported for batch replication
 type BatchReplicateFilter struct {
-	NewerThan     time.Duration `yaml:"newerThan,omitempty" json:"newerThan"`
-	OlderThan     time.Duration `yaml:"olderThan,omitempty" json:"olderThan"`
-	CreatedAfter  time.Time     `yaml:"createdAfter,omitempty" json:"createdAfter"`
-	CreatedBefore time.Time     `yaml:"createdBefore,omitempty" json:"createdBefore"`
-	Tags          []BatchJobKV  `yaml:"tags,omitempty" json:"tags"`
-	Metadata      []BatchJobKV  `yaml:"metadata,omitempty" json:"metadata"`
+	NewerThan     xtime.Duration `yaml:"newerThan,omitempty" json:"newerThan"`
+	OlderThan     xtime.Duration `yaml:"olderThan,omitempty" json:"olderThan"`
+	CreatedAfter  time.Time      `yaml:"createdAfter,omitempty" json:"createdAfter"`
+	CreatedBefore time.Time      `yaml:"createdBefore,omitempty" json:"createdBefore"`
+	Tags          []BatchJobKV   `yaml:"tags,omitempty" json:"tags"`
+	Metadata      []BatchJobKV   `yaml:"metadata,omitempty" json:"metadata"`
 }
 
 // BatchJobReplicateFlags various configurations for replication job definition currently includes
@@ -151,7 +151,7 @@ func (t BatchJobReplicateTarget) ValidPath() bool {
 type BatchJobReplicateSource struct {
 	Type     BatchJobReplicateResourceType `yaml:"type" json:"type"`
 	Bucket   string                        `yaml:"bucket" json:"bucket"`
-	Prefix   string                        `yaml:"prefix" json:"prefix"`
+	Prefix   BatchJobPrefix                `yaml:"prefix" json:"prefix"`
 	Endpoint string                        `yaml:"endpoint" json:"endpoint"`
 	Path     string                        `yaml:"path" json:"path"`
 	Creds    BatchJobReplicateCredentials  `yaml:"credentials" json:"credentials"`

cmd/batch-replicate_gen.go

@@ -411,7 +411,7 @@ func (z *BatchJobReplicateSource) DecodeMsg(dc *msgp.Reader) (err error) {
 				return
 			}
 		case "Prefix":
-			z.Prefix, err = dc.ReadString()
+			err = z.Prefix.DecodeMsg(dc)
 			if err != nil {
 				err = msgp.WrapError(err, "Prefix")
 				return
@@ -514,7 +514,7 @@ func (z *BatchJobReplicateSource) EncodeMsg(en *msgp.Writer) (err error) {
 	if err != nil {
 		return
 	}
-	err = en.WriteString(z.Prefix)
+	err = z.Prefix.EncodeMsg(en)
 	if err != nil {
 		err = msgp.WrapError(err, "Prefix")
 		return
@@ -600,7 +600,11 @@ func (z *BatchJobReplicateSource) MarshalMsg(b []byte) (o []byte, err error) {
 	o = msgp.AppendString(o, z.Bucket)
 	// string "Prefix"
 	o = append(o, 0xa6, 0x50, 0x72, 0x65, 0x66, 0x69, 0x78)
-	o = msgp.AppendString(o, z.Prefix)
+	o, err = z.Prefix.MarshalMsg(o)
+	if err != nil {
+		err = msgp.WrapError(err, "Prefix")
+		return
+	}
 	// string "Endpoint"
 	o = append(o, 0xa8, 0x45, 0x6e, 0x64, 0x70, 0x6f, 0x69, 0x6e, 0x74)
 	o = msgp.AppendString(o, z.Endpoint)
@@ -664,7 +668,7 @@ func (z *BatchJobReplicateSource) UnmarshalMsg(bts []byte) (o []byte, err error)
 				return
 			}
 		case "Prefix":
-			z.Prefix, bts, err = msgp.ReadStringBytes(bts)
+			bts, err = z.Prefix.UnmarshalMsg(bts)
 			if err != nil {
 				err = msgp.WrapError(err, "Prefix")
 				return
@@ -742,7 +746,7 @@ func (z *BatchJobReplicateSource) UnmarshalMsg(bts []byte) (o []byte, err error)
 
 // Msgsize returns an upper bound estimate of the number of bytes occupied by the serialized message
 func (z *BatchJobReplicateSource) Msgsize() (s int) {
-	s = 1 + 5 + msgp.StringPrefixSize + len(string(z.Type)) + 7 + msgp.StringPrefixSize + len(z.Bucket) + 7 + msgp.StringPrefixSize + len(z.Prefix) + 9 + msgp.StringPrefixSize + len(z.Endpoint) + 5 + msgp.StringPrefixSize + len(z.Path) + 6 + 1 + 10 + msgp.StringPrefixSize + len(z.Creds.AccessKey) + 10 + msgp.StringPrefixSize + len(z.Creds.SecretKey) + 13 + msgp.StringPrefixSize + len(z.Creds.SessionToken) + 9 + z.Snowball.Msgsize()
+	s = 1 + 5 + msgp.StringPrefixSize + len(string(z.Type)) + 7 + msgp.StringPrefixSize + len(z.Bucket) + 7 + z.Prefix.Msgsize() + 9 + msgp.StringPrefixSize + len(z.Endpoint) + 5 + msgp.StringPrefixSize + len(z.Path) + 6 + 1 + 10 + msgp.StringPrefixSize + len(z.Creds.AccessKey) + 10 + msgp.StringPrefixSize + len(z.Creds.SecretKey) + 13 + msgp.StringPrefixSize + len(z.Creds.SessionToken) + 9 + z.Snowball.Msgsize()
 	return
 }
 
@@ -1409,13 +1413,13 @@ func (z *BatchReplicateFilter) DecodeMsg(dc *msgp.Reader) (err error) {
 		}
 		switch msgp.UnsafeString(field) {
 		case "NewerThan":
-			z.NewerThan, err = dc.ReadDuration()
+			err = z.NewerThan.DecodeMsg(dc)
 			if err != nil {
 				err = msgp.WrapError(err, "NewerThan")
 				return
 			}
 		case "OlderThan":
-			z.OlderThan, err = dc.ReadDuration()
+			err = z.OlderThan.DecodeMsg(dc)
 			if err != nil {
 				err = msgp.WrapError(err, "OlderThan")
 				return
@@ -1489,7 +1493,7 @@ func (z *BatchReplicateFilter) EncodeMsg(en *msgp.Writer) (err error) {
 	if err != nil {
 		return
 	}
-	err = en.WriteDuration(z.NewerThan)
+	err = z.NewerThan.EncodeMsg(en)
 	if err != nil {
 		err = msgp.WrapError(err, "NewerThan")
 		return
@@ -1499,7 +1503,7 @@ func (z *BatchReplicateFilter) EncodeMsg(en *msgp.Writer) (err error) {
 	if err != nil {
 		return
 	}
-	err = en.WriteDuration(z.OlderThan)
+	err = z.OlderThan.EncodeMsg(en)
 	if err != nil {
 		err = msgp.WrapError(err, "OlderThan")
 		return
@@ -1567,10 +1571,18 @@ func (z *BatchReplicateFilter) MarshalMsg(b []byte) (o []byte, err error) {
 	// map header, size 6
 	// string "NewerThan"
 	o = append(o, 0x86, 0xa9, 0x4e, 0x65, 0x77, 0x65, 0x72, 0x54, 0x68, 0x61, 0x6e)
-	o = msgp.AppendDuration(o, z.NewerThan)
+	o, err = z.NewerThan.MarshalMsg(o)
+	if err != nil {
+		err = msgp.WrapError(err, "NewerThan")
+		return
+	}
 	// string "OlderThan"
 	o = append(o, 0xa9, 0x4f, 0x6c, 0x64, 0x65, 0x72, 0x54, 0x68, 0x61, 0x6e)
-	o = msgp.AppendDuration(o, z.OlderThan)
+	o, err = z.OlderThan.MarshalMsg(o)
+	if err != nil {
+		err = msgp.WrapError(err, "OlderThan")
+		return
+	}
 	// string "CreatedAfter"
 	o = append(o, 0xac, 0x43, 0x72, 0x65, 0x61, 0x74, 0x65, 0x64, 0x41, 0x66, 0x74, 0x65, 0x72)
 	o = msgp.AppendTime(o, z.CreatedAfter)
@@ -1619,13 +1631,13 @@ func (z *BatchReplicateFilter) UnmarshalMsg(bts []byte) (o []byte, err error) {
 		}
 		switch msgp.UnsafeString(field) {
 		case "NewerThan":
-			z.NewerThan, bts, err = msgp.ReadDurationBytes(bts)
+			bts, err = z.NewerThan.UnmarshalMsg(bts)
 			if err != nil {
 				err = msgp.WrapError(err, "NewerThan")
 				return
 			}
 		case "OlderThan":
-			z.OlderThan, bts, err = msgp.ReadDurationBytes(bts)
+			bts, err = z.OlderThan.UnmarshalMsg(bts)
 			if err != nil {
 				err = msgp.WrapError(err, "OlderThan")
 				return
@@ -1694,7 +1706,7 @@ func (z *BatchReplicateFilter) UnmarshalMsg(bts []byte) (o []byte, err error) {
 
 // Msgsize returns an upper bound estimate of the number of bytes occupied by the serialized message
 func (z *BatchReplicateFilter) Msgsize() (s int) {
-	s = 1 + 10 + msgp.DurationSize + 10 + msgp.DurationSize + 13 + msgp.TimeSize + 14 + msgp.TimeSize + 5 + msgp.ArrayHeaderSize
+	s = 1 + 10 + z.NewerThan.Msgsize() + 10 + z.OlderThan.Msgsize() + 13 + msgp.TimeSize + 14 + msgp.TimeSize + 5 + msgp.ArrayHeaderSize
 	for za0001 := range z.Tags {
 		s += z.Tags[za0001].Msgsize()
 	}

cmd/batch-replicate_test.go

@@ -0,0 +1,182 @@
+// Copyright (c) 2015-2024 MinIO, Inc.
+//
+// This file is part of MinIO Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"slices"
+	"testing"
+
+	"gopkg.in/yaml.v3"
+)
+
+func TestParseBatchJobReplicate(t *testing.T) {
+	replicateYaml := `
+replicate:
+  apiVersion: v1
+  # source of the objects to be replicated
+  source:
+    type: minio # valid values are "s3" or "minio"
+    bucket: mytest
+    prefix: object-prefix1 # 'PREFIX' is optional
+    # If your source is the 'local' alias specified to 'mc batch start', then the 'endpoint' and 'credentials' fields are optional and can be omitted
+    # Either the 'source' or 'remote' *must* be the "local" deployment
+#    endpoint: "http://127.0.0.1:9000"
+#    # path: "on|off|auto" # "on" enables path-style bucket lookup. "off" enables virtual host (DNS)-style bucket lookup. Defaults to "auto"
+#    credentials:
+#      accessKey: minioadmin # Required
+#      secretKey: minioadmin # Required
+#    # sessionToken: SESSION-TOKEN # Optional only available when rotating credentials are used
+    snowball: # automatically activated if the source is local
+      disable: true # optionally turn-off snowball archive transfer
+#      batch: 100 # upto this many objects per archive
+#      inmemory: true # indicates if the archive must be staged locally or in-memory
+#      compress: false # S2/Snappy compressed archive
+#      smallerThan: 5MiB # create archive for all objects smaller than 5MiB
+#      skipErrs: false # skips any source side read() errors
+
+  # target where the objects must be replicated
+  target:
+    type: minio # valid values are "s3" or "minio"
+    bucket: mytest
+    prefix: stage # 'PREFIX' is optional
+    # If your source is the 'local' alias specified to 'mc batch start', then the 'endpoint' and 'credentials' fields are optional and can be omitted
+
+    # Either the 'source' or 'remote' *must* be the "local" deployment
+    endpoint: "http://127.0.0.1:9001"
+    # path: "on|off|auto" # "on" enables path-style bucket lookup. "off" enables virtual host (DNS)-style bucket lookup. Defaults to "auto"
+    credentials:
+      accessKey: minioadmin
+      secretKey: minioadmin
+    # sessionToken: SESSION-TOKEN # Optional only available when rotating credentials are used
+
+  # NOTE: All flags are optional
+  # - filtering criteria only applies for all source objects match the criteria
+  # - configurable notification endpoints
+  # - configurable retries for the job (each retry skips successfully previously replaced objects)
+  flags:
+    filter:
+      newerThan: "7d10h31s" # match objects newer than this value (e.g. 7d10h31s)
+      olderThan: "7d" # match objects older than this value (e.g. 7d10h31s)
+#      createdAfter: "date" # match objects created after "date"
+#      createdBefore: "date" # match objects created before "date"
+
+      ## NOTE: tags are not supported when "source" is remote.
+      tags:
+         - key: "name"
+           value: "pick*" # match objects with tag 'name', with all values starting with 'pick'
+
+      metadata:
+         - key: "content-type"
+           value: "image/*" # match objects with 'content-type', with all values starting with 'image/'
+
+#    notify:
+#      endpoint: "https://notify.endpoint" # notification endpoint to receive job status events
+#      token: "Bearer xxxxx" # optional authentication token for the notification endpoint
+#
+#    retry:
+#      attempts: 10 # number of retries for the job before giving up
+#      delay: "500ms" # least amount of delay between each retry
+
+`
+	var job BatchJobRequest
+	err := yaml.Unmarshal([]byte(replicateYaml), &job)
+	if err != nil {
+		t.Fatal("Failed to parse batch-job-replicate yaml", err)
+	}
+	if !slices.Equal(job.Replicate.Source.Prefix.F(), []string{"object-prefix1"}) {
+		t.Fatal("Failed to parse batch-job-replicate yaml", err)
+	}
+	multiPrefixReplicateYaml := `
+replicate:
+  apiVersion: v1
+  # source of the objects to be replicated
+  source:
+    type: minio # valid values are "s3" or "minio"
+    bucket: mytest
+    prefix: # 'PREFIX' is optional
+      - object-prefix1 
+      - object-prefix2
+    # If your source is the 'local' alias specified to 'mc batch start', then the 'endpoint' and 'credentials' fields are optional and can be omitted
+    # Either the 'source' or 'remote' *must* be the "local" deployment
+#    endpoint: "http://127.0.0.1:9000"
+#    # path: "on|off|auto" # "on" enables path-style bucket lookup. "off" enables virtual host (DNS)-style bucket lookup. Defaults to "auto"
+#    credentials:
+#      accessKey: minioadmin # Required
+#      secretKey: minioadmin # Required
+#    # sessionToken: SESSION-TOKEN # Optional only available when rotating credentials are used
+    snowball: # automatically activated if the source is local
+      disable: true # optionally turn-off snowball archive transfer
+#      batch: 100 # upto this many objects per archive
+#      inmemory: true # indicates if the archive must be staged locally or in-memory
+#      compress: false # S2/Snappy compressed archive
+#      smallerThan: 5MiB # create archive for all objects smaller than 5MiB
+#      skipErrs: false # skips any source side read() errors
+
+  # target where the objects must be replicated
+  target:
+    type: minio # valid values are "s3" or "minio"
+    bucket: mytest
+    prefix: stage # 'PREFIX' is optional
+    # If your source is the 'local' alias specified to 'mc batch start', then the 'endpoint' and 'credentials' fields are optional and can be omitted
+
+    # Either the 'source' or 'remote' *must* be the "local" deployment
+    endpoint: "http://127.0.0.1:9001"
+    # path: "on|off|auto" # "on" enables path-style bucket lookup. "off" enables virtual host (DNS)-style bucket lookup. Defaults to "auto"
+    credentials:
+      accessKey: minioadmin
+      secretKey: minioadmin
+    # sessionToken: SESSION-TOKEN # Optional only available when rotating credentials are used
+
+  # NOTE: All flags are optional
+  # - filtering criteria only applies for all source objects match the criteria
+  # - configurable notification endpoints
+  # - configurable retries for the job (each retry skips successfully previously replaced objects)
+  flags:
+    filter:
+      newerThan: "7d10h31s" # match objects newer than this value (e.g. 7d10h31s)
+      olderThan: "7d" # match objects older than this value (e.g. 7d10h31s)
+#      createdAfter: "date" # match objects created after "date"
+#      createdBefore: "date" # match objects created before "date"
+
+      ## NOTE: tags are not supported when "source" is remote.
+      tags:
+         - key: "name"
+           value: "pick*" # match objects with tag 'name', with all values starting with 'pick'
+
+      metadata:
+         - key: "content-type"
+           value: "image/*" # match objects with 'content-type', with all values starting with 'image/'
+
+#    notify:
+#      endpoint: "https://notify.endpoint" # notification endpoint to receive job status events
+#      token: "Bearer xxxxx" # optional authentication token for the notification endpoint
+#
+#    retry:
+#      attempts: 10 # number of retries for the job before giving up
+#      delay: "500ms" # least amount of delay between each retry
+
+`
+	var multiPrefixJob BatchJobRequest
+	err = yaml.Unmarshal([]byte(multiPrefixReplicateYaml), &multiPrefixJob)
+	if err != nil {
+		t.Fatal("Failed to parse batch-job-replicate yaml", err)
+	}
+	if !slices.Equal(multiPrefixJob.Replicate.Source.Prefix.F(), []string{"object-prefix1", "object-prefix2"}) {
+		t.Fatal("Failed to parse batch-job-replicate yaml")
+	}
+}

cmd/batch-rotate.go

@@ -257,7 +257,7 @@ func (r *BatchJobKeyRotateV1) Start(ctx context.Context, api ObjectLayer, job Ba
 		JobType:   string(job.Type()),
 		StartTime: job.Started,
 	}
-	if err := ri.load(ctx, api, job); err != nil {
+	if err := ri.loadOrInit(ctx, api, job); err != nil {
 		return err
 	}
 	if ri.Complete {
@@ -389,6 +389,17 @@ func (r *BatchJobKeyRotateV1) Start(ctx context.Context, api ObjectLayer, job Ba
 					stopFn(result, err)
 					batchLogIf(ctx, err)
 					success = false
+					if attempts >= retryAttempts {
+						auditOptions := AuditLogOptions{
+							Event:     "KeyRotate",
+							APIName:   "StartBatchJob",
+							Bucket:    result.Bucket,
+							Object:    result.Name,
+							VersionID: result.VersionID,
+							Error:     err.Error(),
+						}
+						auditLogInternal(ctx, auditOptions)
+					}
 				} else {
 					stopFn(result, nil)
 				}

cmd/bitrot-streaming.go

@@ -141,13 +141,7 @@ func (b *streamingBitrotReader) Close() error {
 	}
 	if closer, ok := b.rc.(io.Closer); ok {
 		// drain the body for connection reuse at network layer.
-		xhttp.DrainBody(struct {
-			io.Reader
-			io.Closer
-		}{
-			Reader: b.rc,
-			Closer: closeWrapper(func() error { return nil }),
-		})
+		xhttp.DrainBody(io.NopCloser(b.rc))
 		return closer.Close()
 	}
 	return nil

cmd/bootstrap-peer-server.go

@@ -19,9 +19,13 @@ package cmd
 
 import (
 	"context"
+	"crypto/md5"
+	"encoding/hex"
 	"errors"
 	"fmt"
+	"io"
 	"math/rand"
+	"os"
 	"reflect"
 	"strings"
 	"sync"
@@ -43,10 +47,15 @@ type ServerSystemConfig struct {
 	NEndpoints int
 	CmdLines   []string
 	MinioEnv   map[string]string
+	Checksum   string
 }
 
 // Diff - returns error on first difference found in two configs.
 func (s1 *ServerSystemConfig) Diff(s2 *ServerSystemConfig) error {
+	if s1.Checksum != s2.Checksum {
+		return fmt.Errorf("Expected MinIO binary checksum: %s, seen: %s", s1.Checksum, s2.Checksum)
+	}
+
 	ns1 := s1.NEndpoints
 	ns2 := s2.NEndpoints
 	if ns1 != ns2 {
@@ -82,7 +91,7 @@ func (s1 *ServerSystemConfig) Diff(s2 *ServerSystemConfig) error {
 			extra = append(extra, k)
 		}
 	}
-	msg := "Expected same MINIO_ environment variables and values across all servers: "
+	msg := "Expected MINIO_* environment name and values across all servers to be same: "
 	if len(missing) > 0 {
 		msg += fmt.Sprintf(`Missing environment values: %v. `, missing)
 	}
@@ -97,14 +106,17 @@ func (s1 *ServerSystemConfig) Diff(s2 *ServerSystemConfig) error {
 }
 
 var skipEnvs = map[string]struct{}{
-	"MINIO_OPTS":          {},
-	"MINIO_CERT_PASSWD":   {},
-	"MINIO_SERVER_DEBUG":  {},
-	"MINIO_DSYNC_TRACE":   {},
-	"MINIO_ROOT_USER":     {},
-	"MINIO_ROOT_PASSWORD": {},
-	"MINIO_ACCESS_KEY":    {},
-	"MINIO_SECRET_KEY":    {},
+	"MINIO_OPTS":                   {},
+	"MINIO_CERT_PASSWD":            {},
+	"MINIO_SERVER_DEBUG":           {},
+	"MINIO_DSYNC_TRACE":            {},
+	"MINIO_ROOT_USER":              {},
+	"MINIO_ROOT_PASSWORD":          {},
+	"MINIO_ACCESS_KEY":             {},
+	"MINIO_SECRET_KEY":             {},
+	"MINIO_OPERATOR_VERSION":       {},
+	"MINIO_VSPHERE_PLUGIN_VERSION": {},
+	"MINIO_CI_CD":                  {},
 }
 
 func getServerSystemCfg() *ServerSystemConfig {
@@ -120,7 +132,7 @@ func getServerSystemCfg() *ServerSystemConfig {
 		}
 		envValues[envK] = logger.HashString(env.Get(envK, ""))
 	}
-	scfg := &ServerSystemConfig{NEndpoints: globalEndpoints.NEndpoints(), MinioEnv: envValues}
+	scfg := &ServerSystemConfig{NEndpoints: globalEndpoints.NEndpoints(), MinioEnv: envValues, Checksum: binaryChecksum}
 	var cmdLines []string
 	for _, ep := range globalEndpoints {
 		cmdLines = append(cmdLines, ep.CmdLine)
@@ -167,6 +179,18 @@ func (client *bootstrapRESTClient) String() string {
 	return client.gridConn.String()
 }
 
+var binaryChecksum = getBinaryChecksum()
+
+func getBinaryChecksum() string {
+	mw := md5.New()
+	b, err := os.Open(os.Args[0])
+	if err == nil {
+		defer b.Close()
+		io.Copy(mw, b)
+	}
+	return hex.EncodeToString(mw.Sum(nil))
+}
+
 func verifyServerSystemConfig(ctx context.Context, endpointServerPools EndpointServerPools, gm *grid.Manager) error {
 	srcCfg := getServerSystemCfg()
 	clnts := newBootstrapRESTClients(endpointServerPools, gm)
@@ -196,7 +220,7 @@ func verifyServerSystemConfig(ctx context.Context, endpointServerPools EndpointS
 				err := clnt.Verify(ctx, srcCfg)
 				mu.Lock()
 				if err != nil {
-					bootstrapTraceMsg(fmt.Sprintf("clnt.Verify: %v, endpoint: %s", err, clnt))
+					bootstrapTraceMsg(fmt.Sprintf("bootstrapVerify: %v, endpoint: %s", err, clnt))
 					if !isNetworkError(err) {
 						bootLogOnceIf(context.Background(), fmt.Errorf("%s has incorrect configuration: %w", clnt, err), "incorrect_"+clnt.String())
 						incorrectConfigs = append(incorrectConfigs, fmt.Errorf("%s has incorrect configuration: %w", clnt, err))

cmd/bootstrap-peer-server_gen.go

@@ -79,6 +79,12 @@ func (z *ServerSystemConfig) DecodeMsg(dc *msgp.Reader) (err error) {
 				}
 				z.MinioEnv[za0002] = za0003
 			}
+		case "Checksum":
+			z.Checksum, err = dc.ReadString()
+			if err != nil {
+				err = msgp.WrapError(err, "Checksum")
+				return
+			}
 		default:
 			err = dc.Skip()
 			if err != nil {
@@ -92,9 +98,9 @@ func (z *ServerSystemConfig) DecodeMsg(dc *msgp.Reader) (err error) {
 
 // EncodeMsg implements msgp.Encodable
 func (z *ServerSystemConfig) EncodeMsg(en *msgp.Writer) (err error) {
-	// map header, size 3
+	// map header, size 4
 	// write "NEndpoints"
-	err = en.Append(0x83, 0xaa, 0x4e, 0x45, 0x6e, 0x64, 0x70, 0x6f, 0x69, 0x6e, 0x74, 0x73)
+	err = en.Append(0x84, 0xaa, 0x4e, 0x45, 0x6e, 0x64, 0x70, 0x6f, 0x69, 0x6e, 0x74, 0x73)
 	if err != nil {
 		return
 	}
@@ -142,15 +148,25 @@ func (z *ServerSystemConfig) EncodeMsg(en *msgp.Writer) (err error) {
 			return
 		}
 	}
+	// write "Checksum"
+	err = en.Append(0xa8, 0x43, 0x68, 0x65, 0x63, 0x6b, 0x73, 0x75, 0x6d)
+	if err != nil {
+		return
+	}
+	err = en.WriteString(z.Checksum)
+	if err != nil {
+		err = msgp.WrapError(err, "Checksum")
+		return
+	}
 	return
 }
 
 // MarshalMsg implements msgp.Marshaler
 func (z *ServerSystemConfig) MarshalMsg(b []byte) (o []byte, err error) {
 	o = msgp.Require(b, z.Msgsize())
-	// map header, size 3
+	// map header, size 4
 	// string "NEndpoints"
-	o = append(o, 0x83, 0xaa, 0x4e, 0x45, 0x6e, 0x64, 0x70, 0x6f, 0x69, 0x6e, 0x74, 0x73)
+	o = append(o, 0x84, 0xaa, 0x4e, 0x45, 0x6e, 0x64, 0x70, 0x6f, 0x69, 0x6e, 0x74, 0x73)
 	o = msgp.AppendInt(o, z.NEndpoints)
 	// string "CmdLines"
 	o = append(o, 0xa8, 0x43, 0x6d, 0x64, 0x4c, 0x69, 0x6e, 0x65, 0x73)
@@ -165,6 +181,9 @@ func (z *ServerSystemConfig) MarshalMsg(b []byte) (o []byte, err error) {
 		o = msgp.AppendString(o, za0002)
 		o = msgp.AppendString(o, za0003)
 	}
+	// string "Checksum"
+	o = append(o, 0xa8, 0x43, 0x68, 0x65, 0x63, 0x6b, 0x73, 0x75, 0x6d)
+	o = msgp.AppendString(o, z.Checksum)
 	return
 }
 
@@ -241,6 +260,12 @@ func (z *ServerSystemConfig) UnmarshalMsg(bts []byte) (o []byte, err error) {
 				}
 				z.MinioEnv[za0002] = za0003
 			}
+		case "Checksum":
+			z.Checksum, bts, err = msgp.ReadStringBytes(bts)
+			if err != nil {
+				err = msgp.WrapError(err, "Checksum")
+				return
+			}
 		default:
 			bts, err = msgp.Skip(bts)
 			if err != nil {
@@ -266,5 +291,6 @@ func (z *ServerSystemConfig) Msgsize() (s int) {
 			s += msgp.StringPrefixSize + len(za0002) + msgp.StringPrefixSize + len(za0003)
 		}
 	}
+	s += 9 + msgp.StringPrefixSize + len(z.Checksum)
 	return
 }

cmd/bucket-handlers.go

@@ -51,7 +51,6 @@ import (
 	sse "github.com/minio/minio/internal/bucket/encryption"
 	objectlock "github.com/minio/minio/internal/bucket/object/lock"
 	"github.com/minio/minio/internal/bucket/replication"
-	"github.com/minio/minio/internal/config/cache"
 	"github.com/minio/minio/internal/config/dns"
 	"github.com/minio/minio/internal/crypto"
 	"github.com/minio/minio/internal/event"
@@ -92,7 +91,7 @@ const (
 // -- If IP of the entry doesn't match, this means entry is
 //
 //	for another instance. Log an error to console.
-func initFederatorBackend(buckets []BucketInfo, objLayer ObjectLayer) {
+func initFederatorBackend(buckets []string, objLayer ObjectLayer) {
 	if len(buckets) == 0 {
 		return
 	}
@@ -113,10 +112,10 @@ func initFederatorBackend(buckets []BucketInfo, objLayer ObjectLayer) {
 	domainMissing := err == dns.ErrDomainMissing
 	if dnsBuckets != nil {
 		for _, bucket := range buckets {
-			bucketsSet.Add(bucket.Name)
-			r, ok := dnsBuckets[bucket.Name]
+			bucketsSet.Add(bucket)
+			r, ok := dnsBuckets[bucket]
 			if !ok {
-				bucketsToBeUpdated.Add(bucket.Name)
+				bucketsToBeUpdated.Add(bucket)
 				continue
 			}
 			if !globalDomainIPs.Intersection(set.CreateStringSet(getHostsSlice(r)...)).IsEmpty() {
@@ -135,7 +134,7 @@ func initFederatorBackend(buckets []BucketInfo, objLayer ObjectLayer) {
 				// but if we do see a difference with local domain IPs with
 				// hostSlice from etcd then we should update with newer
 				// domainIPs, we proceed to do that here.
-				bucketsToBeUpdated.Add(bucket.Name)
+				bucketsToBeUpdated.Add(bucket)
 				continue
 			}
 
@@ -144,7 +143,7 @@ func initFederatorBackend(buckets []BucketInfo, objLayer ObjectLayer) {
 			// bucket names are globally unique in federation at a given
 			// path prefix, name collision is not allowed. We simply log
 			// an error and continue.
-			bucketsInConflict.Add(bucket.Name)
+			bucketsInConflict.Add(bucket)
 		}
 	}
 
@@ -1342,22 +1341,6 @@ func (api objectAPIHandlers) PostPolicyBucketHandler(w http.ResponseWriter, r *h
 					continue
 				}
 
-				asize, err := objInfo.GetActualSize()
-				if err != nil {
-					asize = objInfo.Size
-				}
-
-				globalCacheConfig.Set(&cache.ObjectInfo{
-					Key:          objInfo.Name,
-					Bucket:       objInfo.Bucket,
-					ETag:         getDecryptedETag(formValues, objInfo, false),
-					ModTime:      objInfo.ModTime,
-					Expires:      objInfo.Expires.UTC().Format(http.TimeFormat),
-					CacheControl: objInfo.CacheControl,
-					Metadata:     cleanReservedKeys(objInfo.UserDefined),
-					Size:         asize,
-				})
-
 				fanOutResp = append(fanOutResp, minio.PutObjectFanOutResponse{
 					Key:          objInfo.Name,
 					ETag:         getDecryptedETag(formValues, objInfo, false),
@@ -1452,22 +1435,6 @@ func (api objectAPIHandlers) PostPolicyBucketHandler(w http.ResponseWriter, r *h
 		w.Header().Set(xhttp.Location, obj)
 	}
 
-	asize, err := objInfo.GetActualSize()
-	if err != nil {
-		asize = objInfo.Size
-	}
-
-	defer globalCacheConfig.Set(&cache.ObjectInfo{
-		Key:          objInfo.Name,
-		Bucket:       objInfo.Bucket,
-		ETag:         etag,
-		ModTime:      objInfo.ModTime,
-		Expires:      objInfo.ExpiresStr(),
-		CacheControl: objInfo.CacheControl,
-		Metadata:     cleanReservedKeys(objInfo.UserDefined),
-		Size:         asize,
-	})
-
 	// Notify object created event.
 	defer sendEvent(eventArgs{
 		EventName:    event.ObjectCreatedPost,
@@ -1715,7 +1682,7 @@ func (api objectAPIHandlers) DeleteBucketHandler(w http.ResponseWriter, r *http.
 	}
 
 	globalNotificationSys.DeleteBucketMetadata(ctx, bucket)
-	globalReplicationPool.deleteResyncMetadata(ctx, bucket)
+	globalReplicationPool.Get().deleteResyncMetadata(ctx, bucket)
 
 	// Call site replication hook.
 	replLogIf(ctx, globalSiteReplicationSys.DeleteBucketHook(ctx, bucket, forceDelete))

cmd/bucket-lifecycle-audit.go

@@ -17,7 +17,11 @@
 
 package cmd
 
-import "github.com/minio/minio/internal/bucket/lifecycle"
+import (
+	"strconv"
+
+	"github.com/minio/minio/internal/bucket/lifecycle"
+)
 
 //go:generate stringer -type lcEventSrc -trimprefix lcEventSrc_ $GOFILE
 type lcEventSrc uint8
@@ -43,7 +47,7 @@ type lcAuditEvent struct {
 	source lcEventSrc
 }
 
-func (lae lcAuditEvent) Tags() map[string]interface{} {
+func (lae lcAuditEvent) Tags() map[string]string {
 	event := lae.Event
 	src := lae.source
 	const (
@@ -55,7 +59,7 @@ func (lae lcAuditEvent) Tags() map[string]interface{} {
 		ilmNewerNoncurrentVersions = "ilm-newer-noncurrent-versions"
 		ilmNoncurrentDays          = "ilm-noncurrent-days"
 	)
-	tags := make(map[string]interface{}, 5)
+	tags := make(map[string]string, 5)
 	if src > lcEventSrc_None {
 		tags[ilmSrc] = src.String()
 	}
@@ -63,7 +67,7 @@ func (lae lcAuditEvent) Tags() map[string]interface{} {
 	tags[ilmRuleID] = event.RuleID
 
 	if !event.Due.IsZero() {
-		tags[ilmDue] = event.Due
+		tags[ilmDue] = event.Due.Format(iso8601Format)
 	}
 
 	// rule with Transition/NoncurrentVersionTransition in effect
@@ -73,10 +77,10 @@ func (lae lcAuditEvent) Tags() map[string]interface{} {
 
 	// rule with NewernoncurrentVersions in effect
 	if event.NewerNoncurrentVersions > 0 {
-		tags[ilmNewerNoncurrentVersions] = event.NewerNoncurrentVersions
+		tags[ilmNewerNoncurrentVersions] = strconv.Itoa(event.NewerNoncurrentVersions)
 	}
 	if event.NoncurrentDays > 0 {
-		tags[ilmNoncurrentDays] = event.NoncurrentDays
+		tags[ilmNoncurrentDays] = strconv.Itoa(event.NoncurrentDays)
 	}
 	return tags
 }

cmd/bucket-lifecycle.go

@@ -71,7 +71,7 @@ func NewLifecycleSys() *LifecycleSys {
 	return &LifecycleSys{}
 }
 
-func ilmTrace(startTime time.Time, duration time.Duration, oi ObjectInfo, event string) madmin.TraceInfo {
+func ilmTrace(startTime time.Time, duration time.Duration, oi ObjectInfo, event string, metadata map[string]string) madmin.TraceInfo {
 	sz, _ := oi.GetActualSize()
 	return madmin.TraceInfo{
 		TraceType: madmin.TraceILM,
@@ -83,16 +83,16 @@ func ilmTrace(startTime time.Time, duration time.Duration, oi ObjectInfo, event
 		Bytes:     sz,
 		Error:     "",
 		Message:   getSource(4),
-		Custom:    map[string]string{"version-id": oi.VersionID},
+		Custom:    metadata,
 	}
 }
 
-func (sys *LifecycleSys) trace(oi ObjectInfo) func(event string) {
+func (sys *LifecycleSys) trace(oi ObjectInfo) func(event string, metadata map[string]string) {
 	startTime := time.Now()
-	return func(event string) {
+	return func(event string, metadata map[string]string) {
 		duration := time.Since(startTime)
 		if globalTrace.NumSubscribers(madmin.TraceILM) > 0 {
-			globalTrace.Publish(ilmTrace(startTime, duration, oi, event))
+			globalTrace.Publish(ilmTrace(startTime, duration, oi, event, metadata))
 		}
 	}
 }
@@ -707,6 +707,11 @@ type auditTierOp struct {
 	Error            string `json:"error,omitempty"`
 }
 
+func (op auditTierOp) String() string {
+	// flattening the auditTierOp{} for audit
+	return fmt.Sprintf("tier:%s,respNS:%d,tx:%d,err:%s", op.Tier, op.TimeToResponseNS, op.OutputBytes, op.Error)
+}
+
 func auditTierActions(ctx context.Context, tier string, bytes int64) func(err error) {
 	startTime := time.Now()
 	return func(err error) {
@@ -730,7 +735,7 @@ func auditTierActions(ctx context.Context, tier string, bytes int64) func(err er
 			globalTierMetrics.logFailure(tier)
 		}
 
-		logger.GetReqInfo(ctx).AppendTags("tierStats", op)
+		logger.GetReqInfo(ctx).AppendTags("tierStats", op.String())
 	}
 }
 
@@ -998,7 +1003,7 @@ func ongoingRestoreObj() restoreObjStatus {
 	}
 }
 
-// completeRestoreObj constructs restoreObjStatus for a completed restore-object with given expiry.
+// completedRestoreObj constructs restoreObjStatus for a completed restore-object with given expiry.
 func completedRestoreObj(expiry time.Time) restoreObjStatus {
 	return restoreObjStatus{
 		ongoing: false,

cmd/bucket-metadata-sys.go

@@ -39,6 +39,7 @@ import (
 	"github.com/minio/minio/internal/logger"
 	"github.com/minio/pkg/v3/policy"
 	"github.com/minio/pkg/v3/sync/errgroup"
+	"golang.org/x/sync/singleflight"
 )
 
 // BucketMetadataSys captures all bucket metadata for a given cluster.
@@ -47,6 +48,7 @@ type BucketMetadataSys struct {
 
 	sync.RWMutex
 	initialized bool
+	group       *singleflight.Group
 	metadataMap map[string]BucketMetadata
 }
 
@@ -62,6 +64,7 @@ func (sys *BucketMetadataSys) Count() int {
 func (sys *BucketMetadataSys) Remove(buckets ...string) {
 	sys.Lock()
 	for _, bucket := range buckets {
+		sys.group.Forget(bucket)
 		delete(sys.metadataMap, bucket)
 		globalBucketMonitor.DeleteBucket(bucket)
 	}
@@ -121,6 +124,7 @@ func (sys *BucketMetadataSys) updateAndParse(ctx context.Context, bucket string,
 		meta.PolicyConfigUpdatedAt = updatedAt
 	case bucketNotificationConfig:
 		meta.NotificationConfigXML = configData
+		meta.NotificationConfigUpdatedAt = updatedAt
 	case bucketLifecycleConfig:
 		meta.LifecycleConfigXML = configData
 		meta.LifecycleConfigUpdatedAt = updatedAt
@@ -150,12 +154,13 @@ func (sys *BucketMetadataSys) updateAndParse(ctx context.Context, bucket string,
 		if err != nil {
 			return updatedAt, fmt.Errorf("Error encrypting bucket target metadata %w", err)
 		}
+		meta.BucketTargetsConfigUpdatedAt = updatedAt
+		meta.BucketTargetsConfigMetaUpdatedAt = updatedAt
 	default:
 		return updatedAt, fmt.Errorf("Unknown bucket %s metadata update requested %s", bucket, configFile)
 	}
 
-	err = sys.save(ctx, meta)
-	return updatedAt, err
+	return updatedAt, sys.save(ctx, meta)
 }
 
 func (sys *BucketMetadataSys) save(ctx context.Context, meta BucketMetadata) error {
@@ -262,6 +267,21 @@ func (sys *BucketMetadataSys) GetVersioningConfig(bucket string) (*versioning.Ve
 	return meta.versioningConfig, meta.VersioningConfigUpdatedAt, nil
 }
 
+// GetBucketPolicy returns configured bucket policy
+func (sys *BucketMetadataSys) GetBucketPolicy(bucket string) (*policy.BucketPolicy, time.Time, error) {
+	meta, _, err := sys.GetConfig(GlobalContext, bucket)
+	if err != nil {
+		if errors.Is(err, errConfigNotFound) {
+			return nil, time.Time{}, BucketPolicyNotFound{Bucket: bucket}
+		}
+		return nil, time.Time{}, err
+	}
+	if meta.policyConfig == nil {
+		return nil, time.Time{}, BucketPolicyNotFound{Bucket: bucket}
+	}
+	return meta.policyConfig, meta.PolicyConfigUpdatedAt, nil
+}
+
 // GetTaggingConfig returns configured tagging config
 // The returned object may not be modified.
 func (sys *BucketMetadataSys) GetTaggingConfig(bucket string) (*tags.Tags, time.Time, error) {
@@ -392,9 +412,7 @@ func (sys *BucketMetadataSys) GetReplicationConfig(ctx context.Context, bucket s
 		return nil, time.Time{}, BucketReplicationConfigNotFound{Bucket: bucket}
 	}
 	if reloaded {
-		globalBucketTargetSys.set(BucketInfo{
-			Name: bucket,
-		}, meta)
+		globalBucketTargetSys.set(bucket, meta)
 	}
 	return meta.replicationConfig, meta.ReplicationConfigUpdatedAt, nil
 }
@@ -413,9 +431,7 @@ func (sys *BucketMetadataSys) GetBucketTargetsConfig(bucket string) (*madmin.Buc
 		return nil, BucketRemoteTargetNotFound{Bucket: bucket}
 	}
 	if reloaded {
-		globalBucketTargetSys.set(BucketInfo{
-			Name: bucket,
-		}, meta)
+		globalBucketTargetSys.set(bucket, meta)
 	}
 	return meta.bucketTargetConfig, nil
 }
@@ -455,13 +471,20 @@ func (sys *BucketMetadataSys) GetConfig(ctx context.Context, bucket string) (met
 	if ok {
 		return meta, reloaded, nil
 	}
-	meta, err = loadBucketMetadata(ctx, objAPI, bucket)
-	if err != nil {
-		if !sys.Initialized() {
-			// bucket metadata not yet initialized
-			return newBucketMetadata(bucket), reloaded, errBucketMetadataNotInitialized
+
+	val, err, _ := sys.group.Do(bucket, func() (val interface{}, err error) {
+		meta, err = loadBucketMetadata(ctx, objAPI, bucket)
+		if err != nil {
+			if !sys.Initialized() {
+				// bucket metadata not yet initialized
+				return newBucketMetadata(bucket), errBucketMetadataNotInitialized
+			}
 		}
-		return meta, reloaded, err
+		return meta, err
+	})
+	meta, _ = val.(BucketMetadata)
+	if err != nil {
+		return meta, false, err
 	}
 	sys.Lock()
 	sys.metadataMap[bucket] = meta
@@ -471,7 +494,7 @@ func (sys *BucketMetadataSys) GetConfig(ctx context.Context, bucket string) (met
 }
 
 // Init - initializes bucket metadata system for all buckets.
-func (sys *BucketMetadataSys) Init(ctx context.Context, buckets []BucketInfo, objAPI ObjectLayer) error {
+func (sys *BucketMetadataSys) Init(ctx context.Context, buckets []string, objAPI ObjectLayer) error {
 	if objAPI == nil {
 		return errServerNotInitialized
 	}
@@ -484,7 +507,7 @@ func (sys *BucketMetadataSys) Init(ctx context.Context, buckets []BucketInfo, ob
 }
 
 // concurrently load bucket metadata to speed up loading bucket metadata.
-func (sys *BucketMetadataSys) concurrentLoad(ctx context.Context, buckets []BucketInfo, failedBuckets map[string]struct{}) {
+func (sys *BucketMetadataSys) concurrentLoad(ctx context.Context, buckets []string) {
 	g := errgroup.WithNErrs(len(buckets))
 	bucketMetas := make([]BucketMetadata, len(buckets))
 	for index := range buckets {
@@ -494,8 +517,8 @@ func (sys *BucketMetadataSys) concurrentLoad(ctx context.Context, buckets []Buck
 			// herd upon start up sequence.
 			time.Sleep(25*time.Millisecond + time.Duration(rand.Int63n(int64(100*time.Millisecond))))
 
-			_, _ = sys.objAPI.HealBucket(ctx, buckets[index].Name, madmin.HealOpts{Recreate: true})
-			meta, err := loadBucketMetadata(ctx, sys.objAPI, buckets[index].Name)
+			_, _ = sys.objAPI.HealBucket(ctx, buckets[index], madmin.HealOpts{Recreate: true})
+			meta, err := loadBucketMetadata(ctx, sys.objAPI, buckets[index])
 			if err != nil {
 				return err
 			}
@@ -508,7 +531,7 @@ func (sys *BucketMetadataSys) concurrentLoad(ctx context.Context, buckets []Buck
 	for index, err := range errs {
 		if err != nil {
 			internalLogOnceIf(ctx, fmt.Errorf("Unable to load bucket metadata, will be retried: %w", err),
-				"load-bucket-metadata-"+buckets[index].Name, logger.WarningKind)
+				"load-bucket-metadata-"+buckets[index], logger.WarningKind)
 		}
 	}
 
@@ -519,16 +542,12 @@ func (sys *BucketMetadataSys) concurrentLoad(ctx context.Context, buckets []Buck
 		if errs[i] != nil {
 			continue
 		}
-		sys.metadataMap[buckets[i].Name] = meta
+		sys.metadataMap[buckets[i]] = meta
 	}
 	sys.Unlock()
 
 	for i, meta := range bucketMetas {
 		if errs[i] != nil {
-			if failedBuckets == nil {
-				failedBuckets = make(map[string]struct{})
-			}
-			failedBuckets[buckets[i].Name] = struct{}{}
 			continue
 		}
 		globalEventNotifier.set(buckets[i], meta)   // set notification targets
@@ -536,7 +555,7 @@ func (sys *BucketMetadataSys) concurrentLoad(ctx context.Context, buckets []Buck
 	}
 }
 
-func (sys *BucketMetadataSys) refreshBucketsMetadataLoop(ctx context.Context, failedBuckets map[string]struct{}) {
+func (sys *BucketMetadataSys) refreshBucketsMetadataLoop(ctx context.Context) {
 	const bucketMetadataRefresh = 15 * time.Minute
 
 	sleeper := newDynamicSleeper(2, 150*time.Millisecond, false)
@@ -548,7 +567,7 @@ func (sys *BucketMetadataSys) refreshBucketsMetadataLoop(ctx context.Context, fa
 		case <-ctx.Done():
 			return
 		case <-t.C:
-			buckets, err := sys.objAPI.ListBuckets(ctx, BucketOptions{})
+			buckets, err := sys.objAPI.ListBuckets(ctx, BucketOptions{NoMetadata: true})
 			if err != nil {
 				internalLogIf(ctx, err, logger.WarningKind)
 				break
@@ -566,7 +585,10 @@ func (sys *BucketMetadataSys) refreshBucketsMetadataLoop(ctx context.Context, fa
 			for i := range buckets {
 				wait := sleeper.Timer(ctx)
 
-				meta, err := loadBucketMetadata(ctx, sys.objAPI, buckets[i].Name)
+				bucket := buckets[i].Name
+				updated := false
+
+				meta, err := loadBucketMetadata(ctx, sys.objAPI, bucket)
 				if err != nil {
 					internalLogIf(ctx, err, logger.WarningKind)
 					wait() // wait to proceed to next entry.
@@ -574,14 +596,16 @@ func (sys *BucketMetadataSys) refreshBucketsMetadataLoop(ctx context.Context, fa
 				}
 
 				sys.Lock()
-				sys.metadataMap[buckets[i].Name] = meta
+				// Update if the bucket metadata in the memory is older than on-disk one
+				if lu := sys.metadataMap[bucket].lastUpdate(); lu.Before(meta.lastUpdate()) {
+					updated = true
+					sys.metadataMap[bucket] = meta
+				}
 				sys.Unlock()
 
-				// Initialize the failed buckets
-				if _, ok := failedBuckets[buckets[i].Name]; ok {
-					globalEventNotifier.set(buckets[i], meta)
-					globalBucketTargetSys.set(buckets[i], meta)
-					delete(failedBuckets, buckets[i].Name)
+				if updated {
+					globalEventNotifier.set(bucket, meta)
+					globalBucketTargetSys.set(bucket, meta)
 				}
 
 				wait() // wait to proceed to next entry.
@@ -600,15 +624,14 @@ func (sys *BucketMetadataSys) Initialized() bool {
 }
 
 // Loads bucket metadata for all buckets into BucketMetadataSys.
-func (sys *BucketMetadataSys) init(ctx context.Context, buckets []BucketInfo) {
-	count := 100 // load 100 bucket metadata at a time.
-	failedBuckets := make(map[string]struct{})
+func (sys *BucketMetadataSys) init(ctx context.Context, buckets []string) {
+	count := globalEndpoints.ESCount() * 10
 	for {
 		if len(buckets) < count {
-			sys.concurrentLoad(ctx, buckets, failedBuckets)
+			sys.concurrentLoad(ctx, buckets)
 			break
 		}
-		sys.concurrentLoad(ctx, buckets[:count], failedBuckets)
+		sys.concurrentLoad(ctx, buckets[:count])
 		buckets = buckets[count:]
 	}
 
@@ -617,7 +640,7 @@ func (sys *BucketMetadataSys) init(ctx context.Context, buckets []BucketInfo) {
 	sys.Unlock()
 
 	if globalIsDistErasure {
-		go sys.refreshBucketsMetadataLoop(ctx, failedBuckets)
+		go sys.refreshBucketsMetadataLoop(ctx)
 	}
 }
 
@@ -634,5 +657,6 @@ func (sys *BucketMetadataSys) Reset() {
 func NewBucketMetadataSys() *BucketMetadataSys {
 	return &BucketMetadataSys{
 		metadataMap: make(map[string]BucketMetadata),
+		group:       &singleflight.Group{},
 	}
 }

cmd/bucket-metadata.go

@@ -81,14 +81,19 @@ type BucketMetadata struct {
 	ReplicationConfigXML        []byte
 	BucketTargetsConfigJSON     []byte
 	BucketTargetsConfigMetaJSON []byte
-	PolicyConfigUpdatedAt       time.Time
-	ObjectLockConfigUpdatedAt   time.Time
-	EncryptionConfigUpdatedAt   time.Time
-	TaggingConfigUpdatedAt      time.Time
-	QuotaConfigUpdatedAt        time.Time
-	ReplicationConfigUpdatedAt  time.Time
-	VersioningConfigUpdatedAt   time.Time
-	LifecycleConfigUpdatedAt    time.Time
+
+	PolicyConfigUpdatedAt            time.Time
+	ObjectLockConfigUpdatedAt        time.Time
+	EncryptionConfigUpdatedAt        time.Time
+	TaggingConfigUpdatedAt           time.Time
+	QuotaConfigUpdatedAt             time.Time
+	ReplicationConfigUpdatedAt       time.Time
+	VersioningConfigUpdatedAt        time.Time
+	LifecycleConfigUpdatedAt         time.Time
+	NotificationConfigUpdatedAt      time.Time
+	BucketTargetsConfigUpdatedAt     time.Time
+	BucketTargetsConfigMetaUpdatedAt time.Time
+	// Add a new UpdatedAt field and update lastUpdate function
 
 	// Unexported fields. Must be updated atomically.
 	policyConfig           *policy.BucketPolicy
@@ -120,6 +125,46 @@ func newBucketMetadata(name string) BucketMetadata {
 	}
 }
 
+// Return the last update of this bucket metadata, which
+// means, the last update of any policy document.
+func (b BucketMetadata) lastUpdate() (t time.Time) {
+	if b.PolicyConfigUpdatedAt.After(t) {
+		t = b.PolicyConfigUpdatedAt
+	}
+	if b.ObjectLockConfigUpdatedAt.After(t) {
+		t = b.ObjectLockConfigUpdatedAt
+	}
+	if b.EncryptionConfigUpdatedAt.After(t) {
+		t = b.EncryptionConfigUpdatedAt
+	}
+	if b.TaggingConfigUpdatedAt.After(t) {
+		t = b.TaggingConfigUpdatedAt
+	}
+	if b.QuotaConfigUpdatedAt.After(t) {
+		t = b.QuotaConfigUpdatedAt
+	}
+	if b.ReplicationConfigUpdatedAt.After(t) {
+		t = b.ReplicationConfigUpdatedAt
+	}
+	if b.VersioningConfigUpdatedAt.After(t) {
+		t = b.VersioningConfigUpdatedAt
+	}
+	if b.LifecycleConfigUpdatedAt.After(t) {
+		t = b.LifecycleConfigUpdatedAt
+	}
+	if b.NotificationConfigUpdatedAt.After(t) {
+		t = b.NotificationConfigUpdatedAt
+	}
+	if b.BucketTargetsConfigUpdatedAt.After(t) {
+		t = b.BucketTargetsConfigUpdatedAt
+	}
+	if b.BucketTargetsConfigMetaUpdatedAt.After(t) {
+		t = b.BucketTargetsConfigMetaUpdatedAt
+	}
+
+	return
+}
+
 // Versioning returns true if versioning is enabled
 func (b BucketMetadata) Versioning() bool {
 	return b.LockEnabled || (b.versioningConfig != nil && b.versioningConfig.Enabled()) || (b.objectLockConfig != nil && b.objectLockConfig.Enabled())
@@ -440,6 +485,18 @@ func (b *BucketMetadata) defaultTimestamps() {
 	if b.LifecycleConfigUpdatedAt.IsZero() {
 		b.LifecycleConfigUpdatedAt = b.Created
 	}
+
+	if b.NotificationConfigUpdatedAt.IsZero() {
+		b.NotificationConfigUpdatedAt = b.Created
+	}
+
+	if b.BucketTargetsConfigUpdatedAt.IsZero() {
+		b.BucketTargetsConfigUpdatedAt = b.Created
+	}
+
+	if b.BucketTargetsConfigMetaUpdatedAt.IsZero() {
+		b.BucketTargetsConfigMetaUpdatedAt = b.Created
+	}
 }
 
 // Save config to supplied ObjectLayer api.

cmd/bucket-metadata_gen.go

@@ -156,6 +156,24 @@ func (z *BucketMetadata) DecodeMsg(dc *msgp.Reader) (err error) {
 				err = msgp.WrapError(err, "LifecycleConfigUpdatedAt")
 				return
 			}
+		case "NotificationConfigUpdatedAt":
+			z.NotificationConfigUpdatedAt, err = dc.ReadTime()
+			if err != nil {
+				err = msgp.WrapError(err, "NotificationConfigUpdatedAt")
+				return
+			}
+		case "BucketTargetsConfigUpdatedAt":
+			z.BucketTargetsConfigUpdatedAt, err = dc.ReadTime()
+			if err != nil {
+				err = msgp.WrapError(err, "BucketTargetsConfigUpdatedAt")
+				return
+			}
+		case "BucketTargetsConfigMetaUpdatedAt":
+			z.BucketTargetsConfigMetaUpdatedAt, err = dc.ReadTime()
+			if err != nil {
+				err = msgp.WrapError(err, "BucketTargetsConfigMetaUpdatedAt")
+				return
+			}
 		default:
 			err = dc.Skip()
 			if err != nil {
@@ -169,9 +187,9 @@ func (z *BucketMetadata) DecodeMsg(dc *msgp.Reader) (err error) {
 
 // EncodeMsg implements msgp.Encodable
 func (z *BucketMetadata) EncodeMsg(en *msgp.Writer) (err error) {
-	// map header, size 22
+	// map header, size 25
 	// write "Name"
-	err = en.Append(0xde, 0x0, 0x16, 0xa4, 0x4e, 0x61, 0x6d, 0x65)
+	err = en.Append(0xde, 0x0, 0x19, 0xa4, 0x4e, 0x61, 0x6d, 0x65)
 	if err != nil {
 		return
 	}
@@ -390,15 +408,45 @@ func (z *BucketMetadata) EncodeMsg(en *msgp.Writer) (err error) {
 		err = msgp.WrapError(err, "LifecycleConfigUpdatedAt")
 		return
 	}
+	// write "NotificationConfigUpdatedAt"
+	err = en.Append(0xbb, 0x4e, 0x6f, 0x74, 0x69, 0x66, 0x69, 0x63, 0x61, 0x74, 0x69, 0x6f, 0x6e, 0x43, 0x6f, 0x6e, 0x66, 0x69, 0x67, 0x55, 0x70, 0x64, 0x61, 0x74, 0x65, 0x64, 0x41, 0x74)
+	if err != nil {
+		return
+	}
+	err = en.WriteTime(z.NotificationConfigUpdatedAt)
+	if err != nil {
+		err = msgp.WrapError(err, "NotificationConfigUpdatedAt")
+		return
+	}
+	// write "BucketTargetsConfigUpdatedAt"
+	err = en.Append(0xbc, 0x42, 0x75, 0x63, 0x6b, 0x65, 0x74, 0x54, 0x61, 0x72, 0x67, 0x65, 0x74, 0x73, 0x43, 0x6f, 0x6e, 0x66, 0x69, 0x67, 0x55, 0x70, 0x64, 0x61, 0x74, 0x65, 0x64, 0x41, 0x74)
+	if err != nil {
+		return
+	}
+	err = en.WriteTime(z.BucketTargetsConfigUpdatedAt)
+	if err != nil {
+		err = msgp.WrapError(err, "BucketTargetsConfigUpdatedAt")
+		return
+	}
+	// write "BucketTargetsConfigMetaUpdatedAt"
+	err = en.Append(0xd9, 0x20, 0x42, 0x75, 0x63, 0x6b, 0x65, 0x74, 0x54, 0x61, 0x72, 0x67, 0x65, 0x74, 0x73, 0x43, 0x6f, 0x6e, 0x66, 0x69, 0x67, 0x4d, 0x65, 0x74, 0x61, 0x55, 0x70, 0x64, 0x61, 0x74, 0x65, 0x64, 0x41, 0x74)
+	if err != nil {
+		return
+	}
+	err = en.WriteTime(z.BucketTargetsConfigMetaUpdatedAt)
+	if err != nil {
+		err = msgp.WrapError(err, "BucketTargetsConfigMetaUpdatedAt")
+		return
+	}
 	return
 }
 
 // MarshalMsg implements msgp.Marshaler
 func (z *BucketMetadata) MarshalMsg(b []byte) (o []byte, err error) {
 	o = msgp.Require(b, z.Msgsize())
-	// map header, size 22
+	// map header, size 25
 	// string "Name"
-	o = append(o, 0xde, 0x0, 0x16, 0xa4, 0x4e, 0x61, 0x6d, 0x65)
+	o = append(o, 0xde, 0x0, 0x19, 0xa4, 0x4e, 0x61, 0x6d, 0x65)
 	o = msgp.AppendString(o, z.Name)
 	// string "Created"
 	o = append(o, 0xa7, 0x43, 0x72, 0x65, 0x61, 0x74, 0x65, 0x64)
@@ -463,6 +511,15 @@ func (z *BucketMetadata) MarshalMsg(b []byte) (o []byte, err error) {
 	// string "LifecycleConfigUpdatedAt"
 	o = append(o, 0xb8, 0x4c, 0x69, 0x66, 0x65, 0x63, 0x79, 0x63, 0x6c, 0x65, 0x43, 0x6f, 0x6e, 0x66, 0x69, 0x67, 0x55, 0x70, 0x64, 0x61, 0x74, 0x65, 0x64, 0x41, 0x74)
 	o = msgp.AppendTime(o, z.LifecycleConfigUpdatedAt)
+	// string "NotificationConfigUpdatedAt"
+	o = append(o, 0xbb, 0x4e, 0x6f, 0x74, 0x69, 0x66, 0x69, 0x63, 0x61, 0x74, 0x69, 0x6f, 0x6e, 0x43, 0x6f, 0x6e, 0x66, 0x69, 0x67, 0x55, 0x70, 0x64, 0x61, 0x74, 0x65, 0x64, 0x41, 0x74)
+	o = msgp.AppendTime(o, z.NotificationConfigUpdatedAt)
+	// string "BucketTargetsConfigUpdatedAt"
+	o = append(o, 0xbc, 0x42, 0x75, 0x63, 0x6b, 0x65, 0x74, 0x54, 0x61, 0x72, 0x67, 0x65, 0x74, 0x73, 0x43, 0x6f, 0x6e, 0x66, 0x69, 0x67, 0x55, 0x70, 0x64, 0x61, 0x74, 0x65, 0x64, 0x41, 0x74)
+	o = msgp.AppendTime(o, z.BucketTargetsConfigUpdatedAt)
+	// string "BucketTargetsConfigMetaUpdatedAt"
+	o = append(o, 0xd9, 0x20, 0x42, 0x75, 0x63, 0x6b, 0x65, 0x74, 0x54, 0x61, 0x72, 0x67, 0x65, 0x74, 0x73, 0x43, 0x6f, 0x6e, 0x66, 0x69, 0x67, 0x4d, 0x65, 0x74, 0x61, 0x55, 0x70, 0x64, 0x61, 0x74, 0x65, 0x64, 0x41, 0x74)
+	o = msgp.AppendTime(o, z.BucketTargetsConfigMetaUpdatedAt)
 	return
 }
 
@@ -616,6 +673,24 @@ func (z *BucketMetadata) UnmarshalMsg(bts []byte) (o []byte, err error) {
 				err = msgp.WrapError(err, "LifecycleConfigUpdatedAt")
 				return
 			}
+		case "NotificationConfigUpdatedAt":
+			z.NotificationConfigUpdatedAt, bts, err = msgp.ReadTimeBytes(bts)
+			if err != nil {
+				err = msgp.WrapError(err, "NotificationConfigUpdatedAt")
+				return
+			}
+		case "BucketTargetsConfigUpdatedAt":
+			z.BucketTargetsConfigUpdatedAt, bts, err = msgp.ReadTimeBytes(bts)
+			if err != nil {
+				err = msgp.WrapError(err, "BucketTargetsConfigUpdatedAt")
+				return
+			}
+		case "BucketTargetsConfigMetaUpdatedAt":
+			z.BucketTargetsConfigMetaUpdatedAt, bts, err = msgp.ReadTimeBytes(bts)
+			if err != nil {
+				err = msgp.WrapError(err, "BucketTargetsConfigMetaUpdatedAt")
+				return
+			}
 		default:
 			bts, err = msgp.Skip(bts)
 			if err != nil {
@@ -630,6 +705,6 @@ func (z *BucketMetadata) UnmarshalMsg(bts []byte) (o []byte, err error) {
 
 // Msgsize returns an upper bound estimate of the number of bytes occupied by the serialized message
 func (z *BucketMetadata) Msgsize() (s int) {
-	s = 3 + 5 + msgp.StringPrefixSize + len(z.Name) + 8 + msgp.TimeSize + 12 + msgp.BoolSize + 17 + msgp.BytesPrefixSize + len(z.PolicyConfigJSON) + 22 + msgp.BytesPrefixSize + len(z.NotificationConfigXML) + 19 + msgp.BytesPrefixSize + len(z.LifecycleConfigXML) + 20 + msgp.BytesPrefixSize + len(z.ObjectLockConfigXML) + 20 + msgp.BytesPrefixSize + len(z.VersioningConfigXML) + 20 + msgp.BytesPrefixSize + len(z.EncryptionConfigXML) + 17 + msgp.BytesPrefixSize + len(z.TaggingConfigXML) + 16 + msgp.BytesPrefixSize + len(z.QuotaConfigJSON) + 21 + msgp.BytesPrefixSize + len(z.ReplicationConfigXML) + 24 + msgp.BytesPrefixSize + len(z.BucketTargetsConfigJSON) + 28 + msgp.BytesPrefixSize + len(z.BucketTargetsConfigMetaJSON) + 22 + msgp.TimeSize + 26 + msgp.TimeSize + 26 + msgp.TimeSize + 23 + msgp.TimeSize + 21 + msgp.TimeSize + 27 + msgp.TimeSize + 26 + msgp.TimeSize + 25 + msgp.TimeSize
+	s = 3 + 5 + msgp.StringPrefixSize + len(z.Name) + 8 + msgp.TimeSize + 12 + msgp.BoolSize + 17 + msgp.BytesPrefixSize + len(z.PolicyConfigJSON) + 22 + msgp.BytesPrefixSize + len(z.NotificationConfigXML) + 19 + msgp.BytesPrefixSize + len(z.LifecycleConfigXML) + 20 + msgp.BytesPrefixSize + len(z.ObjectLockConfigXML) + 20 + msgp.BytesPrefixSize + len(z.VersioningConfigXML) + 20 + msgp.BytesPrefixSize + len(z.EncryptionConfigXML) + 17 + msgp.BytesPrefixSize + len(z.TaggingConfigXML) + 16 + msgp.BytesPrefixSize + len(z.QuotaConfigJSON) + 21 + msgp.BytesPrefixSize + len(z.ReplicationConfigXML) + 24 + msgp.BytesPrefixSize + len(z.BucketTargetsConfigJSON) + 28 + msgp.BytesPrefixSize + len(z.BucketTargetsConfigMetaJSON) + 22 + msgp.TimeSize + 26 + msgp.TimeSize + 26 + msgp.TimeSize + 23 + msgp.TimeSize + 21 + msgp.TimeSize + 27 + msgp.TimeSize + 26 + msgp.TimeSize + 25 + msgp.TimeSize + 28 + msgp.TimeSize + 29 + msgp.TimeSize + 34 + msgp.TimeSize
 	return
 }

cmd/bucket-quota.go

@@ -50,6 +50,9 @@ func (sys *BucketQuotaSys) Init(objAPI ObjectLayer) {
 	bucketStorageCache.InitOnce(10*time.Second,
 		cachevalue.Opts{ReturnLastGood: true, NoWait: true},
 		func(ctx context.Context) (DataUsageInfo, error) {
+			if objAPI == nil {
+				return DataUsageInfo{}, errServerNotInitialized
+			}
 			ctx, done := context.WithTimeout(ctx, 2*time.Second)
 			defer done()
 
@@ -59,7 +62,9 @@ func (sys *BucketQuotaSys) Init(objAPI ObjectLayer) {
 }
 
 // GetBucketUsageInfo return bucket usage info for a given bucket
-func (sys *BucketQuotaSys) GetBucketUsageInfo(ctx context.Context, bucket string) (BucketUsageInfo, error) {
+func (sys *BucketQuotaSys) GetBucketUsageInfo(ctx context.Context, bucket string) BucketUsageInfo {
+	sys.Init(newObjectLayerFn())
+
 	dui, err := bucketStorageCache.GetWithCtx(ctx)
 	timedout := OperationTimedOut{}
 	if err != nil && !errors.Is(err, context.DeadlineExceeded) && !errors.As(err, &timedout) {
@@ -73,10 +78,10 @@ func (sys *BucketQuotaSys) GetBucketUsageInfo(ctx context.Context, bucket string
 	if len(dui.BucketsUsage) > 0 {
 		bui, ok := dui.BucketsUsage[bucket]
 		if ok {
-			return bui, nil
+			return bui
 		}
 	}
-	return BucketUsageInfo{}, nil
+	return BucketUsageInfo{}
 }
 
 // parseBucketQuota parses BucketQuota from json
@@ -118,11 +123,7 @@ func (sys *BucketQuotaSys) enforceQuotaHard(ctx context.Context, bucket string,
 			return BucketQuotaExceeded{Bucket: bucket}
 		}
 
-		bui, err := sys.GetBucketUsageInfo(ctx, bucket)
-		if err != nil {
-			return err
-		}
-
+		bui := sys.GetBucketUsageInfo(ctx, bucket)
 		if bui.Size > 0 && ((bui.Size + uint64(size)) >= quotaSize) {
 			return BucketQuotaExceeded{Bucket: bucket}
 		}

cmd/bucket-replication-handlers.go

@@ -75,7 +75,7 @@ func (api objectAPIHandlers) PutBucketReplicationConfigHandler(w http.ResponseWr
 		writeErrorResponse(ctx, w, apiErr, r.URL)
 		return
 	}
-	sameTarget, apiErr := validateReplicationDestination(ctx, bucket, replicationConfig, true)
+	sameTarget, apiErr := validateReplicationDestination(ctx, bucket, replicationConfig, &validateReplicationDestinationOptions{CheckRemoteBucket: true})
 	if apiErr != noError {
 		writeErrorResponse(ctx, w, apiErr, r.URL)
 		return
@@ -230,7 +230,7 @@ func (api objectAPIHandlers) GetBucketReplicationMetricsHandler(w http.ResponseW
 	w.Header().Set(xhttp.ContentType, string(mimeJSON))
 
 	enc := json.NewEncoder(w)
-	stats := globalReplicationStats.getLatestReplicationStats(bucket)
+	stats := globalReplicationStats.Load().getLatestReplicationStats(bucket)
 	bwRpt := globalNotificationSys.GetBandwidthReports(ctx, bucket)
 	bwMap := bwRpt.BucketStats
 	for arn, st := range stats.ReplicationStats.Stats {
@@ -286,7 +286,7 @@ func (api objectAPIHandlers) GetBucketReplicationMetricsV2Handler(w http.Respons
 	w.Header().Set(xhttp.ContentType, string(mimeJSON))
 
 	enc := json.NewEncoder(w)
-	stats := globalReplicationStats.getLatestReplicationStats(bucket)
+	stats := globalReplicationStats.Load().getLatestReplicationStats(bucket)
 	bwRpt := globalNotificationSys.GetBandwidthReports(ctx, bucket)
 	bwMap := bwRpt.BucketStats
 	for arn, st := range stats.ReplicationStats.Stats {
@@ -422,7 +422,7 @@ func (api objectAPIHandlers) ResetBucketReplicationStartHandler(w http.ResponseW
 		return
 	}
 
-	if err := globalReplicationPool.resyncer.start(ctx, objectAPI, resyncOpts{
+	if err := globalReplicationPool.Get().resyncer.start(ctx, objectAPI, resyncOpts{
 		bucket:       bucket,
 		arn:          arn,
 		resyncID:     resetID,
@@ -559,7 +559,7 @@ func (api objectAPIHandlers) ValidateBucketReplicationCredsHandler(w http.Respon
 		lockEnabled = lcfg.Enabled()
 	}
 
-	sameTarget, apiErr := validateReplicationDestination(ctx, bucket, replicationConfig, true)
+	sameTarget, apiErr := validateReplicationDestination(ctx, bucket, replicationConfig, &validateReplicationDestinationOptions{CheckRemoteBucket: true})
 	if apiErr != noError {
 		writeErrorResponse(ctx, w, apiErr, r.URL)
 		return

cmd/bucket-replication-metrics.go

@@ -119,7 +119,7 @@ func (a *ActiveWorkerStat) update() {
 	if a == nil {
 		return
 	}
-	a.Curr = globalReplicationPool.ActiveWorkers()
+	a.Curr = globalReplicationPool.Get().ActiveWorkers()
 	a.hist.Update(int64(a.Curr))
 	a.Avg = float32(a.hist.Mean())
 	a.Max = int(a.hist.Max())

cmd/bucket-replication-stats.go

@@ -87,6 +87,9 @@ func (r *ReplicationStats) updateMovingAvg() {
 
 // ActiveWorkers returns worker stats
 func (r *ReplicationStats) ActiveWorkers() ActiveWorkerStat {
+	if r == nil {
+		return ActiveWorkerStat{}
+	}
 	r.wlock.RLock()
 	defer r.wlock.RUnlock()
 	w := r.workers.get()
@@ -351,6 +354,9 @@ func NewReplicationStats(ctx context.Context, objectAPI ObjectLayer) *Replicatio
 }
 
 func (r *ReplicationStats) getAllLatest(bucketsUsage map[string]BucketUsageInfo) (bucketsReplicationStats map[string]BucketStats) {
+	if r == nil {
+		return nil
+	}
 	peerBucketStatsList := globalNotificationSys.GetClusterAllBucketStats(GlobalContext)
 	bucketsReplicationStats = make(map[string]BucketStats, len(bucketsUsage))
 
@@ -460,6 +466,9 @@ func (r *ReplicationStats) calculateBucketReplicationStats(bucket string, bucket
 
 // get the most current of in-memory replication stats  and data usage info from crawler.
 func (r *ReplicationStats) getLatestReplicationStats(bucket string) (s BucketStats) {
+	if r == nil {
+		return s
+	}
 	bucketStats := globalNotificationSys.GetClusterBucketStats(GlobalContext, bucket)
 	return r.calculateBucketReplicationStats(bucket, bucketStats)
 }
@@ -495,9 +504,14 @@ func (r *ReplicationStats) decQ(bucket string, sz int64, isDelMarker bool, opTyp
 
 // incProxy increments proxy metrics for proxied calls
 func (r *ReplicationStats) incProxy(bucket string, api replProxyAPI, isErr bool) {
-	r.pCache.inc(bucket, api, isErr)
+	if r != nil {
+		r.pCache.inc(bucket, api, isErr)
+	}
 }
 
 func (r *ReplicationStats) getProxyStats(bucket string) ProxyMetric {
+	if r == nil {
+		return ProxyMetric{}
+	}
 	return r.pCache.getBucketStats(bucket)
 }

cmd/bucket-replication.go

@@ -50,6 +50,7 @@ import (
 	xhttp "github.com/minio/minio/internal/http"
 	xioutil "github.com/minio/minio/internal/ioutil"
 	"github.com/minio/minio/internal/logger"
+	"github.com/minio/minio/internal/once"
 	"github.com/tinylib/msgp/msgp"
 	"github.com/zeebo/xxh3"
 	"golang.org/x/exp/maps"
@@ -91,7 +92,10 @@ func getReplicationConfig(ctx context.Context, bucketName string) (rc *replicati
 
 // validateReplicationDestination returns error if replication destination bucket missing or not configured
 // It also returns true if replication destination is same as this server.
-func validateReplicationDestination(ctx context.Context, bucket string, rCfg *replication.Config, checkRemote bool) (bool, APIError) {
+func validateReplicationDestination(ctx context.Context, bucket string, rCfg *replication.Config, opts *validateReplicationDestinationOptions) (bool, APIError) {
+	if opts == nil {
+		opts = &validateReplicationDestinationOptions{}
+	}
 	var arns []string
 	if rCfg.RoleArn != "" {
 		arns = append(arns, rCfg.RoleArn)
@@ -113,7 +117,7 @@ func validateReplicationDestination(ctx context.Context, bucket string, rCfg *re
 		if clnt == nil {
 			return sameTarget, toAPIError(ctx, BucketRemoteTargetNotFound{Bucket: bucket})
 		}
-		if checkRemote { // validate remote bucket
+		if opts.CheckRemoteBucket { // validate remote bucket
 			found, err := clnt.BucketExists(ctx, arn.Bucket)
 			if err != nil {
 				return sameTarget, errorCodes.ToAPIErrWithErr(ErrRemoteDestinationNotFoundError, err)
@@ -133,23 +137,29 @@ func validateReplicationDestination(ctx context.Context, bucket string, rCfg *re
 				}
 			}
 		}
+		// if checked bucket, then check the ready is unnecessary
+		if !opts.CheckRemoteBucket && opts.CheckReady {
+			endpoint := clnt.EndpointURL().String()
+			if errInt, ok := opts.checkReadyErr.Load(endpoint); !ok {
+				err = checkRemoteEndpoint(ctx, clnt.EndpointURL())
+				opts.checkReadyErr.Store(endpoint, err)
+			} else {
+				if errInt == nil {
+					err = nil
+				} else {
+					err = errInt.(error)
+				}
+			}
+			switch err.(type) {
+			case BucketRemoteIdenticalToSource:
+				return true, errorCodes.ToAPIErrWithErr(ErrBucketRemoteIdenticalToSource, fmt.Errorf("remote target endpoint %s is self referential", clnt.EndpointURL().String()))
+			default:
+			}
+		}
 		// validate replication ARN against target endpoint
-		c := globalBucketTargetSys.GetRemoteTargetClient(bucket, arnStr)
-		if c != nil {
-			if err := checkRemoteEndpoint(ctx, c.EndpointURL()); err != nil {
-				switch err.(type) {
-				case BucketRemoteIdenticalToSource:
-					return true, errorCodes.ToAPIErrWithErr(ErrBucketRemoteIdenticalToSource, fmt.Errorf("remote target endpoint %s is self referential", c.EndpointURL().String()))
-				default:
-				}
-			}
-			if c.EndpointURL().String() == clnt.EndpointURL().String() {
-				selfTarget, _ := isLocalHost(clnt.EndpointURL().Hostname(), clnt.EndpointURL().Port(), globalMinioPort)
-				if !sameTarget {
-					sameTarget = selfTarget
-				}
-				continue
-			}
+		selfTarget, _ := isLocalHost(clnt.EndpointURL().Hostname(), clnt.EndpointURL().Port(), globalMinioPort)
+		if !sameTarget {
+			sameTarget = selfTarget
 		}
 	}
 
@@ -469,7 +479,7 @@ func replicateDelete(ctx context.Context, dobj DeletedObjectReplicationInfo, obj
 	lk := objectAPI.NewNSLock(bucket, "/[replicate]/"+dobj.ObjectName)
 	lkctx, err := lk.GetLock(ctx, globalOperationTimeout)
 	if err != nil {
-		globalReplicationPool.queueMRFSave(dobj.ToMRFEntry())
+		globalReplicationPool.Get().queueMRFSave(dobj.ToMRFEntry())
 		sendEvent(eventArgs{
 			BucketName: bucket,
 			Object: ObjectInfo{
@@ -539,7 +549,7 @@ func replicateDelete(ctx context.Context, dobj DeletedObjectReplicationInfo, obj
 	// to decrement pending count later.
 	for _, rinfo := range rinfos.Targets {
 		if rinfo.ReplicationStatus != rinfo.PrevReplicationStatus {
-			globalReplicationStats.Update(dobj.Bucket, rinfo, replicationStatus,
+			globalReplicationStats.Load().Update(dobj.Bucket, rinfo, replicationStatus,
 				prevStatus)
 		}
 	}
@@ -547,7 +557,7 @@ func replicateDelete(ctx context.Context, dobj DeletedObjectReplicationInfo, obj
 	eventName := event.ObjectReplicationComplete
 	if replicationStatus == replication.Failed {
 		eventName = event.ObjectReplicationFailed
-		globalReplicationPool.queueMRFSave(dobj.ToMRFEntry())
+		globalReplicationPool.Get().queueMRFSave(dobj.ToMRFEntry())
 	}
 	drs := getReplicationState(rinfos, dobj.ReplicationState, dobj.VersionID)
 	if replicationStatus != prevStatus {
@@ -765,21 +775,6 @@ func (m caseInsensitiveMap) Lookup(key string) (string, bool) {
 	return "", false
 }
 
-func getCRCMeta(oi ObjectInfo, partNum int, h http.Header) map[string]string {
-	meta := make(map[string]string)
-	cs := oi.decryptChecksums(partNum, h)
-	for k, v := range cs {
-		cksum := hash.NewChecksumString(k, v)
-		if cksum == nil {
-			continue
-		}
-		if cksum.Valid() {
-			meta[cksum.Type.Key()] = v
-		}
-	}
-	return meta
-}
-
 func putReplicationOpts(ctx context.Context, sc string, objInfo ObjectInfo, partNum int) (putOpts minio.PutObjectOptions, err error) {
 	meta := make(map[string]string)
 	isSSEC := crypto.SSEC.IsEncrypted(objInfo.UserDefined)
@@ -788,11 +783,6 @@ func putReplicationOpts(ctx context.Context, sc string, objInfo ObjectInfo, part
 		// In case of SSE-C objects copy the allowed internal headers as well
 		if !isSSEC || !slices.Contains(maps.Keys(validSSEReplicationHeaders), k) {
 			if stringsHasPrefixFold(k, ReservedMetadataPrefixLower) {
-				if strings.EqualFold(k, ReservedMetadataPrefixLower+"crc") {
-					for k, v := range getCRCMeta(objInfo, partNum, nil) {
-						meta[k] = v
-					}
-				}
 				continue
 			}
 			if isStandardHeader(k) {
@@ -811,8 +801,12 @@ func putReplicationOpts(ctx context.Context, sc string, objInfo ObjectInfo, part
 		if isSSEC {
 			meta[ReplicationSsecChecksumHeader] = base64.StdEncoding.EncodeToString(objInfo.Checksum)
 		} else {
-			for k, v := range getCRCMeta(objInfo, 0, nil) {
-				meta[k] = v
+			for _, pi := range objInfo.Parts {
+				if pi.Number == partNum {
+					for k, v := range pi.Checksums {
+						meta[k] = v
+					}
+				}
 			}
 		}
 	}
@@ -1061,7 +1055,7 @@ func replicateObject(ctx context.Context, ri ReplicateObjectInfo, objectAPI Obje
 			UserAgent:  "Internal: [Replication]",
 			Host:       globalLocalNodeName,
 		})
-		globalReplicationPool.queueMRFSave(ri.ToMRFEntry())
+		globalReplicationPool.Get().queueMRFSave(ri.ToMRFEntry())
 		return
 	}
 	ctx = lkctx.Context()
@@ -1146,7 +1140,7 @@ func replicateObject(ctx context.Context, ri ReplicateObjectInfo, objectAPI Obje
 		for _, rinfo := range rinfos.Targets {
 			if rinfo.ReplicationStatus != rinfo.PrevReplicationStatus {
 				rinfo.OpType = opType // update optype to reflect correct operation.
-				globalReplicationStats.Update(bucket, rinfo, rinfo.ReplicationStatus, rinfo.PrevReplicationStatus)
+				globalReplicationStats.Load().Update(bucket, rinfo, rinfo.ReplicationStatus, rinfo.PrevReplicationStatus)
 			}
 		}
 	}
@@ -1166,7 +1160,7 @@ func replicateObject(ctx context.Context, ri ReplicateObjectInfo, objectAPI Obje
 		ri.EventType = ReplicateMRF
 		ri.ReplicationStatusInternal = rinfos.ReplicationStatusInternal()
 		ri.RetryCount++
-		globalReplicationPool.queueMRFSave(ri.ToMRFEntry())
+		globalReplicationPool.Get().queueMRFSave(ri.ToMRFEntry())
 	}
 }
 
@@ -1306,7 +1300,7 @@ func (ri ReplicateObjectInfo) replicateObject(ctx context.Context, objectAPI Obj
 		HeaderSize: headerSize,
 	}
 	newCtx := ctx
-	if globalBucketMonitor.IsThrottled(bucket, tgt.ARN) {
+	if globalBucketMonitor.IsThrottled(bucket, tgt.ARN) && objInfo.Size < minLargeObjSize {
 		var cancel context.CancelFunc
 		newCtx, cancel = context.WithTimeout(ctx, throttleDeadline)
 		defer cancel()
@@ -1581,7 +1575,7 @@ applyAction:
 			HeaderSize: headerSize,
 		}
 		newCtx := ctx
-		if globalBucketMonitor.IsThrottled(bucket, tgt.ARN) {
+		if globalBucketMonitor.IsThrottled(bucket, tgt.ARN) && objInfo.Size < minLargeObjSize {
 			var cancel context.CancelFunc
 			newCtx, cancel = context.WithTimeout(ctx, throttleDeadline)
 			defer cancel()
@@ -1648,7 +1642,6 @@ func replicateObjectWithMultipart(ctx context.Context, c *minio.Core, bucket, ob
 
 	var (
 		hr     *hash.Reader
-		pInfo  minio.ObjectPart
 		isSSEC = crypto.SSEC.IsEncrypted(objInfo.UserDefined)
 	)
 
@@ -1666,8 +1659,7 @@ func replicateObjectWithMultipart(ctx context.Context, c *minio.Core, bucket, ob
 		cHeader := http.Header{}
 		cHeader.Add(xhttp.MinIOSourceReplicationRequest, "true")
 		if !isSSEC {
-			crc := getCRCMeta(objInfo, partInfo.Number, nil) // No SSE-C keys here.
-			for k, v := range crc {
+			for k, v := range partInfo.Checksums {
 				cHeader.Add(k, v)
 			}
 		}
@@ -1676,18 +1668,19 @@ func replicateObjectWithMultipart(ctx context.Context, c *minio.Core, bucket, ob
 			CustomHeader: cHeader,
 		}
 
+		var size int64
 		if isSSEC {
-			objectSize += partInfo.Size
-			pInfo, err = c.PutObjectPart(ctx, bucket, object, uploadID, partInfo.Number, hr, partInfo.Size, popts)
+			size = partInfo.Size
 		} else {
-			objectSize += partInfo.ActualSize
-			pInfo, err = c.PutObjectPart(ctx, bucket, object, uploadID, partInfo.Number, hr, partInfo.ActualSize, popts)
+			size = partInfo.ActualSize
 		}
+		objectSize += size
+		pInfo, err := c.PutObjectPart(ctx, bucket, object, uploadID, partInfo.Number, hr, size, popts)
 		if err != nil {
 			return err
 		}
-		if !isSSEC && pInfo.Size != partInfo.ActualSize {
-			return fmt.Errorf("Part size mismatch: got %d, want %d", pInfo.Size, partInfo.ActualSize)
+		if pInfo.Size != size {
+			return fmt.Errorf("ssec(%t): Part size mismatch: got %d, want %d", isSSEC, pInfo.Size, size)
 		}
 		uploadedParts = append(uploadedParts, minio.CompletePart{
 			PartNumber:     pInfo.PartNumber,
@@ -1699,7 +1692,7 @@ func replicateObjectWithMultipart(ctx context.Context, c *minio.Core, bucket, ob
 		})
 	}
 	userMeta := map[string]string{
-		validSSEReplicationHeaders[ReservedMetadataPrefix+"Actual-Object-Size"]: objInfo.UserDefined[ReservedMetadataPrefix+"actual-size"],
+		xhttp.MinIOReplicationActualObjectSize: objInfo.UserDefined[ReservedMetadataPrefix+"actual-size"],
 	}
 	if isSSEC && objInfo.UserDefined[ReplicationSsecChecksumHeader] != "" {
 		userMeta[ReplicationSsecChecksumHeader] = objInfo.UserDefined[ReplicationSsecChecksumHeader]
@@ -1795,23 +1788,27 @@ const (
 )
 
 var (
-	globalReplicationPool  *ReplicationPool
-	globalReplicationStats *ReplicationStats
+	globalReplicationPool  = once.NewSingleton[ReplicationPool]()
+	globalReplicationStats atomic.Pointer[ReplicationStats]
 )
 
 // ReplicationPool describes replication pool
 type ReplicationPool struct {
 	// atomic ops:
 	activeWorkers    int32
+	activeLrgWorkers int32
 	activeMRFWorkers int32
 
-	objLayer   ObjectLayer
-	ctx        context.Context
-	priority   string
-	maxWorkers int
-	mu         sync.RWMutex
-	mrfMU      sync.Mutex
-	resyncer   *replicationResyncer
+	objLayer    ObjectLayer
+	ctx         context.Context
+	priority    string
+	maxWorkers  int
+	maxLWorkers int
+	stats       *ReplicationStats
+
+	mu       sync.RWMutex
+	mrfMU    sync.Mutex
+	resyncer *replicationResyncer
 
 	// workers:
 	workers    []chan ReplicationWorkerOperation
@@ -1854,7 +1851,7 @@ const (
 )
 
 // NewReplicationPool creates a pool of replication workers of specified size
-func NewReplicationPool(ctx context.Context, o ObjectLayer, opts replicationPoolOpts) *ReplicationPool {
+func NewReplicationPool(ctx context.Context, o ObjectLayer, opts replicationPoolOpts, stats *ReplicationStats) *ReplicationPool {
 	var workers, failedWorkers int
 	priority := "auto"
 	maxWorkers := WorkerMaxLimit
@@ -1882,9 +1879,13 @@ func NewReplicationPool(ctx context.Context, o ObjectLayer, opts replicationPool
 	if maxWorkers > 0 && failedWorkers > maxWorkers {
 		failedWorkers = maxWorkers
 	}
+	maxLWorkers := LargeWorkerCount
+	if opts.MaxLWorkers > 0 {
+		maxLWorkers = opts.MaxLWorkers
+	}
 	pool := &ReplicationPool{
 		workers:         make([]chan ReplicationWorkerOperation, 0, workers),
-		lrgworkers:      make([]chan ReplicationWorkerOperation, 0, LargeWorkerCount),
+		lrgworkers:      make([]chan ReplicationWorkerOperation, 0, maxLWorkers),
 		mrfReplicaCh:    make(chan ReplicationWorkerOperation, 100000),
 		mrfWorkerKillCh: make(chan struct{}, failedWorkers),
 		resyncer:        newresyncer(),
@@ -1892,11 +1893,13 @@ func NewReplicationPool(ctx context.Context, o ObjectLayer, opts replicationPool
 		mrfStopCh:       make(chan struct{}, 1),
 		ctx:             ctx,
 		objLayer:        o,
+		stats:           stats,
 		priority:        priority,
 		maxWorkers:      maxWorkers,
+		maxLWorkers:     maxLWorkers,
 	}
 
-	pool.AddLargeWorkers()
+	pool.ResizeLrgWorkers(maxLWorkers, 0)
 	pool.ResizeWorkers(workers, 0)
 	pool.ResizeFailedWorkers(failedWorkers)
 	go pool.resyncer.PersistToDisk(ctx, o)
@@ -1918,11 +1921,11 @@ func (p *ReplicationPool) AddMRFWorker() {
 			}
 			switch v := oi.(type) {
 			case ReplicateObjectInfo:
-				globalReplicationStats.incQ(v.Bucket, v.Size, v.DeleteMarker, v.OpType)
+				p.stats.incQ(v.Bucket, v.Size, v.DeleteMarker, v.OpType)
 				atomic.AddInt32(&p.activeMRFWorkers, 1)
 				replicateObject(p.ctx, v, p.objLayer)
 				atomic.AddInt32(&p.activeMRFWorkers, -1)
-				globalReplicationStats.decQ(v.Bucket, v.Size, v.DeleteMarker, v.OpType)
+				p.stats.decQ(v.Bucket, v.Size, v.DeleteMarker, v.OpType)
 
 			default:
 				bugLogIf(p.ctx, fmt.Errorf("unknown mrf replication type: %T", oi), "unknown-mrf-replicate-type")
@@ -1950,9 +1953,9 @@ func (p *ReplicationPool) AddWorker(input <-chan ReplicationWorkerOperation, opT
 				if opTracker != nil {
 					atomic.AddInt32(opTracker, 1)
 				}
-				globalReplicationStats.incQ(v.Bucket, v.Size, v.DeleteMarker, v.OpType)
+				p.stats.incQ(v.Bucket, v.Size, v.DeleteMarker, v.OpType)
 				replicateObject(p.ctx, v, p.objLayer)
-				globalReplicationStats.decQ(v.Bucket, v.Size, v.DeleteMarker, v.OpType)
+				p.stats.decQ(v.Bucket, v.Size, v.DeleteMarker, v.OpType)
 				if opTracker != nil {
 					atomic.AddInt32(opTracker, -1)
 				}
@@ -1960,10 +1963,10 @@ func (p *ReplicationPool) AddWorker(input <-chan ReplicationWorkerOperation, opT
 				if opTracker != nil {
 					atomic.AddInt32(opTracker, 1)
 				}
-				globalReplicationStats.incQ(v.Bucket, 0, true, v.OpType)
+				p.stats.incQ(v.Bucket, 0, true, v.OpType)
 
 				replicateDelete(p.ctx, v, p.objLayer)
-				globalReplicationStats.decQ(v.Bucket, 0, true, v.OpType)
+				p.stats.decQ(v.Bucket, 0, true, v.OpType)
 
 				if opTracker != nil {
 					atomic.AddInt32(opTracker, -1)
@@ -1975,23 +1978,8 @@ func (p *ReplicationPool) AddWorker(input <-chan ReplicationWorkerOperation, opT
 	}
 }
 
-// AddLargeWorkers adds a static number of workers to handle large uploads
-func (p *ReplicationPool) AddLargeWorkers() {
-	for i := 0; i < LargeWorkerCount; i++ {
-		p.lrgworkers = append(p.lrgworkers, make(chan ReplicationWorkerOperation, 100000))
-		i := i
-		go p.AddLargeWorker(p.lrgworkers[i])
-	}
-	go func() {
-		<-p.ctx.Done()
-		for i := 0; i < LargeWorkerCount; i++ {
-			xioutil.SafeClose(p.lrgworkers[i])
-		}
-	}()
-}
-
 // AddLargeWorker adds a replication worker to the static pool for large uploads.
-func (p *ReplicationPool) AddLargeWorker(input <-chan ReplicationWorkerOperation) {
+func (p *ReplicationPool) AddLargeWorker(input <-chan ReplicationWorkerOperation, opTracker *int32) {
 	for {
 		select {
 		case <-p.ctx.Done():
@@ -2002,11 +1990,23 @@ func (p *ReplicationPool) AddLargeWorker(input <-chan ReplicationWorkerOperation
 			}
 			switch v := oi.(type) {
 			case ReplicateObjectInfo:
-				globalReplicationStats.incQ(v.Bucket, v.Size, v.DeleteMarker, v.OpType)
+				if opTracker != nil {
+					atomic.AddInt32(opTracker, 1)
+				}
+				p.stats.incQ(v.Bucket, v.Size, v.DeleteMarker, v.OpType)
 				replicateObject(p.ctx, v, p.objLayer)
-				globalReplicationStats.decQ(v.Bucket, v.Size, v.DeleteMarker, v.OpType)
+				p.stats.decQ(v.Bucket, v.Size, v.DeleteMarker, v.OpType)
+				if opTracker != nil {
+					atomic.AddInt32(opTracker, -1)
+				}
 			case DeletedObjectReplicationInfo:
+				if opTracker != nil {
+					atomic.AddInt32(opTracker, 1)
+				}
 				replicateDelete(p.ctx, v, p.objLayer)
+				if opTracker != nil {
+					atomic.AddInt32(opTracker, -1)
+				}
 			default:
 				bugLogIf(p.ctx, fmt.Errorf("unknown replication type: %T", oi), "unknown-replicate-type")
 			}
@@ -2014,6 +2014,30 @@ func (p *ReplicationPool) AddLargeWorker(input <-chan ReplicationWorkerOperation
 	}
 }
 
+// ResizeLrgWorkers sets replication workers pool for large transfers(>=128MiB) to new size.
+// checkOld can be set to an expected value.
+// If the worker count changed
+func (p *ReplicationPool) ResizeLrgWorkers(n, checkOld int) {
+	p.mu.Lock()
+	defer p.mu.Unlock()
+
+	if (checkOld > 0 && len(p.lrgworkers) != checkOld) || n == len(p.lrgworkers) || n < 1 {
+		// Either already satisfied or worker count changed while we waited for the lock.
+		return
+	}
+	for len(p.lrgworkers) < n {
+		input := make(chan ReplicationWorkerOperation, 100000)
+		p.lrgworkers = append(p.lrgworkers, input)
+
+		go p.AddLargeWorker(input, &p.activeLrgWorkers)
+	}
+	for len(p.lrgworkers) > n {
+		worker := p.lrgworkers[len(p.lrgworkers)-1]
+		p.lrgworkers = p.lrgworkers[:len(p.lrgworkers)-1]
+		xioutil.SafeClose(worker)
+	}
+}
+
 // ActiveWorkers returns the number of active workers handling replication traffic.
 func (p *ReplicationPool) ActiveWorkers() int {
 	return int(atomic.LoadInt32(&p.activeWorkers))
@@ -2024,6 +2048,11 @@ func (p *ReplicationPool) ActiveMRFWorkers() int {
 	return int(atomic.LoadInt32(&p.activeMRFWorkers))
 }
 
+// ActiveLrgWorkers returns the number of active workers handling traffic > 128MiB object size.
+func (p *ReplicationPool) ActiveLrgWorkers() int {
+	return int(atomic.LoadInt32(&p.activeLrgWorkers))
+}
+
 // ResizeWorkers sets replication workers pool to new size.
 // checkOld can be set to an expected value.
 // If the worker count changed
@@ -2049,7 +2078,7 @@ func (p *ReplicationPool) ResizeWorkers(n, checkOld int) {
 }
 
 // ResizeWorkerPriority sets replication failed workers pool size
-func (p *ReplicationPool) ResizeWorkerPriority(pri string, maxWorkers int) {
+func (p *ReplicationPool) ResizeWorkerPriority(pri string, maxWorkers, maxLWorkers int) {
 	var workers, mrfWorkers int
 	p.mu.Lock()
 	switch pri {
@@ -2076,11 +2105,15 @@ func (p *ReplicationPool) ResizeWorkerPriority(pri string, maxWorkers int) {
 	if maxWorkers > 0 && mrfWorkers > maxWorkers {
 		mrfWorkers = maxWorkers
 	}
+	if maxLWorkers <= 0 {
+		maxLWorkers = LargeWorkerCount
+	}
 	p.priority = pri
 	p.maxWorkers = maxWorkers
 	p.mu.Unlock()
 	p.ResizeWorkers(workers, 0)
 	p.ResizeFailedWorkers(mrfWorkers)
+	p.ResizeLrgWorkers(maxLWorkers, 0)
 }
 
 // ResizeFailedWorkers sets replication failed workers pool size
@@ -2124,9 +2157,18 @@ func (p *ReplicationPool) queueReplicaTask(ri ReplicateObjectInfo) {
 		h := xxh3.HashString(ri.Bucket + ri.Name)
 		select {
 		case <-p.ctx.Done():
-		case p.lrgworkers[h%LargeWorkerCount] <- ri:
+		case p.lrgworkers[h%uint64(len(p.lrgworkers))] <- ri:
 		default:
-			globalReplicationPool.queueMRFSave(ri.ToMRFEntry())
+			p.queueMRFSave(ri.ToMRFEntry())
+			p.mu.RLock()
+			maxLWorkers := p.maxLWorkers
+			existing := len(p.lrgworkers)
+			p.mu.RUnlock()
+			maxLWorkers = min(maxLWorkers, LargeWorkerCount)
+			if p.ActiveLrgWorkers() < maxLWorkers {
+				workers := min(existing+1, maxLWorkers)
+				p.ResizeLrgWorkers(workers, existing)
+			}
 		}
 		return
 	}
@@ -2148,7 +2190,7 @@ func (p *ReplicationPool) queueReplicaTask(ri ReplicateObjectInfo) {
 	case healCh <- ri:
 	case ch <- ri:
 	default:
-		globalReplicationPool.queueMRFSave(ri.ToMRFEntry())
+		globalReplicationPool.Get().queueMRFSave(ri.ToMRFEntry())
 		p.mu.RLock()
 		prio := p.priority
 		maxWorkers := p.maxWorkers
@@ -2184,7 +2226,7 @@ func queueReplicateDeletesWrapper(doi DeletedObjectReplicationInfo, existingObje
 			doi.ResetID = v.ResetID
 			doi.TargetArn = k
 
-			globalReplicationPool.queueReplicaDeleteTask(doi)
+			globalReplicationPool.Get().queueReplicaDeleteTask(doi)
 		}
 	}
 }
@@ -2205,7 +2247,7 @@ func (p *ReplicationPool) queueReplicaDeleteTask(doi DeletedObjectReplicationInf
 	case <-p.ctx.Done():
 	case ch <- doi:
 	default:
-		globalReplicationPool.queueMRFSave(doi.ToMRFEntry())
+		p.queueMRFSave(doi.ToMRFEntry())
 		p.mu.RLock()
 		prio := p.priority
 		maxWorkers := p.maxWorkers
@@ -2229,14 +2271,16 @@ func (p *ReplicationPool) queueReplicaDeleteTask(doi DeletedObjectReplicationInf
 }
 
 type replicationPoolOpts struct {
-	Priority   string
-	MaxWorkers int
+	Priority    string
+	MaxWorkers  int
+	MaxLWorkers int
 }
 
 func initBackgroundReplication(ctx context.Context, objectAPI ObjectLayer) {
-	globalReplicationPool = NewReplicationPool(ctx, objectAPI, globalAPIConfig.getReplicationOpts())
-	globalReplicationStats = NewReplicationStats(ctx, objectAPI)
-	go globalReplicationStats.trackEWMA()
+	stats := NewReplicationStats(ctx, objectAPI)
+	globalReplicationPool.Set(NewReplicationPool(ctx, objectAPI, globalAPIConfig.getReplicationOpts(), stats))
+	globalReplicationStats.Store(stats)
+	go stats.trackEWMA()
 }
 
 type proxyResult struct {
@@ -2442,7 +2486,7 @@ func scheduleReplication(ctx context.Context, oi ObjectInfo, o ObjectLayer, dsc
 	if dsc.Synchronous() {
 		replicateObject(ctx, ri, o)
 	} else {
-		globalReplicationPool.queueReplicaTask(ri)
+		globalReplicationPool.Get().queueReplicaTask(ri)
 	}
 }
 
@@ -2570,9 +2614,9 @@ func proxyGetTaggingToRepTarget(ctx context.Context, bucket, object string, opts
 }
 
 func scheduleReplicationDelete(ctx context.Context, dv DeletedObjectReplicationInfo, o ObjectLayer) {
-	globalReplicationPool.queueReplicaDeleteTask(dv)
+	globalReplicationPool.Get().queueReplicaDeleteTask(dv)
 	for arn := range dv.ReplicationState.Targets {
-		globalReplicationStats.Update(dv.Bucket, replicatedTargetInfo{Arn: arn, Size: 0, Duration: 0, OpType: replication.DeleteReplicationType}, replication.Pending, replication.StatusType(""))
+		globalReplicationStats.Load().Update(dv.Bucket, replicatedTargetInfo{Arn: arn, Size: 0, Duration: 0, OpType: replication.DeleteReplicationType}, replication.Pending, replication.StatusType(""))
 	}
 }
 
@@ -3002,9 +3046,9 @@ func (s *replicationResyncer) start(ctx context.Context, objAPI ObjectLayer, opt
 	if len(tgtArns) == 0 {
 		return fmt.Errorf("arn %s specified for resync not found in replication config", opts.arn)
 	}
-	globalReplicationPool.resyncer.RLock()
-	data, ok := globalReplicationPool.resyncer.statusMap[opts.bucket]
-	globalReplicationPool.resyncer.RUnlock()
+	globalReplicationPool.Get().resyncer.RLock()
+	data, ok := globalReplicationPool.Get().resyncer.statusMap[opts.bucket]
+	globalReplicationPool.Get().resyncer.RUnlock()
 	if !ok {
 		data, err = loadBucketResyncMetadata(ctx, opts.bucket, objAPI)
 		if err != nil {
@@ -3030,9 +3074,9 @@ func (s *replicationResyncer) start(ctx context.Context, objAPI ObjectLayer, opt
 		return err
 	}
 
-	globalReplicationPool.resyncer.Lock()
-	defer globalReplicationPool.resyncer.Unlock()
-	brs, ok := globalReplicationPool.resyncer.statusMap[opts.bucket]
+	globalReplicationPool.Get().resyncer.Lock()
+	defer globalReplicationPool.Get().resyncer.Unlock()
+	brs, ok := globalReplicationPool.Get().resyncer.statusMap[opts.bucket]
 	if !ok {
 		brs = BucketReplicationResyncStatus{
 			Version:    resyncMetaVersion,
@@ -3040,8 +3084,8 @@ func (s *replicationResyncer) start(ctx context.Context, objAPI ObjectLayer, opt
 		}
 	}
 	brs.TargetsMap[opts.arn] = status
-	globalReplicationPool.resyncer.statusMap[opts.bucket] = brs
-	go globalReplicationPool.resyncer.resyncBucket(GlobalContext, objAPI, false, opts)
+	globalReplicationPool.Get().resyncer.statusMap[opts.bucket] = brs
+	go globalReplicationPool.Get().resyncer.resyncBucket(GlobalContext, objAPI, false, opts)
 	return nil
 }
 
@@ -3086,7 +3130,7 @@ func (p *ReplicationPool) deleteResyncMetadata(ctx context.Context, bucket strin
 }
 
 // initResync - initializes bucket replication resync for all buckets.
-func (p *ReplicationPool) initResync(ctx context.Context, buckets []BucketInfo, objAPI ObjectLayer) error {
+func (p *ReplicationPool) initResync(ctx context.Context, buckets []string, objAPI ObjectLayer) error {
 	if objAPI == nil {
 		return errServerNotInitialized
 	}
@@ -3095,7 +3139,7 @@ func (p *ReplicationPool) initResync(ctx context.Context, buckets []BucketInfo,
 	return nil
 }
 
-func (p *ReplicationPool) startResyncRoutine(ctx context.Context, buckets []BucketInfo, objAPI ObjectLayer) {
+func (p *ReplicationPool) startResyncRoutine(ctx context.Context, buckets []string, objAPI ObjectLayer) {
 	r := rand.New(rand.NewSource(time.Now().UnixNano()))
 	// Run the replication resync in a loop
 	for {
@@ -3113,13 +3157,13 @@ func (p *ReplicationPool) startResyncRoutine(ctx context.Context, buckets []Buck
 }
 
 // Loads bucket replication resync statuses into memory.
-func (p *ReplicationPool) loadResync(ctx context.Context, buckets []BucketInfo, objAPI ObjectLayer) error {
+func (p *ReplicationPool) loadResync(ctx context.Context, buckets []string, objAPI ObjectLayer) error {
 	// Make sure only one node running resync on the cluster.
 	ctx, cancel := globalLeaderLock.GetLock(ctx)
 	defer cancel()
 
 	for index := range buckets {
-		bucket := buckets[index].Name
+		bucket := buckets[index]
 
 		meta, err := loadBucketResyncMetadata(ctx, bucket, objAPI)
 		if err != nil {
@@ -3373,7 +3417,7 @@ func queueReplicationHeal(ctx context.Context, bucket string, oi ObjectInfo, rcf
 		if roi.ReplicationStatus == replication.Pending ||
 			roi.ReplicationStatus == replication.Failed ||
 			roi.VersionPurgeStatus == Failed || roi.VersionPurgeStatus == Pending {
-			globalReplicationPool.queueReplicaDeleteTask(dv)
+			globalReplicationPool.Get().queueReplicaDeleteTask(dv)
 			return
 		}
 		// if replication status is Complete on DeleteMarker and existing object resync required
@@ -3389,12 +3433,12 @@ func queueReplicationHeal(ctx context.Context, bucket string, oi ObjectInfo, rcf
 	switch roi.ReplicationStatus {
 	case replication.Pending, replication.Failed:
 		roi.EventType = ReplicateHeal
-		globalReplicationPool.queueReplicaTask(roi)
+		globalReplicationPool.Get().queueReplicaTask(roi)
 		return
 	}
 	if roi.ExistingObjResync.mustResync() {
 		roi.EventType = ReplicateExisting
-		globalReplicationPool.queueReplicaTask(roi)
+		globalReplicationPool.Get().queueReplicaTask(roi)
 	}
 	return
 }
@@ -3459,8 +3503,8 @@ func (p *ReplicationPool) queueMRFSave(entry MRFReplicateEntry) {
 		return
 	}
 	if entry.RetryCount > mrfRetryLimit { // let scanner catch up if retry count exceeded
-		atomic.AddUint64(&globalReplicationStats.mrfStats.TotalDroppedCount, 1)
-		atomic.AddUint64(&globalReplicationStats.mrfStats.TotalDroppedBytes, uint64(entry.sz))
+		atomic.AddUint64(&p.stats.mrfStats.TotalDroppedCount, 1)
+		atomic.AddUint64(&p.stats.mrfStats.TotalDroppedBytes, uint64(entry.sz))
 		return
 	}
 
@@ -3473,8 +3517,8 @@ func (p *ReplicationPool) queueMRFSave(entry MRFReplicateEntry) {
 		select {
 		case p.mrfSaveCh <- entry:
 		default:
-			atomic.AddUint64(&globalReplicationStats.mrfStats.TotalDroppedCount, 1)
-			atomic.AddUint64(&globalReplicationStats.mrfStats.TotalDroppedBytes, uint64(entry.sz))
+			atomic.AddUint64(&p.stats.mrfStats.TotalDroppedCount, 1)
+			atomic.AddUint64(&p.stats.mrfStats.TotalDroppedBytes, uint64(entry.sz))
 		}
 	}
 }
@@ -3505,7 +3549,7 @@ func (p *ReplicationPool) persistToDrive(ctx context.Context, v MRFReplicateEntr
 	}
 
 	globalLocalDrivesMu.RLock()
-	localDrives := cloneDrives(globalLocalDrives)
+	localDrives := cloneDrives(globalLocalDrivesMap)
 	globalLocalDrivesMu.RUnlock()
 
 	for _, localDrive := range localDrives {
@@ -3523,7 +3567,7 @@ func (p *ReplicationPool) saveMRFEntries(ctx context.Context, entries map[string
 	if !p.initialized() {
 		return
 	}
-	atomic.StoreUint64(&globalReplicationStats.mrfStats.LastFailedCount, uint64(len(entries)))
+	atomic.StoreUint64(&p.stats.mrfStats.LastFailedCount, uint64(len(entries)))
 	if len(entries) == 0 {
 		return
 	}
@@ -3572,7 +3616,7 @@ func (p *ReplicationPool) loadMRF() (mrfRec MRFReplicateEntries, err error) {
 	}
 
 	globalLocalDrivesMu.RLock()
-	localDrives := cloneDrives(globalLocalDrives)
+	localDrives := cloneDrives(globalLocalDrivesMap)
 	globalLocalDrivesMu.RUnlock()
 
 	for _, localDrive := range localDrives {
@@ -3693,3 +3737,12 @@ func (p *ReplicationPool) getMRF(ctx context.Context, bucket string) (ch <-chan
 
 	return mrfCh, nil
 }
+
+// validateReplicationDestinationOptions is used to configure the validation of the replication destination.
+// validateReplicationDestination uses this to configure the validation.
+type validateReplicationDestinationOptions struct {
+	CheckRemoteBucket bool
+	CheckReady        bool
+
+	checkReadyErr sync.Map
+}

cmd/bucket-stats.go

@@ -20,6 +20,7 @@ package cmd
 import (
 	"fmt"
 	"math"
+	"sync/atomic"
 	"time"
 
 	"github.com/minio/madmin-go/v3"
@@ -127,8 +128,7 @@ func (l *ReplicationLastHour) getTotal() AccElem {
 
 // forwardTo time t, clearing any entries in between.
 func (l *ReplicationLastHour) forwardTo(t int64) {
-	tMin := t / 60
-	if l.LastMin >= tMin {
+	if l.LastMin >= t {
 		return
 	}
 	if t-l.LastMin >= 60 {
@@ -310,10 +310,13 @@ type ReplQNodeStats struct {
 func (r *ReplicationStats) getNodeQueueStats(bucket string) (qs ReplQNodeStats) {
 	qs.NodeName = globalLocalNodeName
 	qs.Uptime = UTCNow().Unix() - globalBootTime.Unix()
-	qs.ActiveWorkers = globalReplicationStats.ActiveWorkers()
+	qs.ActiveWorkers = globalReplicationStats.Load().ActiveWorkers()
 	qs.XferStats = make(map[RMetricName]XferStats)
 	qs.QStats = r.qCache.getBucketStats(bucket)
 	qs.TgtXferStats = make(map[string]map[RMetricName]XferStats)
+	qs.MRFStats = ReplicationMRFStats{
+		LastFailedCount: atomic.LoadUint64(&r.mrfStats.LastFailedCount),
+	}
 
 	r.RLock()
 	defer r.RUnlock()
@@ -399,10 +402,12 @@ func (r *ReplicationStats) getNodeQueueStats(bucket string) (qs ReplQNodeStats)
 func (r *ReplicationStats) getNodeQueueStatsSummary() (qs ReplQNodeStats) {
 	qs.NodeName = globalLocalNodeName
 	qs.Uptime = UTCNow().Unix() - globalBootTime.Unix()
-	qs.ActiveWorkers = globalReplicationStats.ActiveWorkers()
+	qs.ActiveWorkers = globalReplicationStats.Load().ActiveWorkers()
 	qs.XferStats = make(map[RMetricName]XferStats)
 	qs.QStats = r.qCache.getSiteStats()
-
+	qs.MRFStats = ReplicationMRFStats{
+		LastFailedCount: atomic.LoadUint64(&r.mrfStats.LastFailedCount),
+	}
 	r.RLock()
 	defer r.RUnlock()
 	tx := newXferStats()

cmd/bucket-targets.go

@@ -612,7 +612,7 @@ func (sys *BucketTargetSys) UpdateAllTargets(bucket string, tgts *madmin.BucketT
 }
 
 // create minio-go clients for buckets having remote targets
-func (sys *BucketTargetSys) set(bucket BucketInfo, meta BucketMetadata) {
+func (sys *BucketTargetSys) set(bucket string, meta BucketMetadata) {
 	cfg := meta.bucketTargetConfig
 	if cfg == nil || cfg.Empty() {
 		return
@@ -626,9 +626,9 @@ func (sys *BucketTargetSys) set(bucket BucketInfo, meta BucketMetadata) {
 			continue
 		}
 		sys.arnRemotesMap[tgt.Arn] = arnTarget{Client: tgtClient}
-		sys.updateBandwidthLimit(bucket.Name, tgt.Arn, tgt.BandwidthLimit)
+		sys.updateBandwidthLimit(bucket, tgt.Arn, tgt.BandwidthLimit)
 	}
-	sys.targetsMap[bucket.Name] = cfg.Targets
+	sys.targetsMap[bucket] = cfg.Targets
 }
 
 // Returns a minio-go Client configured to access remote host described in replication target config.

cmd/common-main.go

@@ -433,7 +433,6 @@ func buildServerCtxt(ctx *cli.Context, ctxt *serverCtxt) (err error) {
 	ctxt.RecvBufSize = ctx.Int("recv-buf-size")
 	ctxt.IdleTimeout = ctx.Duration("idle-timeout")
 	ctxt.UserTimeout = ctx.Duration("conn-user-timeout")
-	ctxt.ShutdownTimeout = ctx.Duration("shutdown-timeout")
 
 	if conf := ctx.String("config"); len(conf) > 0 {
 		err = mergeServerCtxtFromConfigFile(conf, ctxt)
@@ -686,16 +685,6 @@ func loadEnvVarsFromFiles() {
 		}
 	}
 
-	if env.IsSet(kms.EnvKMSSecretKeyFile) {
-		kmsSecret, err := readFromSecret(env.Get(kms.EnvKMSSecretKeyFile, ""))
-		if err != nil {
-			logger.Fatal(err, "Unable to read the KMS secret key inherited from secret file")
-		}
-		if kmsSecret != "" {
-			os.Setenv(kms.EnvKMSSecretKey, kmsSecret)
-		}
-	}
-
 	if env.IsSet(config.EnvConfigEnvFile) {
 		ekvs, err := minioEnvironFromFile(env.Get(config.EnvConfigEnvFile, ""))
 		if err != nil && !os.IsNotExist(err) {
@@ -769,7 +758,7 @@ func serverHandleEnvVars() {
 	if len(domains) != 0 {
 		for _, domainName := range strings.Split(domains, config.ValueSeparator) {
 			if _, ok := dns2.IsDomainName(domainName); !ok {
-				logger.Fatal(config.ErrInvalidDomainValue(nil).Msg("Unknown value `%s`", domainName),
+				logger.Fatal(config.ErrInvalidDomainValue(nil).Msgf("Unknown value `%s`", domainName),
 					"Invalid MINIO_DOMAIN value in environment variable")
 			}
 			globalDomainNames = append(globalDomainNames, domainName)
@@ -778,7 +767,7 @@ func serverHandleEnvVars() {
 		lcpSuf := lcpSuffix(globalDomainNames)
 		for _, domainName := range globalDomainNames {
 			if domainName == lcpSuf && len(globalDomainNames) > 1 {
-				logger.Fatal(config.ErrOverlappingDomainValue(nil).Msg("Overlapping domains `%s` not allowed", globalDomainNames),
+				logger.Fatal(config.ErrOverlappingDomainValue(nil).Msgf("Overlapping domains `%s` not allowed", globalDomainNames),
 					"Invalid MINIO_DOMAIN value in environment variable")
 			}
 		}
@@ -835,7 +824,7 @@ func serverHandleEnvVars() {
 		}
 	}
 
-	globalDisableFreezeOnBoot = env.Get("_MINIO_DISABLE_API_FREEZE_ON_BOOT", "") == "true" || serverDebugLog
+	globalEnableSyncBoot = env.Get("MINIO_SYNC_BOOT", config.EnableOff) == config.EnableOn
 }
 
 func loadRootCredentials() {
@@ -844,6 +833,7 @@ func loadRootCredentials() {
 	// Check both cases and authenticate them if correctly defined
 	var user, password string
 	var hasCredentials bool
+	var legacyCredentials bool
 	//nolint:gocritic
 	if env.IsSet(config.EnvRootUser) && env.IsSet(config.EnvRootPassword) {
 		user = env.Get(config.EnvRootUser, "")
@@ -852,6 +842,7 @@ func loadRootCredentials() {
 	} else if env.IsSet(config.EnvAccessKey) && env.IsSet(config.EnvSecretKey) {
 		user = env.Get(config.EnvAccessKey, "")
 		password = env.Get(config.EnvSecretKey, "")
+		legacyCredentials = true
 		hasCredentials = true
 	} else if globalServerCtxt.RootUser != "" && globalServerCtxt.RootPwd != "" {
 		user, password = globalServerCtxt.RootUser, globalServerCtxt.RootPwd
@@ -860,8 +851,13 @@ func loadRootCredentials() {
 	if hasCredentials {
 		cred, err := auth.CreateCredentials(user, password)
 		if err != nil {
-			logger.Fatal(config.ErrInvalidCredentials(err),
-				"Unable to validate credentials inherited from the shell environment")
+			if legacyCredentials {
+				logger.Fatal(config.ErrInvalidCredentials(err),
+					"Unable to validate credentials inherited from the shell environment")
+			} else {
+				logger.Fatal(config.ErrInvalidRootUserCredentials(err),
+					"Unable to validate credentials inherited from the shell environment")
+			}
 		}
 		if env.IsSet(config.EnvAccessKey) && env.IsSet(config.EnvSecretKey) {
 			msg := fmt.Sprintf("WARNING: %s and %s are deprecated.\n"+
@@ -875,6 +871,12 @@ func loadRootCredentials() {
 	} else {
 		globalActiveCred = auth.DefaultCredentials
 	}
+
+	var err error
+	globalNodeAuthToken, err = authenticateNode(globalActiveCred.AccessKey, globalActiveCred.SecretKey)
+	if err != nil {
+		logger.Fatal(err, "Unable to generate internode credentials")
+	}
 }
 
 // Initialize KMS global variable after valiadating and loading the configuration.

cmd/config-current.go

@@ -101,7 +101,7 @@ func initHelp() {
 		config.HelpKV{
 			Key:         config.SubnetSubSys,
 			Type:        "string",
-			Description: "register the cluster to MinIO SUBNET",
+			Description: "register Enterprise license for the cluster",
 			Optional:    true,
 		},
 		config.HelpKV{

cmd/data-scanner.go

@@ -31,6 +31,7 @@ import (
 	"strconv"
 	"strings"
 	"sync"
+	"sync/atomic"
 	"time"
 
 	"github.com/minio/madmin-go/v3"
@@ -226,7 +227,9 @@ func runDataScanner(ctx context.Context, objAPI ObjectLayer) {
 				binary.LittleEndian.PutUint64(tmp, cycleInfo.next)
 				tmp, _ = cycleInfo.MarshalMsg(tmp)
 				err = saveConfig(ctx, objAPI, dataUsageBloomNamePath, tmp)
-				scannerLogIf(ctx, err, dataUsageBloomNamePath)
+				if err != nil {
+					scannerLogIf(ctx, fmt.Errorf("%w, Object %s", err, dataUsageBloomNamePath))
+				}
 			}
 		}
 	}
@@ -249,7 +252,8 @@ type folderScanner struct {
 	healObjectSelect      uint32 // Do a heal check on an object once every n cycles. Must divide into healFolderInclude
 	scanMode              madmin.HealScanMode
 
-	weSleep func() bool
+	weSleep    func() bool
+	shouldHeal func() bool
 
 	disks       []StorageAPI
 	disksQuorum int
@@ -304,11 +308,12 @@ type folderScanner struct {
 // The returned cache will always be valid, but may not be updated from the existing.
 // Before each operation sleepDuration is called which can be used to temporarily halt the scanner.
 // If the supplied context is canceled the function will return at the first chance.
-func scanDataFolder(ctx context.Context, disks []StorageAPI, basePath string, cache dataUsageCache, getSize getSizeFn, scanMode madmin.HealScanMode, weSleep func() bool) (dataUsageCache, error) {
+func scanDataFolder(ctx context.Context, disks []StorageAPI, drive *xlStorage, cache dataUsageCache, getSize getSizeFn, scanMode madmin.HealScanMode, weSleep func() bool) (dataUsageCache, error) {
 	switch cache.Info.Name {
 	case "", dataUsageRoot:
 		return cache, errors.New("internal error: root scan attempted")
 	}
+	basePath := drive.drivePath
 	updatePath, closeDisk := globalScannerMetrics.currentPathUpdater(basePath, cache.Info.Name)
 	defer closeDisk()
 
@@ -328,6 +333,26 @@ func scanDataFolder(ctx context.Context, disks []StorageAPI, basePath string, ca
 		disksQuorum:           len(disks) / 2,
 	}
 
+	var skipHeal atomic.Bool
+	if globalIsErasure || cache.Info.SkipHealing {
+		skipHeal.Store(true)
+	}
+
+	// Check if we should do healing at all.
+	s.shouldHeal = func() bool {
+		if skipHeal.Load() {
+			return false
+		}
+		if s.healObjectSelect == 0 {
+			return false
+		}
+		if di, _ := drive.DiskInfo(ctx, DiskInfoOptions{}); di.Healing {
+			skipHeal.Store(true)
+			return false
+		}
+		return true
+	}
+
 	// Enable healing in XL mode.
 	if globalIsErasure && !cache.Info.SkipHealing {
 		// Do a heal check on an object once every n cycles. Must divide into healFolderInclude
@@ -482,14 +507,9 @@ func (f *folderScanner) scanFolder(ctx context.Context, folder cachedFolder, int
 				replication: replicationCfg,
 			}
 
-			item.heal.enabled = thisHash.modAlt(f.oldCache.Info.NextCycle/folder.objectHealProbDiv, f.healObjectSelect/folder.objectHealProbDiv) && globalIsErasure
+			item.heal.enabled = thisHash.modAlt(f.oldCache.Info.NextCycle/folder.objectHealProbDiv, f.healObjectSelect/folder.objectHealProbDiv) && f.shouldHeal()
 			item.heal.bitrot = f.scanMode == madmin.HealDeepScan
 
-			// if the drive belongs to an erasure set
-			// that is already being healed, skip the
-			// healing attempt on this drive.
-			item.heal.enabled = item.heal.enabled && f.healObjectSelect > 0
-
 			sz, err := f.getSize(item)
 			if err != nil && err != errIgnoreFileContrib {
 				wait() // wait to proceed to next entry.
@@ -549,7 +569,7 @@ func (f *folderScanner) scanFolder(ctx context.Context, folder cachedFolder, int
 				APIName: "Scanner",
 				Bucket:  f.root,
 				Object:  prefixName,
-				Tags: map[string]interface{}{
+				Tags: map[string]string{
 					"x-minio-prefixes-total": strconv.Itoa(totalFolders),
 				},
 			})
@@ -652,7 +672,7 @@ func (f *folderScanner) scanFolder(ctx context.Context, folder cachedFolder, int
 		}
 
 		// Scan for healing
-		if f.healObjectSelect == 0 || len(abandonedChildren) == 0 {
+		if len(abandonedChildren) == 0 || !f.shouldHeal() {
 			// If we are not heal scanning, return now.
 			break
 		}
@@ -687,6 +707,9 @@ func (f *folderScanner) scanFolder(ctx context.Context, folder cachedFolder, int
 
 		healObjectsPrefix := color.Green("healObjects:")
 		for k := range abandonedChildren {
+			if !f.shouldHeal() {
+				break
+			}
 			bucket, prefix := path2BucketObject(k)
 			stopFn := globalScannerMetrics.time(scannerMetricCheckMissing)
 			f.updateCurrentPath(k)
@@ -720,6 +743,10 @@ func (f *folderScanner) scanFolder(ctx context.Context, folder cachedFolder, int
 				},
 				// Some disks have data for this.
 				partial: func(entries metaCacheEntries, errs []error) {
+					if !f.shouldHeal() {
+						cancel()
+						return
+					}
 					entry, ok := entries.resolve(&resolver)
 					if !ok {
 						// check if we can get one entry at least
@@ -772,7 +799,9 @@ func (f *folderScanner) scanFolder(ctx context.Context, folder cachedFolder, int
 						}, madmin.HealItemObject)
 						stopFn(int(ver.Size))
 						if !isErrObjectNotFound(err) && !isErrVersionNotFound(err) {
-							scannerLogIf(ctx, err, fiv.Name)
+							if err != nil {
+								scannerLogIf(ctx, fmt.Errorf("%w, Object %s/%s/%s", err, bucket, fiv.Name, ver.VersionID))
+							}
 						}
 						if err == nil {
 							successVersions++
@@ -1109,7 +1138,7 @@ func (i *scannerItem) applyVersionActions(ctx context.Context, o ObjectLayer, fi
 			APIName: "Scanner",
 			Bucket:  i.bucket,
 			Object:  i.objectPath(),
-			Tags: map[string]interface{}{
+			Tags: map[string]string{
 				"x-minio-versions": strconv.Itoa(len(objInfos)),
 			},
 		})
@@ -1141,7 +1170,7 @@ func (i *scannerItem) applyVersionActions(ctx context.Context, o ObjectLayer, fi
 			APIName: "Scanner",
 			Bucket:  i.bucket,
 			Object:  i.objectPath(),
-			Tags: map[string]interface{}{
+			Tags: map[string]string{
 				"x-minio-versions-count": strconv.Itoa(len(objInfos)),
 				"x-minio-versions-size":  strconv.FormatInt(cumulativeSize, 10),
 			},
@@ -1246,7 +1275,7 @@ func applyExpiryOnTransitionedObject(ctx context.Context, objLayer ObjectLayer,
 		if isErrObjectNotFound(err) || isErrVersionNotFound(err) {
 			return false
 		}
-		ilmLogIf(ctx, err)
+		ilmLogIf(ctx, fmt.Errorf("expireTransitionedObject(%s, %s): %w", obj.Bucket, obj.Name, err))
 		return false
 	}
 	timeILM(1)
@@ -1299,7 +1328,7 @@ func applyExpiryOnNonTransitionedObjects(ctx context.Context, objLayer ObjectLay
 			return false
 		}
 		// Assume it is still there.
-		ilmLogOnceIf(ctx, err, "non-transition-expiry")
+		ilmLogOnceIf(ctx, fmt.Errorf("DeleteObject(%s, %s): %w", obj.Bucket, obj.Name, err), "non-transition-expiry"+obj.Name)
 		return false
 	}
 	if dobj.Name == "" {
@@ -1307,6 +1336,8 @@ func applyExpiryOnNonTransitionedObjects(ctx context.Context, objLayer ObjectLay
 	}
 
 	tags := newLifecycleAuditEvent(src, lcEvent).Tags()
+	tags["version-id"] = dobj.VersionID
+
 	// Send audit for the lifecycle delete operation
 	auditLogLifecycle(ctx, dobj, ILMExpiry, tags, traceFn)
 
@@ -1518,7 +1549,7 @@ const (
 	ILMTransition = " ilm:transition"
 )
 
-func auditLogLifecycle(ctx context.Context, oi ObjectInfo, event string, tags map[string]interface{}, traceFn func(event string)) {
+func auditLogLifecycle(ctx context.Context, oi ObjectInfo, event string, tags map[string]string, traceFn func(event string, metadata map[string]string)) {
 	var apiName string
 	switch event {
 	case ILMExpiry:
@@ -1536,5 +1567,5 @@ func auditLogLifecycle(ctx context.Context, oi ObjectInfo, event string, tags ma
 		VersionID: oi.VersionID,
 		Tags:      tags,
 	})
-	traceFn(event)
+	traceFn(event, tags)
 }

cmd/data-usage_test.go

@@ -26,6 +26,9 @@ import (
 	"path"
 	"path/filepath"
 	"testing"
+	"time"
+
+	"github.com/minio/minio/internal/cachevalue"
 )
 
 type usageTestFile struct {
@@ -61,10 +64,13 @@ func TestDataUsageUpdate(t *testing.T) {
 		}
 		return
 	}
-
+	xls := xlStorage{drivePath: base, diskInfoCache: cachevalue.New[DiskInfo]()}
+	xls.diskInfoCache.InitOnce(time.Second, cachevalue.Opts{}, func(ctx context.Context) (DiskInfo, error) {
+		return DiskInfo{Total: 1 << 40, Free: 1 << 40}, nil
+	})
 	weSleep := func() bool { return false }
 
-	got, err := scanDataFolder(context.Background(), nil, base, dataUsageCache{Info: dataUsageCacheInfo{Name: bucket}}, getSize, 0, weSleep)
+	got, err := scanDataFolder(context.Background(), nil, &xls, dataUsageCache{Info: dataUsageCacheInfo{Name: bucket}}, getSize, 0, weSleep)
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -174,7 +180,7 @@ func TestDataUsageUpdate(t *testing.T) {
 	}
 	// Changed dir must be picked up in this many cycles.
 	for i := 0; i < dataUsageUpdateDirCycles; i++ {
-		got, err = scanDataFolder(context.Background(), nil, base, got, getSize, 0, weSleep)
+		got, err = scanDataFolder(context.Background(), nil, &xls, got, getSize, 0, weSleep)
 		got.Info.NextCycle++
 		if err != nil {
 			t.Fatal(err)
@@ -283,8 +289,12 @@ func TestDataUsageUpdatePrefix(t *testing.T) {
 	}
 
 	weSleep := func() bool { return false }
+	xls := xlStorage{drivePath: base, diskInfoCache: cachevalue.New[DiskInfo]()}
+	xls.diskInfoCache.InitOnce(time.Second, cachevalue.Opts{}, func(ctx context.Context) (DiskInfo, error) {
+		return DiskInfo{Total: 1 << 40, Free: 1 << 40}, nil
+	})
 
-	got, err := scanDataFolder(context.Background(), nil, base, dataUsageCache{Info: dataUsageCacheInfo{Name: "bucket"}}, getSize, 0, weSleep)
+	got, err := scanDataFolder(context.Background(), nil, &xls, dataUsageCache{Info: dataUsageCacheInfo{Name: "bucket"}}, getSize, 0, weSleep)
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -419,7 +429,7 @@ func TestDataUsageUpdatePrefix(t *testing.T) {
 	}
 	// Changed dir must be picked up in this many cycles.
 	for i := 0; i < dataUsageUpdateDirCycles; i++ {
-		got, err = scanDataFolder(context.Background(), nil, base, got, getSize, 0, weSleep)
+		got, err = scanDataFolder(context.Background(), nil, &xls, got, getSize, 0, weSleep)
 		got.Info.NextCycle++
 		if err != nil {
 			t.Fatal(err)
@@ -567,8 +577,12 @@ func TestDataUsageCacheSerialize(t *testing.T) {
 		}
 		return
 	}
+	xls := xlStorage{drivePath: base, diskInfoCache: cachevalue.New[DiskInfo]()}
+	xls.diskInfoCache.InitOnce(time.Second, cachevalue.Opts{}, func(ctx context.Context) (DiskInfo, error) {
+		return DiskInfo{Total: 1 << 40, Free: 1 << 40}, nil
+	})
 	weSleep := func() bool { return false }
-	want, err := scanDataFolder(context.Background(), nil, base, dataUsageCache{Info: dataUsageCacheInfo{Name: bucket}}, getSize, 0, weSleep)
+	want, err := scanDataFolder(context.Background(), nil, &xls, dataUsageCache{Info: dataUsageCacheInfo{Name: bucket}}, getSize, 0, weSleep)
 	if err != nil {
 		t.Fatal(err)
 	}

cmd/dummy-data-generator_test.go

@@ -61,10 +61,9 @@ func NewDummyDataGen(totalLength, skipOffset int64) io.ReadSeeker {
 	}
 
 	skipOffset %= int64(len(alphabets))
-	as := make([]byte, 2*len(alphabets))
-	copy(as, alphabets)
-	copy(as[len(alphabets):], alphabets)
-	b := as[skipOffset : skipOffset+int64(len(alphabets))]
+	const multiply = 100
+	as := bytes.Repeat(alphabets, multiply)
+	b := as[skipOffset : skipOffset+int64(len(alphabets)*(multiply-1))]
 	return &DummyDataGen{
 		length: totalLength,
 		b:      b,

cmd/dummy-handlers.go

@@ -44,7 +44,7 @@ func (api objectAPIHandlers) GetBucketWebsiteHandler(w http.ResponseWriter, r *h
 		return
 	}
 
-	// Allow getBucketCors if policy action is set, since this is a dummy call
+	// Allow GetBucketWebsite if policy action is set, since this is a dummy call
 	// we are simply re-purposing the bucketPolicyAction.
 	if s3Error := checkRequestAuthType(ctx, r, policy.GetBucketPolicyAction, bucket, ""); s3Error != ErrNone {
 		writeErrorResponse(ctx, w, errorCodes.ToAPIErr(s3Error), r.URL)
@@ -76,7 +76,7 @@ func (api objectAPIHandlers) GetBucketAccelerateHandler(w http.ResponseWriter, r
 		return
 	}
 
-	// Allow getBucketCors if policy action is set, since this is a dummy call
+	// Allow GetBucketAccelerate if policy action is set, since this is a dummy call
 	// we are simply re-purposing the bucketPolicyAction.
 	if s3Error := checkRequestAuthType(ctx, r, policy.GetBucketPolicyAction, bucket, ""); s3Error != ErrNone {
 		writeErrorResponse(ctx, w, errorCodes.ToAPIErr(s3Error), r.URL)
@@ -109,7 +109,7 @@ func (api objectAPIHandlers) GetBucketRequestPaymentHandler(w http.ResponseWrite
 		return
 	}
 
-	// Allow getBucketCors if policy action is set, since this is a dummy call
+	// Allow GetBucketRequestPaymentHandler if policy action is set, since this is a dummy call
 	// we are simply re-purposing the bucketPolicyAction.
 	if s3Error := checkRequestAuthType(ctx, r, policy.GetBucketPolicyAction, bucket, ""); s3Error != ErrNone {
 		writeErrorResponse(ctx, w, errorCodes.ToAPIErr(s3Error), r.URL)
@@ -143,7 +143,7 @@ func (api objectAPIHandlers) GetBucketLoggingHandler(w http.ResponseWriter, r *h
 		return
 	}
 
-	// Allow getBucketCors if policy action is set, since this is a dummy call
+	// Allow GetBucketLoggingHandler if policy action is set, since this is a dummy call
 	// we are simply re-purposing the bucketPolicyAction.
 	if s3Error := checkRequestAuthType(ctx, r, policy.GetBucketPolicyAction, bucket, ""); s3Error != ErrNone {
 		writeErrorResponse(ctx, w, errorCodes.ToAPIErr(s3Error), r.URL)
@@ -181,9 +181,7 @@ func (api objectAPIHandlers) GetBucketCorsHandler(w http.ResponseWriter, r *http
 		return
 	}
 
-	// Allow getBucketCors if policy action is set, since this is a dummy call
-	// we are simply re-purposing the bucketPolicyAction.
-	if s3Error := checkRequestAuthType(ctx, r, policy.GetBucketPolicyAction, bucket, ""); s3Error != ErrNone {
+	if s3Error := checkRequestAuthType(ctx, r, policy.GetBucketCorsAction, bucket, ""); s3Error != ErrNone {
 		writeErrorResponse(ctx, w, errorCodes.ToAPIErr(s3Error), r.URL)
 		return
 	}
@@ -197,3 +195,63 @@ func (api objectAPIHandlers) GetBucketCorsHandler(w http.ResponseWriter, r *http
 
 	writeErrorResponse(ctx, w, errorCodes.ToAPIErr(ErrNoSuchCORSConfiguration), r.URL)
 }
+
+// PutBucketCorsHandler - PUT bucket cors, a dummy api
+func (api objectAPIHandlers) PutBucketCorsHandler(w http.ResponseWriter, r *http.Request) {
+	ctx := newContext(r, w, "PutBucketCors")
+
+	defer logger.AuditLog(ctx, w, r, mustGetClaimsFromToken(r))
+
+	vars := mux.Vars(r)
+	bucket := vars["bucket"]
+
+	objAPI := api.ObjectAPI()
+	if objAPI == nil {
+		writeErrorResponse(ctx, w, errorCodes.ToAPIErr(ErrServerNotInitialized), r.URL)
+		return
+	}
+
+	if s3Error := checkRequestAuthType(ctx, r, policy.PutBucketCorsAction, bucket, ""); s3Error != ErrNone {
+		writeErrorResponse(ctx, w, errorCodes.ToAPIErr(s3Error), r.URL)
+		return
+	}
+
+	// Validate if bucket exists, before proceeding further...
+	_, err := objAPI.GetBucketInfo(ctx, bucket, BucketOptions{})
+	if err != nil {
+		writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL)
+		return
+	}
+
+	writeErrorResponse(ctx, w, errorCodes.ToAPIErr(ErrNotImplemented), r.URL)
+}
+
+// DeleteBucketCorsHandler - DELETE bucket cors, a dummy api
+func (api objectAPIHandlers) DeleteBucketCorsHandler(w http.ResponseWriter, r *http.Request) {
+	ctx := newContext(r, w, "DeleteBucketCors")
+
+	defer logger.AuditLog(ctx, w, r, mustGetClaimsFromToken(r))
+
+	vars := mux.Vars(r)
+	bucket := vars["bucket"]
+
+	objAPI := api.ObjectAPI()
+	if objAPI == nil {
+		writeErrorResponse(ctx, w, errorCodes.ToAPIErr(ErrServerNotInitialized), r.URL)
+		return
+	}
+
+	if s3Error := checkRequestAuthType(ctx, r, policy.DeleteBucketCorsAction, bucket, ""); s3Error != ErrNone {
+		writeErrorResponse(ctx, w, errorCodes.ToAPIErr(s3Error), r.URL)
+		return
+	}
+
+	// Validate if bucket exists, before proceeding further...
+	_, err := objAPI.GetBucketInfo(ctx, bucket, BucketOptions{})
+	if err != nil {
+		writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL)
+		return
+	}
+
+	writeErrorResponse(ctx, w, errorCodes.ToAPIErr(ErrNotImplemented), r.URL)
+}

cmd/encryption-v1.go

@@ -134,11 +134,16 @@ func DecryptETags(ctx context.Context, k *kms.KMS, objects []ObjectInfo) error {
 		SSES3SinglePartObjects := make(map[int]bool)
 		for i, object := range batch {
 			if kind, ok := crypto.IsEncrypted(object.UserDefined); ok && kind == crypto.S3 && !crypto.IsMultiPart(object.UserDefined) {
-				SSES3SinglePartObjects[i] = true
-
-				metadata = append(metadata, object.UserDefined)
-				buckets = append(buckets, object.Bucket)
-				names = append(names, object.Name)
+				ETag, err := etag.Parse(object.ETag)
+				if err != nil {
+					continue
+				}
+				if ETag.IsEncrypted() {
+					SSES3SinglePartObjects[i] = true
+					metadata = append(metadata, object.UserDefined)
+					buckets = append(buckets, object.Bucket)
+					names = append(names, object.Name)
+				}
 			}
 		}
 
@@ -190,7 +195,7 @@ func DecryptETags(ctx context.Context, k *kms.KMS, objects []ObjectInfo) error {
 				if err != nil {
 					return err
 				}
-				if SSES3SinglePartObjects[i] && ETag.IsEncrypted() {
+				if SSES3SinglePartObjects[i] {
 					ETag, err = etag.Decrypt(keys[0][:], ETag)
 					if err != nil {
 						return err
@@ -596,28 +601,44 @@ func DecryptBlocksRequestR(inputReader io.Reader, h http.Header, seqNumber uint3
 	partEncRelOffset := int64(seqNumber) * (SSEDAREPackageBlockSize + SSEDAREPackageMetaSize)
 
 	w := &DecryptBlocksReader{
-		reader:            inputReader,
-		startSeqNum:       seqNumber,
-		partDecRelOffset:  partDecRelOffset,
-		partEncRelOffset:  partEncRelOffset,
-		parts:             oi.Parts,
-		partIndex:         partStart,
-		header:            h,
-		bucket:            bucket,
-		object:            object,
-		customerKeyHeader: h.Get(xhttp.AmzServerSideEncryptionCustomerKey),
-		copySource:        copySource,
-		metadata:          cloneMSS(oi.UserDefined),
+		reader:           inputReader,
+		startSeqNum:      seqNumber,
+		partDecRelOffset: partDecRelOffset,
+		partEncRelOffset: partEncRelOffset,
+		parts:            oi.Parts,
+		partIndex:        partStart,
 	}
 
-	if w.copySource {
-		w.customerKeyHeader = h.Get(xhttp.AmzServerSideEncryptionCopyCustomerKey)
+	// In case of SSE-C, we have to decrypt the OEK using the client-provided key.
+	// In case of a SSE-C server-side copy, the client might provide two keys,
+	// one for the source and one for the target. This reader is the source.
+	var ssecClientKey []byte
+	if crypto.SSEC.IsEncrypted(oi.UserDefined) {
+		if copySource && crypto.SSECopy.IsRequested(h) {
+			key, err := crypto.SSECopy.ParseHTTP(h)
+			if err != nil {
+				return nil, err
+			}
+			ssecClientKey = key[:]
+		} else {
+			key, err := crypto.SSEC.ParseHTTP(h)
+			if err != nil {
+				return nil, err
+			}
+			ssecClientKey = key[:]
+		}
 	}
 
+	// Decrypt the OEK once and reuse it for all subsequent parts.
+	objectEncryptionKey, err := decryptObjectMeta(ssecClientKey, bucket, object, oi.UserDefined)
+	if err != nil {
+		return nil, err
+	}
+	w.objectEncryptionKey = objectEncryptionKey
+
 	if err := w.buildDecrypter(w.parts[w.partIndex].Number); err != nil {
 		return nil, err
 	}
-
 	return w, nil
 }
 
@@ -633,48 +654,17 @@ type DecryptBlocksReader struct {
 	// Current part index
 	partIndex int
 	// Parts information
-	parts          []ObjectPartInfo
-	header         http.Header
-	bucket, object string
-	metadata       map[string]string
+	parts []ObjectPartInfo
 
+	objectEncryptionKey                []byte
 	partDecRelOffset, partEncRelOffset int64
-
-	copySource bool
-	// Customer Key
-	customerKeyHeader string
 }
 
 func (d *DecryptBlocksReader) buildDecrypter(partID int) error {
-	m := cloneMSS(d.metadata)
-	// Initialize the first decrypter; new decrypters will be
-	// initialized in Read() operation as needed.
-	var key []byte
-	var err error
-	if d.copySource {
-		if crypto.SSEC.IsEncrypted(d.metadata) {
-			d.header.Set(xhttp.AmzServerSideEncryptionCopyCustomerKey, d.customerKeyHeader)
-			key, err = ParseSSECopyCustomerRequest(d.header, d.metadata)
-		}
-	} else {
-		if crypto.SSEC.IsEncrypted(d.metadata) {
-			d.header.Set(xhttp.AmzServerSideEncryptionCustomerKey, d.customerKeyHeader)
-			key, err = ParseSSECustomerHeader(d.header)
-		}
-	}
-	if err != nil {
-		return err
-	}
-
-	objectEncryptionKey, err := decryptObjectMeta(key, d.bucket, d.object, m)
-	if err != nil {
-		return err
-	}
-
 	var partIDbin [4]byte
 	binary.LittleEndian.PutUint32(partIDbin[:], uint32(partID)) // marshal part ID
 
-	mac := hmac.New(sha256.New, objectEncryptionKey) // derive part encryption key from part ID and object key
+	mac := hmac.New(sha256.New, d.objectEncryptionKey) // derive part encryption key from part ID and object key
 	mac.Write(partIDbin[:])
 	partEncryptionKey := mac.Sum(nil)
 
@@ -738,8 +728,9 @@ func (d *DecryptBlocksReader) Read(p []byte) (int, error) {
 // but has an invalid size.
 func (o ObjectInfo) DecryptedSize() (int64, error) {
 	if _, ok := crypto.IsEncrypted(o.UserDefined); !ok {
-		return 0, errors.New("Cannot compute decrypted size of an unencrypted object")
+		return -1, errors.New("Cannot compute decrypted size of an unencrypted object")
 	}
+
 	if !o.isMultipart() {
 		size, err := sio.DecryptedSize(uint64(o.Size))
 		if err != nil {
@@ -752,7 +743,7 @@ func (o ObjectInfo) DecryptedSize() (int64, error) {
 	for _, part := range o.Parts {
 		partSize, err := sio.DecryptedSize(uint64(part.Size))
 		if err != nil {
-			return 0, errObjectTampered
+			return -1, errObjectTampered
 		}
 		size += int64(partSize)
 	}
@@ -1098,7 +1089,7 @@ func (o *ObjectInfo) metadataDecrypter(h http.Header) objectMetaDecryptFn {
 	}
 }
 
-// decryptChecksums will attempt to decode checksums and return it/them if set.
+// decryptPartsChecksums will attempt to decode checksums and return it/them if set.
 // if part > 0, and we have the checksum for the part that will be returned.
 func (o *ObjectInfo) decryptPartsChecksums(h http.Header) {
 	data := o.Checksum
@@ -1167,6 +1158,14 @@ func (o *ObjectInfo) decryptChecksums(part int, h http.Header) map[string]string
 	if len(data) == 0 {
 		return nil
 	}
+	if part > 0 && !crypto.SSEC.IsEncrypted(o.UserDefined) {
+		// already decrypted in ToObjectInfo for multipart objects
+		for _, pi := range o.Parts {
+			if pi.Number == part {
+				return pi.Checksums
+			}
+		}
+	}
 	if _, encrypted := crypto.IsEncrypted(o.UserDefined); encrypted {
 		decrypted, err := o.metadataDecrypter(h)("object-checksum", data)
 		if err != nil {

cmd/endpoint-ellipses.go

@@ -311,7 +311,7 @@ func GetAllSets(setDriveCount uint64, args ...string) ([][]string, error) {
 	for _, sargs := range setArgs {
 		for _, arg := range sargs {
 			if uniqueArgs.Contains(arg) {
-				return nil, config.ErrInvalidErasureEndpoints(nil).Msg(fmt.Sprintf("Input args (%s) has duplicate ellipses", args))
+				return nil, config.ErrInvalidErasureEndpoints(nil).Msgf("Input args (%s) has duplicate ellipses", args)
 			}
 			uniqueArgs.Add(arg)
 		}

cmd/endpoint.go

@@ -253,6 +253,14 @@ type PoolEndpoints struct {
 // EndpointServerPools - list of list of endpoints
 type EndpointServerPools []PoolEndpoints
 
+// ESCount returns the total number of erasure sets in this cluster
+func (l EndpointServerPools) ESCount() (count int) {
+	for _, p := range l {
+		count += p.SetCount
+	}
+	return
+}
+
 // GetNodes returns a sorted list of nodes in this cluster
 func (l EndpointServerPools) GetNodes() (nodes []Node) {
 	nodesMap := make(map[string]Node)

cmd/erasure-common.go

@@ -19,13 +19,9 @@ package cmd
 
 import (
 	"context"
-	"fmt"
-	"io"
 	"math/rand"
 	"sync"
 	"time"
-
-	"github.com/minio/pkg/v3/sync/errgroup"
 )
 
 func (er erasureObjects) getOnlineDisks() (newDisks []StorageAPI) {
@@ -87,89 +83,3 @@ func (er erasureObjects) getLocalDisks() (newDisks []StorageAPI) {
 	}
 	return newDisks
 }
-
-// readMultipleFiles Reads raw data from all specified files from all disks.
-func readMultipleFiles(ctx context.Context, disks []StorageAPI, req ReadMultipleReq, readQuorum int) ([]ReadMultipleResp, error) {
-	resps := make([]chan ReadMultipleResp, len(disks))
-	for i := range resps {
-		resps[i] = make(chan ReadMultipleResp, len(req.Files))
-	}
-	g := errgroup.WithNErrs(len(disks))
-	// Read files in parallel across disks.
-	for index := range disks {
-		index := index
-		g.Go(func() (err error) {
-			if disks[index] == nil {
-				return errDiskNotFound
-			}
-			return disks[index].ReadMultiple(ctx, req, resps[index])
-		}, index)
-	}
-
-	dataArray := make([]ReadMultipleResp, 0, len(req.Files))
-	// Merge results. They should come in order from each.
-	for _, wantFile := range req.Files {
-		quorum := 0
-		toAdd := ReadMultipleResp{
-			Bucket: req.Bucket,
-			Prefix: req.Prefix,
-			File:   wantFile,
-		}
-		for i := range resps {
-			if disks[i] == nil {
-				continue
-			}
-			select {
-			case <-ctx.Done():
-			case gotFile, ok := <-resps[i]:
-				if !ok {
-					continue
-				}
-				if gotFile.Error != "" || !gotFile.Exists {
-					continue
-				}
-				if gotFile.File != wantFile || gotFile.Bucket != req.Bucket || gotFile.Prefix != req.Prefix {
-					continue
-				}
-				quorum++
-				if toAdd.Modtime.After(gotFile.Modtime) || len(gotFile.Data) < len(toAdd.Data) {
-					// Pick latest, or largest to avoid possible truncated entries.
-					continue
-				}
-				toAdd = gotFile
-			}
-		}
-		if quorum < readQuorum {
-			toAdd.Exists = false
-			toAdd.Error = errErasureReadQuorum.Error()
-			toAdd.Data = nil
-		}
-		dataArray = append(dataArray, toAdd)
-	}
-
-	ignoredErrs := []error{
-		errFileNotFound,
-		errVolumeNotFound,
-		errFileVersionNotFound,
-		io.ErrUnexpectedEOF, // some times we would read without locks, ignore these errors
-		io.EOF,              // some times we would read without locks, ignore these errors
-		context.DeadlineExceeded,
-		context.Canceled,
-	}
-	ignoredErrs = append(ignoredErrs, objectOpIgnoredErrs...)
-
-	errs := g.Wait()
-	for index, err := range errs {
-		if err == nil {
-			continue
-		}
-		if !IsErr(err, ignoredErrs...) {
-			storageLogOnceIf(ctx, fmt.Errorf("Drive %s, path (%s/%s) returned an error (%w)",
-				disks[index], req.Bucket, req.Prefix, err),
-				disks[index].String())
-		}
-	}
-
-	// Return all the metadata.
-	return dataArray, nil
-}

cmd/erasure-decode.go

@@ -48,14 +48,14 @@ func newParallelReader(readers []io.ReaderAt, e Erasure, offset, totalLength int
 		r2b[i] = i
 	}
 	bufs := make([][]byte, len(readers))
-	// Fill buffers
-	b := globalBytePoolCap.Load().Get()
 	shardSize := int(e.ShardSize())
-	if cap(b) < len(readers)*shardSize {
-		// We should always have enough capacity, but older objects may be bigger.
-		globalBytePoolCap.Load().Put(b)
-		b = nil
-	} else {
+	var b []byte
+
+	// We should always have enough capacity, but older objects may be bigger
+	// we do not need stashbuffer for them.
+	if globalBytePoolCap.Load().WidthCap() >= len(readers)*shardSize {
+		// Fill buffers
+		b = globalBytePoolCap.Load().Get()
 		// Seed the buffers.
 		for i := range bufs {
 			bufs[i] = b[i*shardSize : (i+1)*shardSize]

cmd/erasure-healing-common_test.go

@@ -23,6 +23,7 @@ import (
 	"fmt"
 	"os"
 	"path/filepath"
+	"runtime"
 	"testing"
 	"time"
 
@@ -147,6 +148,10 @@ func TestCommonTime(t *testing.T) {
 // TestListOnlineDisks - checks if listOnlineDisks and outDatedDisks
 // are consistent with each other.
 func TestListOnlineDisks(t *testing.T) {
+	if runtime.GOOS == globalWindowsOSName {
+		t.Skip()
+	}
+
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
 
@@ -230,7 +235,7 @@ func TestListOnlineDisks(t *testing.T) {
 	}
 
 	object := "object"
-	data := bytes.Repeat([]byte("a"), smallFileThreshold*16)
+	data := bytes.Repeat([]byte("a"), smallFileThreshold*32)
 	z := obj.(*erasureServerPools)
 
 	erasureDisks, err := z.GetDisks(0, 0)

cmd/erasure-healing.go

@@ -226,21 +226,28 @@ func (er *erasureObjects) auditHealObject(ctx context.Context, bucket, object, v
 	opts := AuditLogOptions{
 		Event:     "HealObject",
 		Bucket:    bucket,
-		Object:    object,
+		Object:    decodeDirObject(object),
 		VersionID: versionID,
 	}
 	if err != nil {
 		opts.Error = err.Error()
 	}
 
-	opts.Tags = map[string]interface{}{
-		"healResult": result,
-		"objectLocation": auditObjectOp{
-			Name:   decodeDirObject(object),
-			Pool:   er.poolIndex + 1,
-			Set:    er.setIndex + 1,
-			Drives: er.getEndpointStrings(),
-		},
+	b, a := result.GetCorruptedCounts()
+	if b == a {
+		opts.Error = fmt.Sprintf("unable to heal %d corrupted blocks on drives", b)
+	}
+	b, a = result.GetMissingCounts()
+	if b == a {
+		opts.Error = fmt.Sprintf("unable to heal %d missing blocks on drives", b)
+	}
+
+	opts.Tags = map[string]string{
+		"healObject": auditObjectOp{
+			Name: opts.Object,
+			Pool: er.poolIndex + 1,
+			Set:  er.setIndex + 1,
+		}.String(),
 	}
 
 	auditLogInternal(ctx, opts)
@@ -629,7 +636,7 @@ func (er *erasureObjects) healObject(ctx context.Context, bucket string, object
 		}
 
 		for i, v := range result.Before.Drives {
-			if v.Endpoint == disk.String() {
+			if v.Endpoint == disk.Endpoint().String() {
 				result.After.Drives[i].State = madmin.DriveStateOk
 			}
 		}

cmd/erasure-metadata.go

@@ -26,6 +26,7 @@ import (
 	"time"
 
 	"github.com/minio/minio/internal/amztime"
+	"github.com/minio/minio/internal/bucket/lifecycle"
 	"github.com/minio/minio/internal/bucket/replication"
 	"github.com/minio/minio/internal/crypto"
 	"github.com/minio/minio/internal/hash/sha256"
@@ -173,6 +174,7 @@ func (fi FileInfo) ToObjectInfo(bucket, object string, versioned bool) ObjectInf
 		}
 	}
 	objInfo.Checksum = fi.Checksum
+	objInfo.decryptPartsChecksums(nil)
 	objInfo.Inlined = fi.InlineData()
 	// Success.
 	return objInfo
@@ -287,13 +289,14 @@ func findFileInfoInQuorum(ctx context.Context, metaArr []FileInfo, modTime time.
 		mtimeValid := meta.ModTime.Equal(modTime)
 		if mtimeValid || etagOnly {
 			fmt.Fprintf(h, "%v", meta.XLV1)
-			if !etagOnly {
-				// Verify dataDir is same only when mtime is valid and etag is not considered.
-				fmt.Fprintf(h, "%v", meta.GetDataDir())
-			}
 			for _, part := range meta.Parts {
 				fmt.Fprintf(h, "part.%d", part.Number)
+				fmt.Fprintf(h, "part.%d", part.Size)
 			}
+			// Previously we checked if we had quorum on DataDir value.
+			// We have removed this check to allow reading objects with different DataDir
+			// values in a few drives (due to a rebalance-stop race bug)
+			// provided their their etags or ModTimes match.
 
 			if !meta.Deleted && meta.Size != 0 {
 				fmt.Fprintf(h, "%v+%v", meta.Erasure.DataBlocks, meta.Erasure.ParityBlocks)
@@ -388,8 +391,7 @@ func pickValidFileInfo(ctx context.Context, metaArr []FileInfo, modTime time.Tim
 	return findFileInfoInQuorum(ctx, metaArr, modTime, etag, quorum)
 }
 
-// writeUniqueFileInfo - writes unique `xl.meta` content for each disk concurrently.
-func writeUniqueFileInfo(ctx context.Context, disks []StorageAPI, origbucket, bucket, prefix string, files []FileInfo, quorum int) ([]StorageAPI, error) {
+func writeAllMetadataWithRevert(ctx context.Context, disks []StorageAPI, origbucket, bucket, prefix string, files []FileInfo, quorum int, revert bool) ([]StorageAPI, error) {
 	g := errgroup.WithNErrs(len(disks))
 
 	// Start writing `xl.meta` to all disks in parallel.
@@ -413,9 +415,37 @@ func writeUniqueFileInfo(ctx context.Context, disks []StorageAPI, origbucket, bu
 	mErrs := g.Wait()
 
 	err := reduceWriteQuorumErrs(ctx, mErrs, objectOpIgnoredErrs, quorum)
+	if err != nil && revert {
+		ng := errgroup.WithNErrs(len(disks))
+		for index := range disks {
+			if mErrs[index] != nil {
+				continue
+			}
+			index := index
+			ng.Go(func() error {
+				if disks[index] == nil {
+					return errDiskNotFound
+				}
+				return disks[index].Delete(ctx, bucket, pathJoin(prefix, xlStorageFormatFile), DeleteOptions{
+					Recursive: true,
+				})
+			}, index)
+		}
+		ng.Wait()
+	}
+
 	return evalDisks(disks, mErrs), err
 }
 
+func writeAllMetadata(ctx context.Context, disks []StorageAPI, origbucket, bucket, prefix string, files []FileInfo, quorum int) ([]StorageAPI, error) {
+	return writeAllMetadataWithRevert(ctx, disks, origbucket, bucket, prefix, files, quorum, true)
+}
+
+// writeUniqueFileInfo - writes unique `xl.meta` content for each disk concurrently.
+func writeUniqueFileInfo(ctx context.Context, disks []StorageAPI, origbucket, bucket, prefix string, files []FileInfo, quorum int) ([]StorageAPI, error) {
+	return writeAllMetadataWithRevert(ctx, disks, origbucket, bucket, prefix, files, quorum, false)
+}
+
 func commonParity(parities []int, defaultParityCount int) int {
 	N := len(parities)
 
@@ -456,6 +486,7 @@ func commonParity(parities []int, defaultParityCount int) int {
 }
 
 func listObjectParities(partsMetadata []FileInfo, errs []error) (parities []int) {
+	totalShards := len(partsMetadata)
 	parities = make([]int, len(partsMetadata))
 	for index, metadata := range partsMetadata {
 		if errs[index] != nil {
@@ -466,9 +497,13 @@ func listObjectParities(partsMetadata []FileInfo, errs []error) (parities []int)
 			parities[index] = -1
 			continue
 		}
+		//nolint:gocritic
 		// Delete marker or zero byte objects take highest parity.
 		if metadata.Deleted || metadata.Size == 0 {
-			parities[index] = len(partsMetadata) / 2
+			parities[index] = totalShards / 2
+		} else if metadata.TransitionStatus == lifecycle.TransitionComplete {
+			// For tiered objects, read quorum is N/2+1 to ensure simple majority on xl.meta. It is not equal to EcM because the data integrity is entrusted with the warm tier.
+			parities[index] = totalShards - (totalShards/2 + 1)
 		} else {
 			parities[index] = metadata.Erasure.ParityBlocks
 		}

cmd/erasure-metadata_test.go

@@ -19,6 +19,8 @@ package cmd
 
 import (
 	"context"
+	"fmt"
+	"slices"
 	"strconv"
 	"testing"
 	"time"
@@ -359,3 +361,124 @@ func TestSkipTierFreeVersion(t *testing.T) {
 		t.Fatal("Expected SkipTierFreeVersion to be set on FileInfo but wasn't")
 	}
 }
+
+func TestListObjectParities(t *testing.T) {
+	mkMetaArr := func(N, parity, agree int) []FileInfo {
+		fi := newFileInfo("obj-1", N-parity, parity)
+		fi.TransitionTier = "WARM-TIER"
+		fi.TransitionedObjName = mustGetUUID()
+		fi.TransitionStatus = "complete"
+		fi.Size = 1 << 20
+
+		metaArr := make([]FileInfo, N)
+		for i := range N {
+			fi.Erasure.Index = i + 1
+			metaArr[i] = fi
+			if i < agree {
+				continue
+			}
+			metaArr[i].TransitionTier, metaArr[i].TransitionedObjName = "", ""
+			metaArr[i].TransitionStatus = ""
+		}
+		return metaArr
+	}
+	mkParities := func(N, agreedParity, disagreedParity, agree int) []int {
+		ps := make([]int, N)
+		for i := range N {
+			if i < agree {
+				ps[i] = agreedParity
+				continue
+			}
+			ps[i] = disagreedParity // disagree
+		}
+		return ps
+	}
+
+	mkTest := func(N, parity, agree int) (res struct {
+		metaArr  []FileInfo
+		errs     []error
+		parities []int
+		parity   int
+	},
+	) {
+		res.metaArr = mkMetaArr(N, parity, agree)
+		res.parities = mkParities(N, N-(N/2+1), parity, agree)
+		res.errs = make([]error, N)
+		if agree >= N/2+1 { // simple majority consensus
+			res.parity = N - (N/2 + 1)
+		} else {
+			res.parity = -1
+		}
+		return res
+	}
+
+	nonTieredTest := func(N, parity, agree int) (res struct {
+		metaArr  []FileInfo
+		errs     []error
+		parities []int
+		parity   int
+	},
+	) {
+		fi := newFileInfo("obj-1", N-parity, parity)
+		fi.Size = 1 << 20
+		metaArr := make([]FileInfo, N)
+		parities := make([]int, N)
+		for i := range N {
+			fi.Erasure.Index = i + 1
+			metaArr[i] = fi
+			parities[i] = parity
+			if i < agree {
+				continue
+			}
+			metaArr[i].Erasure.Index = 0 // creates invalid fi on remaining drives
+			parities[i] = -1             // invalid fi are assigned parity -1
+		}
+		res.metaArr = metaArr
+		res.parities = parities
+		res.errs = make([]error, N)
+		if agree >= N-parity {
+			res.parity = parity
+		} else {
+			res.parity = -1
+		}
+
+		return res
+	}
+	tests := []struct {
+		metaArr  []FileInfo
+		errs     []error
+		parities []int
+		parity   int
+	}{
+		// More than simple majority consensus
+		mkTest(15, 3, 11),
+		// No simple majority consensus
+		mkTest(15, 3, 7),
+		// Exact simple majority consensus
+		mkTest(15, 3, 8),
+		// More than simple majority consensus
+		mkTest(16, 4, 11),
+		// No simple majority consensus
+		mkTest(16, 4, 8),
+		// Exact simple majority consensus
+		mkTest(16, 4, 9),
+		// non-tiered object require read quorum of EcM
+		nonTieredTest(15, 3, 12),
+		// non-tiered object with fewer than EcM in consensus
+		nonTieredTest(15, 3, 11),
+		// non-tiered object require read quorum of EcM
+		nonTieredTest(16, 4, 12),
+		// non-tiered object with fewer than EcM in consensus
+		nonTieredTest(16, 4, 11),
+	}
+	for i, test := range tests {
+		t.Run(fmt.Sprintf("Test %d", i+1), func(t *testing.T) {
+			if got := listObjectParities(test.metaArr, test.errs); !slices.Equal(got, test.parities) {
+				t.Fatalf("Expected parities %v but got %v", test.parities, got)
+			}
+			if got := commonParity(test.parities, len(test.metaArr)/2); got != test.parity {
+				t.Fatalf("Expected common parity %v but got %v", test.parity, got)
+			}
+		})
+	}
+}

cmd/erasure-multipart.go

@@ -80,6 +80,14 @@ func (er erasureObjects) checkUploadIDExists(ctx context.Context, bucket, object
 		return fi, nil, err
 	}
 
+	if readQuorum < 0 {
+		return fi, nil, errErasureReadQuorum
+	}
+
+	if writeQuorum < 0 {
+		return fi, nil, errErasureWriteQuorum
+	}
+
 	quorum := readQuorum
 	if write {
 		quorum = writeQuorum
@@ -88,14 +96,13 @@ func (er erasureObjects) checkUploadIDExists(ctx context.Context, bucket, object
 	// List all online disks.
 	_, modTime, etag := listOnlineDisks(storageDisks, partsMetadata, errs, quorum)
 
-	var reducedErr error
 	if write {
-		reducedErr = reduceWriteQuorumErrs(ctx, errs, objectOpIgnoredErrs, writeQuorum)
+		err = reduceWriteQuorumErrs(ctx, errs, objectOpIgnoredErrs, writeQuorum)
 	} else {
-		reducedErr = reduceReadQuorumErrs(ctx, errs, objectOpIgnoredErrs, readQuorum)
+		err = reduceReadQuorumErrs(ctx, errs, objectOpIgnoredErrs, readQuorum)
 	}
-	if reducedErr != nil {
-		return fi, nil, reducedErr
+	if err != nil {
+		return fi, nil, err
 	}
 
 	// Pick one from the first valid metadata.
@@ -103,10 +110,9 @@ func (er erasureObjects) checkUploadIDExists(ctx context.Context, bucket, object
 	return fi, partsMetadata, err
 }
 
-// Removes part.meta given by partName belonging to a multipart upload from minioMetaBucket
-func (er erasureObjects) removePartMeta(bucket, object, uploadID, dataDir string, partNumber int) {
-	uploadIDPath := er.getUploadIDDir(bucket, object, uploadID)
-	curpartPath := pathJoin(uploadIDPath, dataDir, fmt.Sprintf("part.%d", partNumber))
+// cleanupMultipartPath removes all extraneous files and parts from the multipart folder, this is used per CompleteMultipart.
+// do not use this function outside of completeMultipartUpload()
+func (er erasureObjects) cleanupMultipartPath(ctx context.Context, paths ...string) {
 	storageDisks := er.getDisks()
 
 	g := errgroup.WithNErrs(len(storageDisks))
@@ -116,42 +122,7 @@ func (er erasureObjects) removePartMeta(bucket, object, uploadID, dataDir string
 		}
 		index := index
 		g.Go(func() error {
-			_ = storageDisks[index].Delete(context.TODO(), minioMetaMultipartBucket, curpartPath+".meta", DeleteOptions{
-				Recursive: false,
-				Immediate: false,
-			})
-
-			return nil
-		}, index)
-	}
-	g.Wait()
-}
-
-// Removes part given by partName belonging to a multipart upload from minioMetaBucket
-func (er erasureObjects) removeObjectPart(bucket, object, uploadID, dataDir string, partNumber int) {
-	uploadIDPath := er.getUploadIDDir(bucket, object, uploadID)
-	curpartPath := pathJoin(uploadIDPath, dataDir, fmt.Sprintf("part.%d", partNumber))
-	storageDisks := er.getDisks()
-
-	g := errgroup.WithNErrs(len(storageDisks))
-	for index, disk := range storageDisks {
-		if disk == nil {
-			continue
-		}
-		index := index
-		g.Go(func() error {
-			// Ignoring failure to remove parts that weren't present in CompleteMultipartUpload
-			// requests. xl.meta is the authoritative source of truth on which parts constitute
-			// the object. The presence of parts that don't belong in the object doesn't affect correctness.
-			_ = storageDisks[index].Delete(context.TODO(), minioMetaMultipartBucket, curpartPath, DeleteOptions{
-				Recursive: false,
-				Immediate: false,
-			})
-			_ = storageDisks[index].Delete(context.TODO(), minioMetaMultipartBucket, curpartPath+".meta", DeleteOptions{
-				Recursive: false,
-				Immediate: false,
-			})
-
+			_ = storageDisks[index].DeleteBulk(ctx, minioMetaMultipartBucket, paths...)
 			return nil
 		}, index)
 	}
@@ -159,7 +130,7 @@ func (er erasureObjects) removeObjectPart(bucket, object, uploadID, dataDir stri
 }
 
 // Clean-up the old multipart uploads. Should be run in a Go routine.
-func (er erasureObjects) cleanupStaleUploads(ctx context.Context, expiry time.Duration) {
+func (er erasureObjects) cleanupStaleUploads(ctx context.Context) {
 	// run multiple cleanup's local to this server.
 	var wg sync.WaitGroup
 	for _, disk := range er.getLocalDisks() {
@@ -167,7 +138,7 @@ func (er erasureObjects) cleanupStaleUploads(ctx context.Context, expiry time.Du
 			wg.Add(1)
 			go func(disk StorageAPI) {
 				defer wg.Done()
-				er.cleanupStaleUploadsOnDisk(ctx, disk, expiry)
+				er.cleanupStaleUploadsOnDisk(ctx, disk)
 			}(disk)
 		}
 	}
@@ -193,7 +164,7 @@ func (er erasureObjects) deleteAll(ctx context.Context, bucket, prefix string) {
 }
 
 // Remove the old multipart uploads on the given disk.
-func (er erasureObjects) cleanupStaleUploadsOnDisk(ctx context.Context, disk StorageAPI, expiry time.Duration) {
+func (er erasureObjects) cleanupStaleUploadsOnDisk(ctx context.Context, disk StorageAPI) {
 	drivePath := disk.Endpoint().Path
 
 	readDirFn(pathJoin(drivePath, minioMetaMultipartBucket), func(shaDir string, typ os.FileMode) error {
@@ -219,7 +190,7 @@ func (er erasureObjects) cleanupStaleUploadsOnDisk(ctx context.Context, disk Sto
 				modTime = fi.ModTime
 				wait()
 			}
-			if time.Since(modTime) < expiry {
+			if time.Since(modTime) < globalAPIConfig.getStaleUploadsExpiry() {
 				return nil
 			}
 			w := xioutil.NewDeadlineWorker(globalDriveConfig.GetMaxTimeout())
@@ -238,7 +209,7 @@ func (er erasureObjects) cleanupStaleUploadsOnDisk(ctx context.Context, disk Sto
 			return nil
 		}
 		// Modtime is returned in the Created field. See (*xlStorage).StatVol
-		if time.Since(vi.Created) < expiry {
+		if time.Since(vi.Created) < globalAPIConfig.getStaleUploadsExpiry() {
 			return nil
 		}
 		w := xioutil.NewDeadlineWorker(globalDriveConfig.GetMaxTimeout())
@@ -267,7 +238,7 @@ func (er erasureObjects) cleanupStaleUploadsOnDisk(ctx context.Context, disk Sto
 		w := xioutil.NewDeadlineWorker(globalDriveConfig.GetMaxTimeout())
 		return w.Run(func() error {
 			wait := deleteMultipartCleanupSleeper.Timer(ctx)
-			if time.Since(vi.Created) > expiry {
+			if time.Since(vi.Created) > globalAPIConfig.getStaleUploadsExpiry() {
 				pathUUID := mustGetUUID()
 				targetPath := pathJoin(drivePath, minioMetaTmpDeletedBucket, pathUUID)
 
@@ -287,7 +258,7 @@ func (er erasureObjects) cleanupStaleUploadsOnDisk(ctx context.Context, disk Sto
 // towards simplification of multipart APIs.
 // The resulting ListMultipartsInfo structure is unmarshalled directly as XML.
 func (er erasureObjects) ListMultipartUploads(ctx context.Context, bucket, object, keyMarker, uploadIDMarker, delimiter string, maxUploads int) (result ListMultipartsInfo, err error) {
-	auditObjectErasureSet(ctx, object, &er)
+	auditObjectErasureSet(ctx, "ListMultipartUploads", object, &er)
 
 	result.MaxUploads = maxUploads
 	result.KeyMarker = keyMarker
@@ -526,9 +497,10 @@ func (er erasureObjects) newMultipartUpload(ctx context.Context, bucket string,
 	uploadIDPath := er.getUploadIDDir(bucket, object, uploadUUID)
 
 	// Write updated `xl.meta` to all disks.
-	if _, err := writeUniqueFileInfo(ctx, onlineDisks, bucket, minioMetaMultipartBucket, uploadIDPath, partsMetadata, writeQuorum); err != nil {
+	if _, err := writeAllMetadata(ctx, onlineDisks, bucket, minioMetaMultipartBucket, uploadIDPath, partsMetadata, writeQuorum); err != nil {
 		return nil, toObjectErr(err, bucket, object)
 	}
+
 	return &NewMultipartUploadResult{
 		UploadID:     uploadID,
 		ChecksumAlgo: userDefined[hash.MinIOMultipartChecksum],
@@ -542,14 +514,14 @@ func (er erasureObjects) newMultipartUpload(ctx context.Context, bucket string,
 // Implements S3 compatible initiate multipart API.
 func (er erasureObjects) NewMultipartUpload(ctx context.Context, bucket, object string, opts ObjectOptions) (*NewMultipartUploadResult, error) {
 	if !opts.NoAuditLog {
-		auditObjectErasureSet(ctx, object, &er)
+		auditObjectErasureSet(ctx, "NewMultipartUpload", object, &er)
 	}
 
 	return er.newMultipartUpload(ctx, bucket, object, opts)
 }
 
 // renamePart - renames multipart part to its relevant location under uploadID.
-func renamePart(ctx context.Context, disks []StorageAPI, srcBucket, srcEntry, dstBucket, dstEntry string, writeQuorum int) ([]StorageAPI, error) {
+func (er erasureObjects) renamePart(ctx context.Context, disks []StorageAPI, srcBucket, srcEntry, dstBucket, dstEntry string, optsMeta []byte, writeQuorum int) ([]StorageAPI, error) {
 	g := errgroup.WithNErrs(len(disks))
 
 	// Rename file on all underlying storage disks.
@@ -559,60 +531,25 @@ func renamePart(ctx context.Context, disks []StorageAPI, srcBucket, srcEntry, ds
 			if disks[index] == nil {
 				return errDiskNotFound
 			}
-			return disks[index].RenameFile(ctx, srcBucket, srcEntry, dstBucket, dstEntry)
+			return disks[index].RenamePart(ctx, srcBucket, srcEntry, dstBucket, dstEntry, optsMeta)
 		}, index)
 	}
 
 	// Wait for all renames to finish.
 	errs := g.Wait()
 
-	// Do not need to undo partial successful operation since those will be cleaned up
-	// in 24hrs via multipart cleaner, never rename() back to `.minio.sys/tmp` as there
-	// is no way to clean them.
-
-	// We can safely allow RenameFile errors up to len(er.getDisks()) - writeQuorum
-	// otherwise return failure. Cleanup successful renames.
-	return evalDisks(disks, errs), reduceWriteQuorumErrs(ctx, errs, objectOpIgnoredErrs, writeQuorum)
-}
-
-// writeAllDisks - writes 'b' to all provided disks.
-// If write cannot reach quorum, the files will be deleted from all disks.
-func writeAllDisks(ctx context.Context, disks []StorageAPI, dstBucket, dstEntry string, b []byte, writeQuorum int) ([]StorageAPI, error) {
-	g := errgroup.WithNErrs(len(disks))
-
-	// Write file to all underlying storage disks.
-	for index := range disks {
-		index := index
-		g.Go(func() error {
-			if disks[index] == nil {
-				return errDiskNotFound
-			}
-			return disks[index].WriteAll(ctx, dstBucket, dstEntry, b)
-		}, index)
+	paths := []string{
+		dstEntry,
+		dstEntry + ".meta",
 	}
 
-	// Wait for all renames to finish.
-	errs := g.Wait()
-
-	// We can safely allow RenameFile errors up to len(er.getDisks()) - writeQuorum
-	// otherwise return failure. Cleanup successful renames.
 	err := reduceWriteQuorumErrs(ctx, errs, objectOpIgnoredErrs, writeQuorum)
-	if errors.Is(err, errErasureWriteQuorum) {
-		// Remove all written
-		g := errgroup.WithNErrs(len(disks))
-		for index := range disks {
-			if disks[index] == nil || errs[index] != nil {
-				continue
-			}
-			index := index
-			g.Go(func() error {
-				return disks[index].Delete(ctx, dstBucket, dstEntry, DeleteOptions{Immediate: true})
-			}, index)
-		}
-		// Ignore these errors.
-		g.WaitErr()
+	if err != nil {
+		er.cleanupMultipartPath(ctx, paths...)
 	}
 
+	// We can safely allow RenameFile errors up to len(er.getDisks()) - writeQuorum
+	// otherwise return failure. Cleanup successful renames.
 	return evalDisks(disks, errs), err
 }
 
@@ -623,7 +560,7 @@ func writeAllDisks(ctx context.Context, disks []StorageAPI, dstBucket, dstEntry
 // Implements S3 compatible Upload Part API.
 func (er erasureObjects) PutObjectPart(ctx context.Context, bucket, object, uploadID string, partID int, r *PutObjReader, opts ObjectOptions) (pi PartInfo, err error) {
 	if !opts.NoAuditLog {
-		auditObjectErasureSet(ctx, object, &er)
+		auditObjectErasureSet(ctx, "PutObjectPart", object, &er)
 	}
 
 	data := r.Reader
@@ -768,19 +705,6 @@ func (er erasureObjects) PutObjectPart(ctx context.Context, bucket, object, uplo
 
 	// Rename temporary part file to its final location.
 	partPath := pathJoin(uploadIDPath, fi.DataDir, partSuffix)
-	onlineDisks, err = renamePart(ctx, onlineDisks, minioMetaTmpBucket, tmpPartPath, minioMetaMultipartBucket, partPath, writeQuorum)
-	if err != nil {
-		if errors.Is(err, errFileNotFound) {
-			// An in-quorum errFileNotFound means that client stream
-			// prematurely closed and we do not find any xl.meta or
-			// part.1's - in such a scenario we must return as if client
-			// disconnected. This means that erasure.Encode() CreateFile()
-			// did not do anything.
-			return pi, IncompleteBody{Bucket: bucket, Object: object}
-		}
-
-		return pi, toObjectErr(err, minioMetaMultipartBucket, partPath)
-	}
 
 	md5hex := r.MD5CurrentHexString()
 	if opts.PreserveETag != "" {
@@ -802,15 +726,22 @@ func (er erasureObjects) PutObjectPart(ctx context.Context, bucket, object, uplo
 		Checksums:  r.ContentCRC(),
 	}
 
-	fi.Parts = []ObjectPartInfo{partInfo}
-	partFI, err := fi.MarshalMsg(nil)
+	partFI, err := partInfo.MarshalMsg(nil)
 	if err != nil {
 		return pi, toObjectErr(err, minioMetaMultipartBucket, partPath)
 	}
 
-	// Write part metadata to all disks.
-	onlineDisks, err = writeAllDisks(ctx, onlineDisks, minioMetaMultipartBucket, partPath+".meta", partFI, writeQuorum)
+	onlineDisks, err = er.renamePart(ctx, onlineDisks, minioMetaTmpBucket, tmpPartPath, minioMetaMultipartBucket, partPath, partFI, writeQuorum)
 	if err != nil {
+		if errors.Is(err, errFileNotFound) {
+			// An in-quorum errFileNotFound means that client stream
+			// prematurely closed and we do not find any xl.meta or
+			// part.1's - in such a scenario we must return as if client
+			// disconnected. This means that erasure.Encode() CreateFile()
+			// did not do anything.
+			return pi, IncompleteBody{Bucket: bucket, Object: object}
+		}
+
 		return pi, toObjectErr(err, minioMetaMultipartBucket, partPath)
 	}
 
@@ -835,7 +766,7 @@ func (er erasureObjects) PutObjectPart(ctx context.Context, bucket, object, uplo
 // Does not contain currently uploaded parts by design.
 func (er erasureObjects) GetMultipartInfo(ctx context.Context, bucket, object, uploadID string, opts ObjectOptions) (MultipartInfo, error) {
 	if !opts.NoAuditLog {
-		auditObjectErasureSet(ctx, object, &er)
+		auditObjectErasureSet(ctx, "GetMultipartInfo", object, &er)
 	}
 
 	result := MultipartInfo{
@@ -873,7 +804,7 @@ func (er erasureObjects) GetMultipartInfo(ctx context.Context, bucket, object, u
 // replied back to the client.
 func (er erasureObjects) ListObjectParts(ctx context.Context, bucket, object, uploadID string, partNumberMarker, maxParts int, opts ObjectOptions) (result ListPartsInfo, err error) {
 	if !opts.NoAuditLog {
-		auditObjectErasureSet(ctx, object, &er)
+		auditObjectErasureSet(ctx, "ListObjectParts", object, &er)
 	}
 
 	uploadIDLock := er.NewNSLock(bucket, pathJoin(object, uploadID))
@@ -953,7 +884,7 @@ func (er erasureObjects) ListObjectParts(ctx context.Context, bucket, object, up
 
 	g := errgroup.WithNErrs(len(req.Files)).WithConcurrency(32)
 
-	partsInfo := make([]ObjectPartInfo, len(req.Files))
+	partsInfo := make([]*ObjectPartInfo, len(req.Files))
 	for i, file := range req.Files {
 		file := file
 		partN := i + start
@@ -965,21 +896,17 @@ func (er erasureObjects) ListObjectParts(ctx context.Context, bucket, object, up
 				return err
 			}
 
-			var pfi FileInfo
-			_, err = pfi.UnmarshalMsg(buf)
+			pinfo := &ObjectPartInfo{}
+			_, err = pinfo.UnmarshalMsg(buf)
 			if err != nil {
 				return err
 			}
 
-			if len(pfi.Parts) != 1 {
-				return errors.New("invalid number of parts expected 1, got 0")
+			if partN != pinfo.Number {
+				return fmt.Errorf("part.%d.meta has incorrect corresponding part number: expected %d, got %d", partN, partN, pinfo.Number)
 			}
 
-			if partN != pfi.Parts[0].Number {
-				return fmt.Errorf("part.%d.meta has incorrect corresponding part number: expected %d, got %d", partN, partN, pfi.Parts[0].Number)
-			}
-
-			partsInfo[i] = pfi.Parts[0]
+			partsInfo[i] = pinfo
 			return nil
 		}, i)
 	}
@@ -987,7 +914,7 @@ func (er erasureObjects) ListObjectParts(ctx context.Context, bucket, object, up
 	g.Wait()
 
 	for _, part := range partsInfo {
-		if part.Number != 0 && !part.ModTime.IsZero() {
+		if part != nil && part.Number != 0 && !part.ModTime.IsZero() {
 			fi.AddObjectPart(part.Number, part.ETag, part.Size, part.ActualSize, part.ModTime, part.Index, part.Checksums)
 		}
 	}
@@ -1023,6 +950,106 @@ func (er erasureObjects) ListObjectParts(ctx context.Context, bucket, object, up
 	return result, nil
 }
 
+func readParts(ctx context.Context, disks []StorageAPI, bucket string, partMetaPaths []string, partNumbers []int, readQuorum int) ([]*ObjectPartInfo, error) {
+	g := errgroup.WithNErrs(len(disks))
+
+	objectPartInfos := make([][]*ObjectPartInfo, len(disks))
+	// Rename file on all underlying storage disks.
+	for index := range disks {
+		index := index
+		g.Go(func() (err error) {
+			if disks[index] == nil {
+				return errDiskNotFound
+			}
+			objectPartInfos[index], err = disks[index].ReadParts(ctx, bucket, partMetaPaths...)
+			return err
+		}, index)
+	}
+
+	if err := reduceReadQuorumErrs(ctx, g.Wait(), objectOpIgnoredErrs, readQuorum); err != nil {
+		return nil, err
+	}
+
+	partInfosInQuorum := make([]*ObjectPartInfo, len(partMetaPaths))
+	partMetaQuorumMap := make(map[string]int, len(partNumbers))
+	for pidx := range partMetaPaths {
+		var pinfos []*ObjectPartInfo
+		for idx := range disks {
+			if len(objectPartInfos[idx]) == 0 {
+				partMetaQuorumMap[partMetaPaths[pidx]]++
+				continue
+			}
+
+			pinfo := objectPartInfos[idx][pidx]
+			if pinfo == nil {
+				partMetaQuorumMap[partMetaPaths[pidx]]++
+				continue
+			}
+
+			if pinfo.ETag == "" {
+				partMetaQuorumMap[partMetaPaths[pidx]]++
+			} else {
+				pinfos = append(pinfos, pinfo)
+				partMetaQuorumMap[pinfo.ETag]++
+			}
+		}
+
+		var maxQuorum int
+		var maxETag string
+		var maxPartMeta string
+		for etag, quorum := range partMetaQuorumMap {
+			if maxQuorum < quorum {
+				maxQuorum = quorum
+				maxETag = etag
+				maxPartMeta = etag
+			}
+		}
+
+		var pinfo *ObjectPartInfo
+		for _, pinfo = range pinfos {
+			if pinfo != nil && maxETag != "" && pinfo.ETag == maxETag {
+				break
+			}
+			if maxPartMeta != "" && path.Base(maxPartMeta) == fmt.Sprintf("part.%d.meta", pinfo.Number) {
+				break
+			}
+		}
+
+		if pinfo != nil && pinfo.ETag != "" && partMetaQuorumMap[maxETag] >= readQuorum {
+			partInfosInQuorum[pidx] = pinfo
+			continue
+		}
+
+		if partMetaQuorumMap[maxPartMeta] == len(disks) {
+			if pinfo != nil && pinfo.Error != "" {
+				partInfosInQuorum[pidx] = &ObjectPartInfo{Error: pinfo.Error}
+			} else {
+				partInfosInQuorum[pidx] = &ObjectPartInfo{
+					Error: InvalidPart{
+						PartNumber: partNumbers[pidx],
+					}.Error(),
+				}
+			}
+		} else {
+			partInfosInQuorum[pidx] = &ObjectPartInfo{Error: errErasureReadQuorum.Error()}
+		}
+	}
+	return partInfosInQuorum, nil
+}
+
+func errStrToPartErr(errStr string) error {
+	if strings.Contains(errStr, "file not found") {
+		return InvalidPart{}
+	}
+	if strings.Contains(errStr, "Specified part could not be found") {
+		return InvalidPart{}
+	}
+	if strings.Contains(errStr, errErasureReadQuorum.Error()) {
+		return errErasureReadQuorum
+	}
+	return errors.New(errStr)
+}
+
 // CompleteMultipartUpload - completes an ongoing multipart
 // transaction after receiving all the parts indicated by the client.
 // Returns an md5sum calculated by concatenating all the individual
@@ -1031,7 +1058,7 @@ func (er erasureObjects) ListObjectParts(ctx context.Context, bucket, object, up
 // Implements S3 compatible Complete multipart API.
 func (er erasureObjects) CompleteMultipartUpload(ctx context.Context, bucket string, object string, uploadID string, parts []CompletePart, opts ObjectOptions) (oi ObjectInfo, err error) {
 	if !opts.NoAuditLog {
-		auditObjectErasureSet(ctx, object, &er)
+		auditObjectErasureSet(ctx, "CompleteMultipartUpload", object, &er)
 	}
 
 	if opts.CheckPrecondFn != nil {
@@ -1076,24 +1103,22 @@ func (er erasureObjects) CompleteMultipartUpload(ctx context.Context, bucket str
 	uploadIDPath := er.getUploadIDDir(bucket, object, uploadID)
 	onlineDisks := er.getDisks()
 	writeQuorum := fi.WriteQuorum(er.defaultWQuorum())
+	readQuorum := fi.ReadQuorum(er.defaultRQuorum())
 
 	// Read Part info for all parts
 	partPath := pathJoin(uploadIDPath, fi.DataDir) + "/"
-	req := ReadMultipleReq{
-		Bucket:       minioMetaMultipartBucket,
-		Prefix:       partPath,
-		MaxSize:      1 << 20, // Each part should realistically not be > 1MiB.
-		Files:        make([]string, 0, len(parts)),
-		AbortOn404:   true,
-		MetadataOnly: true,
+	partMetaPaths := make([]string, len(parts))
+	partNumbers := make([]int, len(parts))
+	for idx, part := range parts {
+		partMetaPaths[idx] = pathJoin(partPath, fmt.Sprintf("part.%d.meta", part.PartNumber))
+		partNumbers[idx] = part.PartNumber
 	}
-	for _, part := range parts {
-		req.Files = append(req.Files, fmt.Sprintf("part.%d.meta", part.PartNumber))
-	}
-	partInfoFiles, err := readMultipleFiles(ctx, onlineDisks, req, writeQuorum)
+
+	partInfoFiles, err := readParts(ctx, onlineDisks, minioMetaMultipartBucket, partMetaPaths, partNumbers, readQuorum)
 	if err != nil {
 		return oi, err
 	}
+
 	if len(partInfoFiles) != len(parts) {
 		// Should only happen through internal error
 		err := fmt.Errorf("unexpected part result count: %d, want %d", len(partInfoFiles), len(parts))
@@ -1155,35 +1180,22 @@ func (er erasureObjects) CompleteMultipartUpload(ctx context.Context, bucket str
 		opts.EncryptFn = metadataEncrypter(key)
 	}
 
-	for i, part := range partInfoFiles {
-		partID := parts[i].PartNumber
-		if part.Error != "" || !part.Exists {
-			return oi, InvalidPart{
-				PartNumber: partID,
-			}
-		}
-
-		var pfi FileInfo
-		_, err := pfi.UnmarshalMsg(part.Data)
-		if err != nil {
-			// Maybe crash or similar.
+	for idx, part := range partInfoFiles {
+		if part.Error != "" {
+			err = errStrToPartErr(part.Error)
 			bugLogIf(ctx, err)
-			return oi, InvalidPart{
-				PartNumber: partID,
-			}
+			return oi, err
 		}
 
-		partI := pfi.Parts[0]
-		partNumber := partI.Number
-		if partID != partNumber {
-			internalLogIf(ctx, fmt.Errorf("part.%d.meta has incorrect corresponding part number: expected %d, got %d", partID, partID, partI.Number))
+		if parts[idx].PartNumber != part.Number {
+			internalLogIf(ctx, fmt.Errorf("part.%d.meta has incorrect corresponding part number: expected %d, got %d", parts[idx].PartNumber, parts[idx].PartNumber, part.Number))
 			return oi, InvalidPart{
-				PartNumber: partID,
+				PartNumber: part.Number,
 			}
 		}
 
 		// Add the current part.
-		fi.AddObjectPart(partI.Number, partI.ETag, partI.Size, partI.ActualSize, partI.ModTime, partI.Index, partI.Checksums)
+		fi.AddObjectPart(part.Number, part.ETag, part.Size, part.ActualSize, part.ModTime, part.Index, part.Checksums)
 	}
 
 	// Calculate full object size.
@@ -1337,7 +1349,9 @@ func (er erasureObjects) CompleteMultipartUpload(ctx context.Context, bucket str
 
 	// Save the consolidated actual size.
 	if opts.ReplicationRequest {
-		fi.Metadata[ReservedMetadataPrefix+"actual-size"] = opts.UserDefined["X-Minio-Internal-Actual-Object-Size"]
+		if v := opts.UserDefined[ReservedMetadataPrefix+"Actual-Object-Size"]; v != "" {
+			fi.Metadata[ReservedMetadataPrefix+"actual-size"] = v
+		}
 	} else {
 		fi.Metadata[ReservedMetadataPrefix+"actual-size"] = strconv.FormatInt(objectActualSize, 10)
 	}
@@ -1359,10 +1373,10 @@ func (er erasureObjects) CompleteMultipartUpload(ctx context.Context, bucket str
 		}
 	}
 
+	paths := make([]string, 0, len(currentFI.Parts))
 	// Remove parts that weren't present in CompleteMultipartUpload request.
 	for _, curpart := range currentFI.Parts {
-		// Remove part.meta which is not needed anymore.
-		er.removePartMeta(bucket, object, uploadID, currentFI.DataDir, curpart.Number)
+		paths = append(paths, pathJoin(uploadIDPath, currentFI.DataDir, fmt.Sprintf("part.%d.meta", curpart.Number)))
 
 		if objectPartIndex(fi.Parts, curpart.Number) == -1 {
 			// Delete the missing part files. e.g,
@@ -1371,10 +1385,12 @@ func (er erasureObjects) CompleteMultipartUpload(ctx context.Context, bucket str
 			// Request 3: PutObjectPart 2
 			// Request 4: CompleteMultipartUpload --part 2
 			// N.B. 1st part is not present. This part should be removed from the storage.
-			er.removeObjectPart(bucket, object, uploadID, currentFI.DataDir, curpart.Number)
+			paths = append(paths, pathJoin(uploadIDPath, currentFI.DataDir, fmt.Sprintf("part.%d", curpart.Number)))
 		}
 	}
 
+	er.cleanupMultipartPath(ctx, paths...) // cleanup all part.N.meta, and skipped part.N's before final rename().
+
 	defer func() {
 		if err == nil {
 			er.deleteAll(context.Background(), minioMetaMultipartBucket, uploadIDPath)
@@ -1393,13 +1409,13 @@ func (er erasureObjects) CompleteMultipartUpload(ctx context.Context, bucket str
 	}
 
 	if !opts.Speedtest && len(versions) > 0 {
-		globalMRFState.addPartialOp(partialOperation{
-			bucket:    bucket,
-			object:    object,
-			queued:    time.Now(),
-			versions:  versions,
-			setIndex:  er.setIndex,
-			poolIndex: er.poolIndex,
+		globalMRFState.addPartialOp(PartialOperation{
+			Bucket:    bucket,
+			Object:    object,
+			Queued:    time.Now(),
+			Versions:  versions,
+			SetIndex:  er.setIndex,
+			PoolIndex: er.poolIndex,
 		})
 	}
 
@@ -1439,7 +1455,7 @@ func (er erasureObjects) CompleteMultipartUpload(ctx context.Context, bucket str
 // operation.
 func (er erasureObjects) AbortMultipartUpload(ctx context.Context, bucket, object, uploadID string, opts ObjectOptions) (err error) {
 	if !opts.NoAuditLog {
-		auditObjectErasureSet(ctx, object, &er)
+		auditObjectErasureSet(ctx, "AbortMultipartUpload", object, &er)
 	}
 
 	lk := er.NewNSLock(bucket, pathJoin(object, uploadID))

cmd/erasure-object.go

@@ -70,7 +70,7 @@ func countOnlineDisks(onlineDisks []StorageAPI) (online int) {
 // update metadata.
 func (er erasureObjects) CopyObject(ctx context.Context, srcBucket, srcObject, dstBucket, dstObject string, srcInfo ObjectInfo, srcOpts, dstOpts ObjectOptions) (oi ObjectInfo, err error) {
 	if !dstOpts.NoAuditLog {
-		auditObjectErasureSet(ctx, dstObject, &er)
+		auditObjectErasureSet(ctx, "CopyObject", dstObject, &er)
 	}
 
 	// This call shouldn't be used for anything other than metadata updates or adding self referential versions.
@@ -97,7 +97,7 @@ func (er erasureObjects) CopyObject(ctx context.Context, srcBucket, srcObject, d
 	if srcOpts.VersionID != "" {
 		metaArr, errs = readAllFileInfo(ctx, storageDisks, "", srcBucket, srcObject, srcOpts.VersionID, true, false)
 	} else {
-		metaArr, errs = readAllXL(ctx, storageDisks, srcBucket, srcObject, true, false, true)
+		metaArr, errs = readAllXL(ctx, storageDisks, srcBucket, srcObject, true, false)
 	}
 
 	readQuorum, writeQuorum, err := objectQuorumFromMeta(ctx, metaArr, errs, er.defaultParityCount)
@@ -199,7 +199,7 @@ func (er erasureObjects) CopyObject(ctx context.Context, srcBucket, srcObject, d
 // Read(Closer). When err != nil, the returned reader is always nil.
 func (er erasureObjects) GetObjectNInfo(ctx context.Context, bucket, object string, rs *HTTPRangeSpec, h http.Header, opts ObjectOptions) (gr *GetObjectReader, err error) {
 	if !opts.NoAuditLog {
-		auditObjectErasureSet(ctx, object, &er)
+		auditObjectErasureSet(ctx, "GetObject", object, &er)
 	}
 
 	var unlockOnDefer bool
@@ -283,7 +283,7 @@ func (er erasureObjects) GetObjectNInfo(ctx context.Context, bucket, object stri
 	}
 
 	if unlockOnDefer {
-		unlockOnDefer = fi.InlineData()
+		unlockOnDefer = fi.InlineData() || len(fi.Data) > 0
 	}
 
 	pr, pw := xioutil.WaitPipe()
@@ -395,24 +395,16 @@ func (er erasureObjects) getObjectWithFileInfo(ctx context.Context, bucket, obje
 			// that we have some parts or data blocks missing or corrupted
 			// - attempt a heal to successfully heal them for future calls.
 			if written == partLength {
-				var scan madmin.HealScanMode
-				switch {
-				case errors.Is(err, errFileNotFound):
-					scan = madmin.HealNormalScan
-				case errors.Is(err, errFileCorrupt):
-					scan = madmin.HealDeepScan
-				}
-				switch scan {
-				case madmin.HealNormalScan, madmin.HealDeepScan:
+				if errors.Is(err, errFileNotFound) || errors.Is(err, errFileCorrupt) {
 					healOnce.Do(func() {
-						globalMRFState.addPartialOp(partialOperation{
-							bucket:    bucket,
-							object:    object,
-							versionID: fi.VersionID,
-							queued:    time.Now(),
-							setIndex:  er.setIndex,
-							poolIndex: er.poolIndex,
-							scanMode:  scan,
+						globalMRFState.addPartialOp(PartialOperation{
+							Bucket:     bucket,
+							Object:     object,
+							VersionID:  fi.VersionID,
+							Queued:     time.Now(),
+							SetIndex:   er.setIndex,
+							PoolIndex:  er.poolIndex,
+							BitrotScan: errors.Is(err, errFileCorrupt),
 						})
 					})
 					// Healing is triggered and we have written
@@ -439,7 +431,7 @@ func (er erasureObjects) getObjectWithFileInfo(ctx context.Context, bucket, obje
 // GetObjectInfo - reads object metadata and replies back ObjectInfo.
 func (er erasureObjects) GetObjectInfo(ctx context.Context, bucket, object string, opts ObjectOptions) (info ObjectInfo, err error) {
 	if !opts.NoAuditLog {
-		auditObjectErasureSet(ctx, object, &er)
+		auditObjectErasureSet(ctx, "GetObjectInfo", object, &er)
 	}
 
 	if !opts.NoLock {
@@ -456,7 +448,7 @@ func (er erasureObjects) GetObjectInfo(ctx context.Context, bucket, object strin
 	return er.getObjectInfo(ctx, bucket, object, opts)
 }
 
-func auditDanglingObjectDeletion(ctx context.Context, bucket, object, versionID string, tags map[string]interface{}) {
+func auditDanglingObjectDeletion(ctx context.Context, bucket, object, versionID string, tags map[string]string) {
 	if len(logger.AuditTargets()) == 0 {
 		return
 	}
@@ -472,13 +464,16 @@ func auditDanglingObjectDeletion(ctx context.Context, bucket, object, versionID
 	auditLogInternal(ctx, opts)
 }
 
-func joinErrs(errs []error) []string {
-	s := make([]string, len(errs))
+func joinErrs(errs []error) string {
+	var s string
 	for i := range s {
+		if s != "" {
+			s += ","
+		}
 		if errs[i] == nil {
-			s[i] = "<nil>"
+			s += "<nil>"
 		} else {
-			s[i] = errs[i].Error()
+			s += errs[i].Error()
 		}
 	}
 	return s
@@ -491,20 +486,18 @@ func (er erasureObjects) deleteIfDangling(ctx context.Context, bucket, object st
 		// can be deleted safely, in such a scenario return ReadQuorum error.
 		return FileInfo{}, errErasureReadQuorum
 	}
-	tags := make(map[string]interface{}, 4)
-	tags["set"] = er.setIndex
-	tags["pool"] = er.poolIndex
+	tags := make(map[string]string, 16)
+	tags["set"] = strconv.Itoa(er.setIndex)
+	tags["pool"] = strconv.Itoa(er.poolIndex)
 	tags["merrs"] = joinErrs(errs)
-	tags["derrs"] = dataErrsByPart
+	tags["derrs"] = fmt.Sprintf("%v", dataErrsByPart)
 	if m.IsValid() {
-		tags["size"] = m.Size
-		tags["mtime"] = m.ModTime.Format(http.TimeFormat)
-		tags["data"] = m.Erasure.DataBlocks
-		tags["parity"] = m.Erasure.ParityBlocks
+		tags["sz"] = strconv.FormatInt(m.Size, 10)
+		tags["mt"] = m.ModTime.Format(iso8601Format)
+		tags["d:p"] = fmt.Sprintf("%d:%d", m.Erasure.DataBlocks, m.Erasure.ParityBlocks)
 	} else {
-		tags["invalid-meta"] = true
-		tags["data"] = er.setDriveCount - er.defaultParityCount
-		tags["parity"] = er.defaultParityCount
+		tags["invalid"] = "1"
+		tags["d:p"] = fmt.Sprintf("%d:%d", er.setDriveCount-er.defaultParityCount, er.defaultParityCount)
 	}
 
 	// count the number of offline disks
@@ -527,7 +520,7 @@ func (er erasureObjects) deleteIfDangling(ctx context.Context, bucket, object st
 		}
 	}
 	if offline > 0 {
-		tags["offline"] = offline
+		tags["offline"] = strconv.Itoa(offline)
 	}
 
 	_, file, line, cok := runtime.Caller(1)
@@ -556,55 +549,24 @@ func (er erasureObjects) deleteIfDangling(ctx context.Context, bucket, object st
 		}, index)
 	}
 
-	rmDisks := make(map[string]string, len(disks))
 	for index, err := range g.Wait() {
-		var errStr, diskName string
+		var errStr string
 		if err != nil {
 			errStr = err.Error()
 		} else {
 			errStr = "<nil>"
 		}
-		if disks[index] != nil {
-			diskName = disks[index].String()
-		} else {
-			diskName = fmt.Sprintf("disk-%d", index)
-		}
-		rmDisks[diskName] = errStr
+		tags[fmt.Sprintf("ddisk-%d", index)] = errStr
 	}
-	tags["cleanupResult"] = rmDisks
+
 	return m, nil
 }
 
-func fileInfoFromRaw(ri RawFileInfo, bucket, object string, readData, inclFreeVers, allParts bool) (FileInfo, error) {
-	var xl xlMetaV2
-	if err := xl.LoadOrConvert(ri.Buf); err != nil {
-		return FileInfo{}, errFileCorrupt
-	}
-
-	fi, err := xl.ToFileInfo(bucket, object, "", inclFreeVers, allParts)
-	if err != nil {
-		return FileInfo{}, err
-	}
-
-	if !fi.IsValid() {
-		return FileInfo{}, errFileCorrupt
-	}
-
-	versionID := fi.VersionID
-	if versionID == "" {
-		versionID = nullVersionID
-	}
-
-	fileInfo, err := xl.ToFileInfo(bucket, object, versionID, inclFreeVers, allParts)
-	if err != nil {
-		return FileInfo{}, err
-	}
-
-	if readData {
-		fileInfo.Data = xl.data.find(versionID)
-	}
-
-	return fileInfo, nil
+func fileInfoFromRaw(ri RawFileInfo, bucket, object string, readData, inclFreeVers bool) (FileInfo, error) {
+	return getFileInfo(ri.Buf, bucket, object, "", fileInfoOpts{
+		Data:             readData,
+		InclFreeVersions: inclFreeVers,
+	})
 }
 
 func readAllRawFileInfo(ctx context.Context, disks []StorageAPI, bucket, object string, readData bool) ([]RawFileInfo, []error) {
@@ -628,7 +590,7 @@ func readAllRawFileInfo(ctx context.Context, disks []StorageAPI, bucket, object
 	return rawFileInfos, g.Wait()
 }
 
-func pickLatestQuorumFilesInfo(ctx context.Context, rawFileInfos []RawFileInfo, errs []error, bucket, object string, readData, inclFreeVers, allParts bool) ([]FileInfo, []error) {
+func pickLatestQuorumFilesInfo(ctx context.Context, rawFileInfos []RawFileInfo, errs []error, bucket, object string, readData, inclFreeVers bool) ([]FileInfo, []error) {
 	metadataArray := make([]*xlMetaV2, len(rawFileInfos))
 	metaFileInfos := make([]FileInfo, len(rawFileInfos))
 	metadataShallowVersions := make([][]xlMetaV2ShallowVersion, len(rawFileInfos))
@@ -665,7 +627,7 @@ func pickLatestQuorumFilesInfo(ctx context.Context, rawFileInfos []RawFileInfo,
 
 	readQuorum := (len(rawFileInfos) + 1) / 2
 	meta := &xlMetaV2{versions: mergeXLV2Versions(readQuorum, false, 1, metadataShallowVersions...)}
-	lfi, err := meta.ToFileInfo(bucket, object, "", inclFreeVers, allParts)
+	lfi, err := meta.ToFileInfo(bucket, object, "", inclFreeVers, true)
 	if err != nil {
 		for i := range errs {
 			if errs[i] == nil {
@@ -694,7 +656,7 @@ func pickLatestQuorumFilesInfo(ctx context.Context, rawFileInfos []RawFileInfo,
 		}
 
 		// make sure to preserve this for diskmtime based healing bugfix.
-		metaFileInfos[index], errs[index] = metadataArray[index].ToFileInfo(bucket, object, versionID, inclFreeVers, allParts)
+		metaFileInfos[index], errs[index] = metadataArray[index].ToFileInfo(bucket, object, versionID, inclFreeVers, true)
 		if errs[index] != nil {
 			continue
 		}
@@ -739,9 +701,9 @@ func shouldCheckForDangling(err error, errs []error, bucket string) bool {
 	return false
 }
 
-func readAllXL(ctx context.Context, disks []StorageAPI, bucket, object string, readData, inclFreeVers, allParts bool) ([]FileInfo, []error) {
+func readAllXL(ctx context.Context, disks []StorageAPI, bucket, object string, readData, inclFreeVers bool) ([]FileInfo, []error) {
 	rawFileInfos, errs := readAllRawFileInfo(ctx, disks, bucket, object, readData)
-	return pickLatestQuorumFilesInfo(ctx, rawFileInfos, errs, bucket, object, readData, inclFreeVers, allParts)
+	return pickLatestQuorumFilesInfo(ctx, rawFileInfos, errs, bucket, object, readData, inclFreeVers)
 }
 
 func (er erasureObjects) getObjectFileInfo(ctx context.Context, bucket, object string, opts ObjectOptions, readData bool) (FileInfo, []FileInfo, []StorageAPI, error) {
@@ -756,8 +718,9 @@ func (er erasureObjects) getObjectFileInfo(ctx context.Context, bucket, object s
 	disks := er.getDisks()
 
 	ropts := ReadOptions{
-		ReadData: readData,
-		Healing:  false,
+		ReadData:         readData,
+		InclFreeVersions: opts.InclFreeVersions,
+		Healing:          false,
 	}
 
 	mrfCheck := make(chan FileInfo)
@@ -797,7 +760,7 @@ func (er erasureObjects) getObjectFileInfo(ctx context.Context, bucket, object s
 					// Read the latest version
 					rfi, err = disk.ReadXL(ctx, bucket, object, readData)
 					if err == nil {
-						fi, err = fileInfoFromRaw(rfi, bucket, object, readData, opts.InclFreeVersions, true)
+						fi, err = fileInfoFromRaw(rfi, bucket, object, readData, opts.InclFreeVersions)
 					}
 				}
 
@@ -843,13 +806,13 @@ func (er erasureObjects) getObjectFileInfo(ctx context.Context, bucket, object s
 		// additionally do not heal delete markers inline, let them be
 		// healed upon regular heal process.
 		if missingBlocks > 0 && missingBlocks < fi.Erasure.DataBlocks {
-			globalMRFState.addPartialOp(partialOperation{
-				bucket:    fi.Volume,
-				object:    fi.Name,
-				versionID: fi.VersionID,
-				queued:    time.Now(),
-				setIndex:  er.setIndex,
-				poolIndex: er.poolIndex,
+			globalMRFState.addPartialOp(PartialOperation{
+				Bucket:    fi.Volume,
+				Object:    fi.Name,
+				VersionID: fi.VersionID,
+				Queued:    time.Now(),
+				SetIndex:  er.setIndex,
+				PoolIndex: er.poolIndex,
 			})
 		}
 
@@ -931,14 +894,16 @@ func (er erasureObjects) getObjectFileInfo(ctx context.Context, bucket, object s
 		}
 
 		rw.Lock()
+		// when its a versioned bucket and empty versionID - at totalResp == setDriveCount
+		// we must use rawFileInfo to resolve versions to figure out the latest version.
 		if opts.VersionID == "" && totalResp == er.setDriveCount {
 			fi, onlineMeta, onlineDisks, modTime, etag, err = calcQuorum(pickLatestQuorumFilesInfo(ctx,
-				rawArr, errs, bucket, object, readData, opts.InclFreeVersions, true))
+				rawArr, errs, bucket, object, readData, opts.InclFreeVersions))
 		} else {
 			fi, onlineMeta, onlineDisks, modTime, etag, err = calcQuorum(metaArr, errs)
 		}
 		rw.Unlock()
-		if err == nil && fi.InlineData() {
+		if err == nil && (fi.InlineData() || len(fi.Data) > 0) {
 			break
 		}
 	}
@@ -1273,7 +1238,7 @@ func (er erasureObjects) PutObject(ctx context.Context, bucket string, object st
 // putObject wrapper for erasureObjects PutObject
 func (er erasureObjects) putObject(ctx context.Context, bucket string, object string, r *PutObjReader, opts ObjectOptions) (objInfo ObjectInfo, err error) {
 	if !opts.NoAuditLog {
-		auditObjectErasureSet(ctx, object, &er)
+		auditObjectErasureSet(ctx, "PutObject", object, &er)
 	}
 
 	data := r.Reader
@@ -1422,7 +1387,7 @@ func (er erasureObjects) putObject(ctx context.Context, bucket string, object st
 	writers := make([]io.Writer, len(onlineDisks))
 	var inlineBuffers []*bytes.Buffer
 	if shardFileSize >= 0 {
-		if !opts.Versioned && shardFileSize < inlineBlock {
+		if !opts.Versioned && shardFileSize <= inlineBlock {
 			inlineBuffers = make([]*bytes.Buffer, len(onlineDisks))
 		} else if shardFileSize < inlineBlock/8 {
 			inlineBuffers = make([]*bytes.Buffer, len(onlineDisks))
@@ -1430,7 +1395,7 @@ func (er erasureObjects) putObject(ctx context.Context, bucket string, object st
 	} else {
 		// If compressed, use actual size to determine.
 		if sz := erasure.ShardFileSize(data.ActualSize()); sz > 0 {
-			if !opts.Versioned && sz < inlineBlock {
+			if !opts.Versioned && sz <= inlineBlock {
 				inlineBuffers = make([]*bytes.Buffer, len(onlineDisks))
 			} else if sz < inlineBlock/8 {
 				inlineBuffers = make([]*bytes.Buffer, len(onlineDisks))
@@ -1599,13 +1564,13 @@ func (er erasureObjects) putObject(ctx context.Context, bucket string, object st
 				break
 			}
 		} else {
-			globalMRFState.addPartialOp(partialOperation{
-				bucket:    bucket,
-				object:    object,
-				queued:    time.Now(),
-				versions:  versions,
-				setIndex:  er.setIndex,
-				poolIndex: er.poolIndex,
+			globalMRFState.addPartialOp(PartialOperation{
+				Bucket:    bucket,
+				Object:    object,
+				Queued:    time.Now(),
+				Versions:  versions,
+				SetIndex:  er.setIndex,
+				PoolIndex: er.poolIndex,
 			})
 		}
 	}
@@ -1648,7 +1613,7 @@ func (er erasureObjects) deleteObjectVersion(ctx context.Context, bucket, object
 func (er erasureObjects) DeleteObjects(ctx context.Context, bucket string, objects []ObjectToDelete, opts ObjectOptions) ([]DeletedObject, []error) {
 	if !opts.NoAuditLog {
 		for _, obj := range objects {
-			auditObjectErasureSet(ctx, obj.ObjectV.ObjectName, &er)
+			auditObjectErasureSet(ctx, "DeleteObjects", obj.ObjectV.ObjectName, &er)
 		}
 	}
 
@@ -1868,7 +1833,7 @@ func (er erasureObjects) deletePrefix(ctx context.Context, bucket, prefix string
 // response to the client request.
 func (er erasureObjects) DeleteObject(ctx context.Context, bucket, object string, opts ObjectOptions) (objInfo ObjectInfo, err error) {
 	if !opts.NoAuditLog {
-		auditObjectErasureSet(ctx, object, &er)
+		auditObjectErasureSet(ctx, "DeleteObject", object, &er)
 	}
 
 	var lc *lifecycle.Lifecycle
@@ -2134,11 +2099,11 @@ func (er erasureObjects) DeleteObject(ctx context.Context, bucket, object string
 // Send the successful but partial upload/delete, however ignore
 // if the channel is blocked by other items.
 func (er erasureObjects) addPartial(bucket, object, versionID string) {
-	globalMRFState.addPartialOp(partialOperation{
-		bucket:    bucket,
-		object:    object,
-		versionID: versionID,
-		queued:    time.Now(),
+	globalMRFState.addPartialOp(PartialOperation{
+		Bucket:    bucket,
+		Object:    object,
+		VersionID: versionID,
+		Queued:    time.Now(),
 	})
 }
 
@@ -2163,7 +2128,7 @@ func (er erasureObjects) PutObjectMetadata(ctx context.Context, bucket, object s
 	if opts.VersionID != "" {
 		metaArr, errs = readAllFileInfo(ctx, disks, "", bucket, object, opts.VersionID, false, false)
 	} else {
-		metaArr, errs = readAllXL(ctx, disks, bucket, object, false, false, true)
+		metaArr, errs = readAllXL(ctx, disks, bucket, object, false, false)
 	}
 
 	readQuorum, _, err := objectQuorumFromMeta(ctx, metaArr, errs, er.defaultParityCount)
@@ -2240,7 +2205,7 @@ func (er erasureObjects) PutObjectTags(ctx context.Context, bucket, object strin
 	if opts.VersionID != "" {
 		metaArr, errs = readAllFileInfo(ctx, disks, "", bucket, object, opts.VersionID, false, false)
 	} else {
-		metaArr, errs = readAllXL(ctx, disks, bucket, object, false, false, true)
+		metaArr, errs = readAllXL(ctx, disks, bucket, object, false, false)
 	}
 
 	readQuorum, _, err := objectQuorumFromMeta(ctx, metaArr, errs, er.defaultParityCount)

cmd/erasure-server-pool-decom.go

@@ -358,7 +358,7 @@ func (p *poolMeta) validate(pools []*erasureSets) (bool, error) {
 			update = true
 		}
 		if ok && pi.completed {
-			return false, fmt.Errorf("pool(%s) = %s is decommissioned, please remove from server command line", humanize.Ordinal(pi.position+1), k)
+			logger.LogIf(GlobalContext, "decommission", fmt.Errorf("pool(%s) = %s is decommissioned, please remove from server command line", humanize.Ordinal(pi.position+1), k))
 		}
 	}
 
@@ -608,7 +608,7 @@ func (z *erasureServerPools) IsDecommissionRunning() bool {
 	return false
 }
 
-func (z *erasureServerPools) decommissionObject(ctx context.Context, bucket string, gr *GetObjectReader) (err error) {
+func (z *erasureServerPools) decommissionObject(ctx context.Context, idx int, bucket string, gr *GetObjectReader) (err error) {
 	objInfo := gr.ObjInfo
 
 	defer func() {
@@ -623,9 +623,11 @@ func (z *erasureServerPools) decommissionObject(ctx context.Context, bucket stri
 
 	if objInfo.isMultipart() {
 		res, err := z.NewMultipartUpload(ctx, bucket, objInfo.Name, ObjectOptions{
-			VersionID:   objInfo.VersionID,
-			UserDefined: objInfo.UserDefined,
-			NoAuditLog:  true,
+			VersionID:    objInfo.VersionID,
+			UserDefined:  objInfo.UserDefined,
+			NoAuditLog:   true,
+			SrcPoolIdx:   idx,
+			DataMovement: true,
 		})
 		if err != nil {
 			return fmt.Errorf("decommissionObject: NewMultipartUpload() %w", err)
@@ -660,6 +662,7 @@ func (z *erasureServerPools) decommissionObject(ctx context.Context, bucket stri
 			}
 		}
 		_, err = z.CompleteMultipartUpload(ctx, bucket, objInfo.Name, res.UploadID, parts, ObjectOptions{
+			SrcPoolIdx:   idx,
 			DataMovement: true,
 			MTime:        objInfo.ModTime,
 			NoAuditLog:   true,
@@ -681,6 +684,7 @@ func (z *erasureServerPools) decommissionObject(ctx context.Context, bucket stri
 		NewPutObjReader(hr),
 		ObjectOptions{
 			DataMovement: true,
+			SrcPoolIdx:   idx,
 			VersionID:    objInfo.VersionID,
 			MTime:        objInfo.ModTime,
 			UserDefined:  objInfo.UserDefined,
@@ -855,6 +859,7 @@ func (z *erasureServerPools) decommissionPool(ctx context.Context, idx int, pool
 					versionID = nullVersionID
 				}
 
+				var failure, ignore bool
 				if version.Deleted {
 					_, err := z.DeleteObject(ctx,
 						bi.Name,
@@ -867,14 +872,19 @@ func (z *erasureServerPools) decommissionPool(ctx context.Context, idx int, pool
 							VersionID:          versionID,
 							MTime:              version.ModTime,
 							DeleteReplication:  version.ReplicationState,
+							SrcPoolIdx:         idx,
+							DataMovement:       true,
 							DeleteMarker:       true, // make sure we create a delete marker
 							SkipDecommissioned: true, // make sure we skip the decommissioned pool
 							NoAuditLog:         true,
 						})
-					var failure bool
 					if err != nil {
-						if isErrObjectNotFound(err) || isErrVersionNotFound(err) {
-							err = nil
+						// This can happen when rebalance stop races with ongoing rebalance workers.
+						// These rebalance failures can be ignored.
+						if isErrObjectNotFound(err) || isErrVersionNotFound(err) || isDataMovementOverWriteErr(err) {
+							ignore = true
+							stopFn(0, nil)
+							continue
 						}
 					}
 					stopFn(version.Size, err)
@@ -893,22 +903,26 @@ func (z *erasureServerPools) decommissionPool(ctx context.Context, idx int, pool
 					continue
 				}
 
-				var failure, ignore bool
 				// gr.Close() is ensured by decommissionObject().
 				for try := 0; try < 3; try++ {
 					if version.IsRemote() {
 						if err := z.DecomTieredObject(ctx, bi.Name, version.Name, version, ObjectOptions{
-							VersionID:   versionID,
-							MTime:       version.ModTime,
-							UserDefined: version.Metadata,
+							VersionID:    versionID,
+							MTime:        version.ModTime,
+							UserDefined:  version.Metadata,
+							SrcPoolIdx:   idx,
+							DataMovement: true,
 						}); err != nil {
-							stopFn(version.Size, err)
-							failure = true
-							decomLogIf(ctx, err)
-							continue
+							if isErrObjectNotFound(err) || isErrVersionNotFound(err) || isDataMovementOverWriteErr(err) {
+								ignore = true
+								stopFn(0, nil)
+							}
+						}
+						if !ignore {
+							stopFn(version.Size, err)
+							failure = err != nil
+							decomLogIf(ctx, err)
 						}
-						stopFn(version.Size, nil)
-						failure = false
 						break
 					}
 					gr, err := set.GetObjectNInfo(ctx,
@@ -925,7 +939,7 @@ func (z *erasureServerPools) decommissionPool(ctx context.Context, idx int, pool
 					if isErrObjectNotFound(err) || isErrVersionNotFound(err) {
 						// object deleted by the application, nothing to do here we move on.
 						ignore = true
-						stopFn(version.Size, nil)
+						stopFn(0, nil)
 						break
 					}
 					if err != nil && !ignore {
@@ -943,7 +957,12 @@ func (z *erasureServerPools) decommissionPool(ctx context.Context, idx int, pool
 						stopFn(version.Size, err)
 						continue
 					}
-					if err = z.decommissionObject(ctx, bi.Name, gr); err != nil {
+					if err = z.decommissionObject(ctx, idx, bi.Name, gr); err != nil {
+						if isErrObjectNotFound(err) || isErrVersionNotFound(err) || isDataMovementOverWriteErr(err) {
+							ignore = true
+							stopFn(0, nil)
+							break
+						}
 						stopFn(version.Size, err)
 						failure = true
 						decomLogIf(ctx, err)

cmd/erasure-server-pool-decom_test.go

@@ -134,7 +134,7 @@ func TestPoolMetaValidate(t *testing.T) {
 			meta:           nmeta1,
 			pools:          pools,
 			name:           "Invalid-Completed-Pool-Not-Removed",
-			expectedErr:    true,
+			expectedErr:    false,
 			expectedUpdate: false,
 		},
 		{

cmd/erasure-server-pool-rebalance.go

@@ -119,11 +119,8 @@ func (z *erasureServerPools) loadRebalanceMeta(ctx context.Context) error {
 	}
 
 	z.rebalMu.Lock()
-	if len(r.PoolStats) == len(z.serverPools) {
-		z.rebalMeta = r
-	} else {
-		z.updateRebalanceStats(ctx)
-	}
+	z.rebalMeta = r
+	z.updateRebalanceStats(ctx)
 	z.rebalMu.Unlock()
 
 	return nil
@@ -147,24 +144,16 @@ func (z *erasureServerPools) updateRebalanceStats(ctx context.Context) error {
 		}
 	}
 	if ok {
-		lock := z.serverPools[0].NewNSLock(minioMetaBucket, rebalMetaName)
-		lkCtx, err := lock.GetLock(ctx, globalOperationTimeout)
-		if err != nil {
-			rebalanceLogIf(ctx, fmt.Errorf("failed to acquire write lock on %s/%s: %w", minioMetaBucket, rebalMetaName, err))
-			return err
-		}
-		defer lock.Unlock(lkCtx)
-
-		ctx = lkCtx.Context()
-
-		noLockOpts := ObjectOptions{NoLock: true}
-		return z.rebalMeta.saveWithOpts(ctx, z.serverPools[0], noLockOpts)
+		return z.rebalMeta.save(ctx, z.serverPools[0])
 	}
 
 	return nil
 }
 
 func (z *erasureServerPools) findIndex(index int) int {
+	if z.rebalMeta == nil {
+		return 0
+	}
 	for i := 0; i < len(z.rebalMeta.PoolStats); i++ {
 		if i == index {
 			return index
@@ -277,6 +266,10 @@ func (z *erasureServerPools) bucketRebalanceDone(bucket string, poolIdx int) {
 	z.rebalMu.Lock()
 	defer z.rebalMu.Unlock()
 
+	if z.rebalMeta == nil {
+		return
+	}
+
 	ps := z.rebalMeta.PoolStats[poolIdx]
 	if ps == nil {
 		return
@@ -331,6 +324,10 @@ func (r *rebalanceMeta) loadWithOpts(ctx context.Context, store objectIO, opts O
 }
 
 func (r *rebalanceMeta) saveWithOpts(ctx context.Context, store objectIO, opts ObjectOptions) error {
+	if r == nil {
+		return nil
+	}
+
 	data := make([]byte, 4, r.Msgsize()+4)
 
 	// Initialize the header.
@@ -353,8 +350,15 @@ func (z *erasureServerPools) IsRebalanceStarted() bool {
 	z.rebalMu.RLock()
 	defer z.rebalMu.RUnlock()
 
-	if r := z.rebalMeta; r != nil {
-		if r.StoppedAt.IsZero() {
+	r := z.rebalMeta
+	if r == nil {
+		return false
+	}
+	if !r.StoppedAt.IsZero() {
+		return false
+	}
+	for _, ps := range r.PoolStats {
+		if ps.Participating && ps.Info.Status != rebalCompleted {
 			return true
 		}
 	}
@@ -369,7 +373,7 @@ func (z *erasureServerPools) IsPoolRebalancing(poolIndex int) bool {
 		if !r.StoppedAt.IsZero() {
 			return false
 		}
-		ps := z.rebalMeta.PoolStats[poolIndex]
+		ps := r.PoolStats[poolIndex]
 		return ps.Participating && ps.Info.Status == rebalStarted
 	}
 	return false
@@ -622,14 +626,18 @@ func (z *erasureServerPools) rebalanceBucket(ctx context.Context, bucket string,
 
 			var rebalanced, expired int
 			for _, version := range fivs.Versions {
+				stopFn := globalRebalanceMetrics.log(rebalanceMetricRebalanceObject, poolIdx, bucket, version.Name, version.VersionID)
+
 				// Skip transitioned objects for now. TBD
 				if version.IsRemote() {
+					stopFn(version.Size, errors.New("ILM Tiered version will be skipped for now"))
 					continue
 				}
 
 				// Apply lifecycle rules on the objects that are expired.
 				if filterLifecycle(bucket, version.Name, version) {
 					expired++
+					stopFn(version.Size, errors.New("ILM expired object/version will be skipped"))
 					continue
 				}
 
@@ -639,6 +647,7 @@ func (z *erasureServerPools) rebalanceBucket(ctx context.Context, bucket string,
 				remainingVersions := len(fivs.Versions) - expired
 				if version.Deleted && remainingVersions == 1 {
 					rebalanced++
+					stopFn(version.Size, errors.New("DELETE marked object with no other non-current versions will be skipped"))
 					continue
 				}
 
@@ -647,6 +656,7 @@ func (z *erasureServerPools) rebalanceBucket(ctx context.Context, bucket string,
 					versionID = nullVersionID
 				}
 
+				var failure, ignore bool
 				if version.Deleted {
 					_, err := z.DeleteObject(ctx,
 						bucket,
@@ -656,16 +666,26 @@ func (z *erasureServerPools) rebalanceBucket(ctx context.Context, bucket string,
 							VersionID:         versionID,
 							MTime:             version.ModTime,
 							DeleteReplication: version.ReplicationState,
+							SrcPoolIdx:        poolIdx,
+							DataMovement:      true,
 							DeleteMarker:      true, // make sure we create a delete marker
 							SkipRebalancing:   true, // make sure we skip the decommissioned pool
 							NoAuditLog:        true,
 						})
-					var failure bool
-					if err != nil && !isErrObjectNotFound(err) && !isErrVersionNotFound(err) {
-						rebalanceLogIf(ctx, err)
-						failure = true
+					// This can happen when rebalance stop races with ongoing rebalance workers.
+					// These rebalance failures can be ignored.
+					if err != nil {
+						// This can happen when rebalance stop races with ongoing rebalance workers.
+						// These rebalance failures can be ignored.
+						if isErrObjectNotFound(err) || isErrVersionNotFound(err) || isDataMovementOverWriteErr(err) {
+							ignore = true
+							stopFn(0, nil)
+							continue
+						}
 					}
-
+					stopFn(version.Size, err)
+					rebalanceLogIf(ctx, err)
+					failure = err != nil
 					if !failure {
 						z.updatePoolStats(poolIdx, bucket, version)
 						rebalanced++
@@ -674,10 +694,8 @@ func (z *erasureServerPools) rebalanceBucket(ctx context.Context, bucket string,
 					continue
 				}
 
-				var failure, ignore bool
 				for try := 0; try < 3; try++ {
 					// GetObjectReader.Close is called by rebalanceObject
-					stopFn := globalRebalanceMetrics.log(rebalanceMetricRebalanceObject, poolIdx, bucket, version.Name, version.VersionID)
 					gr, err := set.GetObjectNInfo(ctx,
 						bucket,
 						encodeDirObject(version.Name),
@@ -702,7 +720,14 @@ func (z *erasureServerPools) rebalanceBucket(ctx context.Context, bucket string,
 						continue
 					}
 
-					if err = z.rebalanceObject(ctx, bucket, gr); err != nil {
+					if err = z.rebalanceObject(ctx, poolIdx, bucket, gr); err != nil {
+						// This can happen when rebalance stop races with ongoing rebalance workers.
+						// These rebalance failures can be ignored.
+						if isErrObjectNotFound(err) || isErrVersionNotFound(err) || isDataMovementOverWriteErr(err) {
+							ignore = true
+							stopFn(0, nil)
+							break
+						}
 						failure = true
 						rebalanceLogIf(ctx, err)
 						stopFn(version.Size, err)
@@ -794,7 +819,9 @@ func (z *erasureServerPools) saveRebalanceStats(ctx context.Context, poolIdx int
 	case rebalSaveStoppedAt:
 		r.StoppedAt = time.Now()
 	case rebalSaveStats:
-		r.PoolStats[poolIdx] = z.rebalMeta.PoolStats[poolIdx]
+		if z.rebalMeta != nil {
+			r.PoolStats[poolIdx] = z.rebalMeta.PoolStats[poolIdx]
+		}
 	}
 	z.rebalMeta = r
 
@@ -816,7 +843,7 @@ func auditLogRebalance(ctx context.Context, apiName, bucket, object, versionID s
 	})
 }
 
-func (z *erasureServerPools) rebalanceObject(ctx context.Context, bucket string, gr *GetObjectReader) (err error) {
+func (z *erasureServerPools) rebalanceObject(ctx context.Context, poolIdx int, bucket string, gr *GetObjectReader) (err error) {
 	oi := gr.ObjInfo
 
 	defer func() {
@@ -831,9 +858,11 @@ func (z *erasureServerPools) rebalanceObject(ctx context.Context, bucket string,
 
 	if oi.isMultipart() {
 		res, err := z.NewMultipartUpload(ctx, bucket, oi.Name, ObjectOptions{
-			VersionID:   oi.VersionID,
-			UserDefined: oi.UserDefined,
-			NoAuditLog:  true,
+			VersionID:    oi.VersionID,
+			UserDefined:  oi.UserDefined,
+			NoAuditLog:   true,
+			DataMovement: true,
+			SrcPoolIdx:   poolIdx,
 		})
 		if err != nil {
 			return fmt.Errorf("rebalanceObject: NewMultipartUpload() %w", err)
@@ -885,6 +914,7 @@ func (z *erasureServerPools) rebalanceObject(ctx context.Context, bucket string,
 		oi.Name,
 		NewPutObjReader(hr),
 		ObjectOptions{
+			SrcPoolIdx:   poolIdx,
 			DataMovement: true,
 			VersionID:    oi.VersionID,
 			MTime:        oi.ModTime,
@@ -975,6 +1005,8 @@ const (
 	rebalanceMetricSaveMetadata
 )
 
+var errDataMovementSrcDstPoolSame = errors.New("source and destination pool are the same")
+
 func rebalanceTrace(r rebalanceMetric, poolIdx int, startTime time.Time, duration time.Duration, err error, path string, sz int64) madmin.TraceInfo {
 	var errStr string
 	if err != nil {

cmd/erasure-server-pool.go

@@ -168,7 +168,7 @@ func newErasureServerPools(ctx context.Context, endpointServerPools EndpointServ
 
 	if !globalIsDistErasure {
 		globalLocalDrivesMu.Lock()
-		globalLocalDrives = localDrives
+		globalLocalDrivesMap = make(map[string]StorageAPI, len(localDrives))
 		for _, drive := range localDrives {
 			globalLocalDrivesMap[drive.Endpoint().String()] = drive
 		}
@@ -187,19 +187,12 @@ func newErasureServerPools(ctx context.Context, endpointServerPools EndpointServ
 		globalLeaderLock = newSharedLock(GlobalContext, z, "leader.lock")
 	})
 
-	// Enable background operations on
-	//
-	// - Disk auto healing
-	// - MRF (most recently failed) healing
-	// - Background expiration routine for lifecycle policies
+	// Start self healing after the object initialization
+	// so various tasks will be useful
 	bootstrapTrace("initAutoHeal", func() {
 		initAutoHeal(GlobalContext, z)
 	})
 
-	bootstrapTrace("initHealMRF", func() {
-		go globalMRFState.healRoutine(z)
-	})
-
 	// initialize the object layer.
 	defer setObjectLayer(z)
 
@@ -1087,6 +1080,14 @@ func (z *erasureServerPools) PutObject(ctx context.Context, bucket string, objec
 		return ObjectInfo{}, err
 	}
 
+	if opts.DataMovement && idx == opts.SrcPoolIdx {
+		return ObjectInfo{}, DataMovementOverwriteErr{
+			Bucket:    bucket,
+			Object:    object,
+			VersionID: opts.VersionID,
+			Err:       errDataMovementSrcDstPoolSame,
+		}
+	}
 	// Overwrite the object at the right pool
 	return z.serverPools[idx].PutObject(ctx, bucket, object, data, opts)
 }
@@ -1139,6 +1140,16 @@ func (z *erasureServerPools) DeleteObject(ctx context.Context, bucket string, ob
 		return pinfo.ObjInfo, nil
 	}
 
+	// Datamovement must never be allowed on the same pool.
+	if opts.DataMovement && opts.SrcPoolIdx == pinfo.Index {
+		return pinfo.ObjInfo, DataMovementOverwriteErr{
+			Bucket:    bucket,
+			Object:    object,
+			VersionID: opts.VersionID,
+			Err:       errDataMovementSrcDstPoolSame,
+		}
+	}
+
 	// Delete concurrently in all server pools with read quorum error for unversioned objects.
 	if len(noReadQuorumPools) > 0 && !opts.Versioned && !opts.VersionSuspended {
 		return z.deleteObjectFromAllPools(ctx, bucket, object, opts, noReadQuorumPools)
@@ -1526,14 +1537,12 @@ func (z *erasureServerPools) listObjectsGeneric(ctx context.Context, bucket, pre
 			loi.NextMarker = last.Name
 		}
 
-		if merged.lastSkippedEntry != "" {
-			if merged.lastSkippedEntry > loi.NextMarker {
-				// An object hidden by ILM was found during listing. Since the number of entries
-				// fetched from drives is limited, set IsTruncated to true to ask the s3 client
-				// to continue listing if it wishes in order to find if there is more objects.
-				loi.IsTruncated = true
-				loi.NextMarker = merged.lastSkippedEntry
-			}
+		if loi.IsTruncated && merged.lastSkippedEntry > loi.NextMarker {
+			// An object hidden by ILM was found during a truncated listing. Since the number of entries
+			// fetched from drives is limited by max-keys, we should use the last ILM filtered entry
+			// as a continuation token if it is lexially higher than the last visible object so that the
+			// next call of WalkDir() with the max-keys can reach new objects not seen previously.
+			loi.NextMarker = merged.lastSkippedEntry
 		}
 
 		if loi.NextMarker != "" {
@@ -1754,6 +1763,15 @@ func (z *erasureServerPools) NewMultipartUpload(ctx context.Context, bucket, obj
 		return nil, err
 	}
 
+	if opts.DataMovement && idx == opts.SrcPoolIdx {
+		return nil, DataMovementOverwriteErr{
+			Bucket:    bucket,
+			Object:    object,
+			VersionID: opts.VersionID,
+			Err:       errDataMovementSrcDstPoolSame,
+		}
+	}
+
 	return z.serverPools[idx].NewMultipartUpload(ctx, bucket, object, opts)
 }
 
@@ -1962,6 +1980,34 @@ func (z *erasureServerPools) DeleteBucket(ctx context.Context, bucket string, op
 		defer lk.Unlock(lkctx)
 	}
 
+	if !opts.Force {
+		results := make(chan itemOrErr[ObjectInfo])
+
+		ctx, cancel := context.WithTimeout(ctx, time.Minute)
+		defer cancel()
+		err := z.Walk(ctx, bucket, "", results, WalkOptions{Limit: 1})
+		if err != nil {
+			s3LogIf(ctx, fmt.Errorf("unable to verify if the bucket %s is empty: %w", bucket, err))
+			return toObjectErr(err, bucket)
+		}
+
+		select {
+		case <-ctx.Done():
+			return ctx.Err()
+		case r, found := <-results:
+			if found {
+				if r.Err != nil {
+					s3LogIf(ctx, fmt.Errorf("unable to verify if the bucket %s is empty: %w", bucket, r.Err))
+					return toObjectErr(r.Err, bucket)
+				}
+				return toObjectErr(errVolumeNotEmpty, bucket)
+			}
+		}
+
+		// Always pass force to the lower level
+		opts.Force = true
+	}
+
 	err := z.s3Peer.DeleteBucket(ctx, bucket, opts)
 	if err == nil || isErrBucketNotFound(err) {
 		// If site replication is configured, hold on to deleted bucket state until sites sync
@@ -2008,10 +2054,12 @@ func (z *erasureServerPools) ListBuckets(ctx context.Context, opts BucketOptions
 				if err != nil {
 					return nil, err
 				}
-				for i := range buckets {
-					createdAt, err := globalBucketMetadataSys.CreatedAt(buckets[i].Name)
-					if err == nil {
-						buckets[i].Created = createdAt
+				if !opts.NoMetadata {
+					for i := range buckets {
+						createdAt, err := globalBucketMetadataSys.CreatedAt(buckets[i].Name)
+						if err == nil {
+							buckets[i].Created = createdAt
+						}
 					}
 				}
 				return buckets, nil
@@ -2025,10 +2073,13 @@ func (z *erasureServerPools) ListBuckets(ctx context.Context, opts BucketOptions
 	if err != nil {
 		return nil, err
 	}
-	for i := range buckets {
-		createdAt, err := globalBucketMetadataSys.CreatedAt(buckets[i].Name)
-		if err == nil {
-			buckets[i].Created = createdAt
+
+	if !opts.NoMetadata {
+		for i := range buckets {
+			createdAt, err := globalBucketMetadataSys.CreatedAt(buckets[i].Name)
+			if err == nil {
+				buckets[i].Created = createdAt
+			}
 		}
 	}
 	return buckets, nil
@@ -2082,7 +2133,6 @@ func (z *erasureServerPools) HealBucket(ctx context.Context, bucket string, opts
 
 // Walk a bucket, optionally prefix recursively, until we have returned
 // all the contents of the provided bucket+prefix.
-// TODO: Note that most errors will result in a truncated listing.
 func (z *erasureServerPools) Walk(ctx context.Context, bucket, prefix string, results chan<- itemOrErr[ObjectInfo], opts WalkOptions) error {
 	if err := checkListObjsArgs(ctx, bucket, prefix, ""); err != nil {
 		xioutil.SafeClose(results)
@@ -2100,7 +2150,7 @@ func (z *erasureServerPools) Walk(ctx context.Context, bucket, prefix string, re
 			disks, infos, _ := set.getOnlineDisksWithHealingAndInfo(true)
 			if len(disks) == 0 {
 				xioutil.SafeClose(results)
-				err := fmt.Errorf("Walk: no online disks found in pool %d, set %d", setIdx, poolIdx)
+				err := fmt.Errorf("Walk: no online disks found in (set:%d pool:%d) %w", setIdx, poolIdx, errErasureReadQuorum)
 				cancelCause(err)
 				return err
 			}
@@ -2176,6 +2226,7 @@ func (z *erasureServerPools) Walk(ctx context.Context, bucket, prefix string, re
 					filterPrefix:   filterPrefix,
 					recursive:      true,
 					forwardTo:      opts.Marker,
+					perDiskLimit:   opts.Limit,
 					minDisks:       listingQuorum,
 					reportNotFound: false,
 					agreed:         send,
@@ -2199,9 +2250,8 @@ func (z *erasureServerPools) Walk(ctx context.Context, bucket, prefix string, re
 	// Convert and filter merged entries.
 	merged := make(chan metaCacheEntry, 100)
 	vcfg, _ := globalBucketVersioningSys.Get(bucket)
+	errCh := make(chan error, 1)
 	go func() {
-		defer cancelCause(nil)
-		defer xioutil.SafeClose(results)
 		sentErr := false
 		sendErr := func(err error) {
 			if !sentErr {
@@ -2212,6 +2262,15 @@ func (z *erasureServerPools) Walk(ctx context.Context, bucket, prefix string, re
 				}
 			}
 		}
+		defer func() {
+			select {
+			case <-ctx.Done():
+				sendErr(ctx.Err())
+			default:
+			}
+			xioutil.SafeClose(results)
+			cancelCause(nil)
+		}()
 		send := func(oi ObjectInfo) bool {
 			select {
 			case results <- itemOrErr[ObjectInfo]{Item: oi}:
@@ -2266,12 +2325,16 @@ func (z *erasureServerPools) Walk(ctx context.Context, bucket, prefix string, re
 				}
 			}
 		}
+		if err := <-errCh; err != nil {
+			sendErr(err)
+		}
 	}()
 	go func() {
+		defer close(errCh)
 		// Merge all entries from all disks.
 		// We leave quorum at 1, since entries are already resolved to have the desired quorum.
 		// mergeEntryChannels will close 'merged' channel upon completion or cancellation.
-		storageLogIf(ctx, mergeEntryChannels(ctx, entries, merged, 1))
+		errCh <- mergeEntryChannels(ctx, entries, merged, 1)
 	}()
 
 	return nil
@@ -2327,12 +2390,18 @@ func (z *erasureServerPools) HealObjects(ctx context.Context, bucket, prefix str
 
 	var poolErrs [][]error
 	for idx, erasureSet := range z.serverPools {
+		if opts.Pool != nil && *opts.Pool != idx {
+			continue
+		}
 		if z.IsSuspended(idx) {
 			continue
 		}
 		errs := make([]error, len(erasureSet.sets))
 		var wg sync.WaitGroup
 		for idx, set := range erasureSet.sets {
+			if opts.Set != nil && *opts.Set != idx {
+				continue
+			}
 			wg.Add(1)
 			go func(idx int, set *erasureObjects) {
 				defer wg.Done()
@@ -2425,6 +2494,7 @@ const (
 type HealthOptions struct {
 	Maintenance    bool
 	DeploymentType string
+	NoLogging      bool
 }
 
 // HealthResult returns the current state of the system, also
@@ -2449,6 +2519,24 @@ type HealthResult struct {
 	UsingDefaults bool
 }
 
+func (hr HealthResult) String() string {
+	var str strings.Builder
+	for i, es := range hr.ESHealth {
+		str.WriteString("(Pool: ")
+		str.WriteString(strconv.Itoa(es.PoolID))
+		str.WriteString(" Set: ")
+		str.WriteString(strconv.Itoa(es.SetID))
+		str.WriteString(" Healthy: ")
+		str.WriteString(strconv.FormatBool(es.Healthy))
+		if i == 0 {
+			str.WriteString(")")
+		} else {
+			str.WriteString(") | ")
+		}
+	}
+	return str.String()
+}
+
 // Health - returns current status of the object layer health,
 // provides if write access exists across sets, additionally
 // can be used to query scenarios if health may be lost
@@ -2566,18 +2654,18 @@ func (z *erasureServerPools) Health(ctx context.Context, opts HealthOptions) Hea
 			})
 
 			healthy := erasureSetUpCount[poolIdx][setIdx].online >= poolWriteQuorums[poolIdx]
-			if !healthy {
+			if !healthy && !opts.NoLogging {
 				storageLogIf(logger.SetReqInfo(ctx, reqInfo),
-					fmt.Errorf("Write quorum may be lost on pool: %d, set: %d, expected write quorum: %d",
-						poolIdx, setIdx, poolWriteQuorums[poolIdx]), logger.FatalKind)
+					fmt.Errorf("Write quorum could not be established on pool: %d, set: %d, expected write quorum: %d, drives-online: %d",
+						poolIdx, setIdx, poolWriteQuorums[poolIdx], erasureSetUpCount[poolIdx][setIdx].online), logger.FatalKind)
 			}
 			result.Healthy = result.Healthy && healthy
 
 			healthyRead := erasureSetUpCount[poolIdx][setIdx].online >= poolReadQuorums[poolIdx]
-			if !healthyRead {
+			if !healthyRead && !opts.NoLogging {
 				storageLogIf(logger.SetReqInfo(ctx, reqInfo),
-					fmt.Errorf("Read quorum may be lost on pool: %d, set: %d, expected read quorum: %d",
-						poolIdx, setIdx, poolReadQuorums[poolIdx]))
+					fmt.Errorf("Read quorum could not be established on pool: %d, set: %d, expected read quorum: %d, drives-online: %d",
+						poolIdx, setIdx, poolReadQuorums[poolIdx], erasureSetUpCount[poolIdx][setIdx].online))
 			}
 			result.HealthyRead = result.HealthyRead && healthyRead
 		}
@@ -2741,5 +2829,14 @@ func (z *erasureServerPools) DecomTieredObject(ctx context.Context, bucket, obje
 		return err
 	}
 
+	if opts.DataMovement && idx == opts.SrcPoolIdx {
+		return DataMovementOverwriteErr{
+			Bucket:    bucket,
+			Object:    object,
+			VersionID: opts.VersionID,
+			Err:       errDataMovementSrcDstPoolSame,
+		}
+	}
+
 	return z.serverPools[idx].DecomTieredObject(ctx, bucket, object, fi, opts)
 }

cmd/erasure-sets.go

@@ -36,7 +36,6 @@ import (
 	"github.com/minio/minio-go/v7/pkg/set"
 	"github.com/minio/minio-go/v7/pkg/tags"
 	"github.com/minio/minio/internal/dsync"
-	xioutil "github.com/minio/minio/internal/ioutil"
 	"github.com/minio/minio/internal/logger"
 	"github.com/minio/pkg/v3/console"
 	"github.com/minio/pkg/v3/sync/errgroup"
@@ -80,10 +79,6 @@ type erasureSets struct {
 
 	poolIndex int
 
-	// A channel to send the set index to the MRF when
-	// any disk belonging to that set is connected
-	setReconnectEvent chan int
-
 	// Distribution algorithm of choice.
 	distributionAlgo string
 	deploymentID     [16]byte
@@ -125,13 +120,6 @@ func connectEndpoint(endpoint Endpoint) (StorageAPI, *formatErasureV3, error) {
 
 	format, err := loadFormatErasure(disk, false)
 	if err != nil {
-		if errors.Is(err, errUnformattedDisk) {
-			info, derr := disk.DiskInfo(context.TODO(), DiskInfoOptions{})
-			if derr != nil && info.RootDisk {
-				disk.Close()
-				return nil, nil, fmt.Errorf("Drive: %s is a root drive", disk)
-			}
-		}
 		disk.Close()
 		return nil, nil, fmt.Errorf("Drive: %s returned %w", disk, err) // make sure to '%w' to wrap the error
 	}
@@ -201,7 +189,7 @@ func (s *erasureSets) Legacy() (ok bool) {
 
 // connectDisks - attempt to connect all the endpoints, loads format
 // and re-arranges the disks in proper position.
-func (s *erasureSets) connectDisks() {
+func (s *erasureSets) connectDisks(log bool) {
 	defer func() {
 		s.lastConnectDisksOpTime = time.Now()
 	}()
@@ -235,13 +223,18 @@ func (s *erasureSets) connectDisks() {
 			if err != nil {
 				if endpoint.IsLocal && errors.Is(err, errUnformattedDisk) {
 					globalBackgroundHealState.pushHealLocalDisks(endpoint)
-				} else {
-					printEndpointError(endpoint, err, true)
+				} else if !errors.Is(err, errDriveIsRoot) {
+					if log {
+						printEndpointError(endpoint, err, true)
+					}
 				}
 				return
 			}
-			if disk.IsLocal() && disk.Healing() != nil {
-				globalBackgroundHealState.pushHealLocalDisks(disk.Endpoint())
+			if disk.IsLocal() {
+				h := disk.Healing()
+				if h != nil && !h.Finished {
+					globalBackgroundHealState.pushHealLocalDisks(disk.Endpoint())
+				}
 			}
 			s.erasureDisksMu.Lock()
 			setIndex, diskIndex, err := findDiskIndex(s.format, format)
@@ -272,13 +265,7 @@ func (s *erasureSets) connectDisks() {
 				if globalIsDistErasure {
 					globalLocalSetDrives[s.poolIndex][setIndex][diskIndex] = disk
 				}
-				for i, ldisk := range globalLocalDrives {
-					_, k, l := ldisk.GetDiskLoc()
-					if k == setIndex && l == diskIndex {
-						globalLocalDrives[i] = disk
-						break
-					}
-				}
+				globalLocalDrivesMap[disk.Endpoint().String()] = disk
 				globalLocalDrivesMu.Unlock()
 			}
 			s.erasureDisksMu.Unlock()
@@ -297,7 +284,7 @@ func (s *erasureSets) monitorAndConnectEndpoints(ctx context.Context, monitorInt
 	time.Sleep(time.Duration(r.Float64() * float64(time.Second)))
 
 	// Pre-emptively connect the disks if possible.
-	s.connectDisks()
+	s.connectDisks(false)
 
 	monitor := time.NewTimer(monitorInterval)
 	defer monitor.Stop()
@@ -311,7 +298,7 @@ func (s *erasureSets) monitorAndConnectEndpoints(ctx context.Context, monitorInt
 				console.Debugln("running drive monitoring")
 			}
 
-			s.connectDisks()
+			s.connectDisks(true)
 
 			// Reset the timer for next interval
 			monitor.Reset(monitorInterval)
@@ -380,7 +367,6 @@ func newErasureSets(ctx context.Context, endpoints PoolEndpoints, storageDisks [
 		setDriveCount:      setDriveCount,
 		defaultParityCount: defaultParityCount,
 		format:             format,
-		setReconnectEvent:  make(chan int),
 		distributionAlgo:   format.Erasure.DistributionAlgo,
 		deploymentID:       uuid.MustParse(format.ID),
 		poolIndex:          poolIdx,
@@ -542,7 +528,7 @@ func (s *erasureSets) cleanupStaleUploads(ctx context.Context) {
 					if set == nil {
 						return
 					}
-					set.cleanupStaleUploads(ctx, globalAPIConfig.getStaleUploadsExpiry())
+					set.cleanupStaleUploads(ctx)
 				}(set)
 			}
 			wg.Wait()
@@ -554,26 +540,29 @@ func (s *erasureSets) cleanupStaleUploads(ctx context.Context) {
 }
 
 type auditObjectOp struct {
-	Name   string   `json:"name"`
-	Pool   int      `json:"poolId"`
-	Set    int      `json:"setId"`
-	Drives []string `json:"drives"`
+	Name string `json:"name"`
+	Pool int    `json:"poolId"`
+	Set  int    `json:"setId"`
+}
+
+func (op auditObjectOp) String() string {
+	// Flatten the auditObjectOp
+	return fmt.Sprintf("name=%s,pool=%d,set=%d", op.Name, op.Pool, op.Set)
 }
 
 // Add erasure set information to the current context
-func auditObjectErasureSet(ctx context.Context, object string, set *erasureObjects) {
+func auditObjectErasureSet(ctx context.Context, api, object string, set *erasureObjects) {
 	if len(logger.AuditTargets()) == 0 {
 		return
 	}
 
 	op := auditObjectOp{
-		Name:   decodeDirObject(object),
-		Pool:   set.poolIndex + 1,
-		Set:    set.setIndex + 1,
-		Drives: set.getEndpointStrings(),
+		Name: decodeDirObject(object),
+		Pool: set.poolIndex + 1,
+		Set:  set.setIndex + 1,
 	}
 
-	logger.GetReqInfo(ctx).AppendTags("objectLocation", op)
+	logger.GetReqInfo(ctx).AppendTags(api, op.String())
 }
 
 // NewNSLock - initialize a new namespace RWLocker instance.
@@ -662,14 +651,6 @@ func (s *erasureSets) Shutdown(ctx context.Context) error {
 			return err
 		}
 	}
-	select {
-	case _, ok := <-s.setReconnectEvent:
-		if ok {
-			xioutil.SafeClose(s.setReconnectEvent)
-		}
-	default:
-		xioutil.SafeClose(s.setReconnectEvent)
-	}
 	return nil
 }
 
@@ -1154,13 +1135,7 @@ func (s *erasureSets) HealFormat(ctx context.Context, dryRun bool) (res madmin.H
 					if globalIsDistErasure {
 						globalLocalSetDrives[s.poolIndex][m][n] = disk
 					}
-					for i, ldisk := range globalLocalDrives {
-						_, k, l := ldisk.GetDiskLoc()
-						if k == m && l == n {
-							globalLocalDrives[i] = disk
-							break
-						}
-					}
+					globalLocalDrivesMap[disk.Endpoint().String()] = disk
 					globalLocalDrivesMu.Unlock()
 				}
 			}

cmd/erasure.go

@@ -90,6 +90,11 @@ func (er erasureObjects) defaultWQuorum() int {
 	return dataCount
 }
 
+// defaultRQuorum read quorum based on setDriveCount and defaultParityCount
+func (er erasureObjects) defaultRQuorum() int {
+	return er.setDriveCount - er.defaultParityCount
+}
+
 func diskErrToDriveState(err error) (state string) {
 	switch {
 	case errors.Is(err, errDiskNotFound) || errors.Is(err, context.DeadlineExceeded):
@@ -102,6 +107,8 @@ func diskErrToDriveState(err error) (state string) {
 		state = madmin.DriveStatePermission
 	case errors.Is(err, errFaultyDisk):
 		state = madmin.DriveStateFaulty
+	case errors.Is(err, errDriveIsRoot):
+		state = madmin.DriveStateRootMount
 	case err == nil:
 		state = madmin.DriveStateOk
 	default:
@@ -196,11 +203,9 @@ func getDisksInfo(disks []StorageAPI, endpoints []Endpoint, metrics bool) (disks
 			di.State = diskErrToDriveState(err)
 			di.FreeInodes = info.FreeInodes
 			di.UsedInodes = info.UsedInodes
-			if info.Healing {
-				if hi := disks[index].Healing(); hi != nil {
-					hd := hi.toHealingDisk()
-					di.HealInfo = &hd
-				}
+			if hi := disks[index].Healing(); hi != nil {
+				hd := hi.toHealingDisk()
+				di.HealInfo = &hd
 			}
 			di.Metrics = &madmin.DiskMetrics{
 				LastMinute:              make(map[string]madmin.TimedAction, len(info.Metrics.LastMinute)),

cmd/event-notification.go

@@ -49,23 +49,18 @@ func NewEventNotifier(ctx context.Context) *EventNotifier {
 }
 
 // GetARNList - returns available ARNs.
-func (evnot *EventNotifier) GetARNList(onlyActive bool) []string {
+func (evnot *EventNotifier) GetARNList() []string {
 	arns := []string{}
 	if evnot == nil {
 		return arns
 	}
 	region := globalSite.Region()
-	for targetID, target := range evnot.targetList.TargetMap() {
+	for targetID := range evnot.targetList.TargetMap() {
 		// httpclient target is part of ListenNotification
 		// which doesn't need to be listed as part of the ARN list
 		// This list is only meant for external targets, filter
 		// this out pro-actively.
 		if !strings.HasPrefix(targetID.ID, "httpclient+") {
-			if onlyActive {
-				if _, err := target.IsActive(); err != nil {
-					continue
-				}
-			}
 			arns = append(arns, targetID.ToARN(region).String())
 		}
 	}
@@ -74,7 +69,7 @@ func (evnot *EventNotifier) GetARNList(onlyActive bool) []string {
 }
 
 // Loads notification policies for all buckets into EventNotifier.
-func (evnot *EventNotifier) set(bucket BucketInfo, meta BucketMetadata) {
+func (evnot *EventNotifier) set(bucket string, meta BucketMetadata) {
 	config := meta.notificationConfig
 	if config == nil {
 		return
@@ -86,7 +81,7 @@ func (evnot *EventNotifier) set(bucket BucketInfo, meta BucketMetadata) {
 			internalLogIf(GlobalContext, err)
 		}
 	}
-	evnot.AddRulesMap(bucket.Name, config.ToRulesMap())
+	evnot.AddRulesMap(bucket, config.ToRulesMap())
 }
 
 // Targets returns all the registered targets

cmd/fmt-gen.go

@@ -0,0 +1,119 @@
+// Copyright (c) 2015-2024 MinIO, Inc.
+//
+// This file is part of MinIO Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"encoding/json"
+	"log"
+	"os"
+	"path/filepath"
+
+	"github.com/klauspost/compress/zip"
+	"github.com/minio/cli"
+)
+
+var fmtGenFlags = []cli.Flag{
+	cli.IntFlag{
+		Name:  "parity",
+		Usage: "specify erasure code parity",
+	},
+	cli.StringFlag{
+		Name:  "deployment-id",
+		Usage: "deployment-id of the MinIO cluster for which format.json is needed",
+	},
+	cli.StringFlag{
+		Name:   "address",
+		Value:  ":" + GlobalMinioDefaultPort,
+		Usage:  "bind to a specific ADDRESS:PORT, ADDRESS can be an IP or hostname",
+		EnvVar: "MINIO_ADDRESS",
+	},
+}
+
+var fmtGenCmd = cli.Command{
+	Name:   "fmt-gen",
+	Usage:  "Generate format.json files for an erasure server pool",
+	Flags:  append(fmtGenFlags, GlobalFlags...),
+	Action: fmtGenMain,
+	Hidden: true,
+	CustomHelpTemplate: `NAME:
+  {{.HelpName}} - {{.Usage}}
+
+USAGE:
+  {{.HelpName}} {{if .VisibleFlags}}[FLAGS] {{end}}DIR1 [DIR2..]
+  {{.HelpName}} {{if .VisibleFlags}}[FLAGS] {{end}}DIR{1...64}
+  {{.HelpName}} {{if .VisibleFlags}}[FLAGS] {{end}}DIR{1...64} DIR{65...128}
+
+DIR:
+  DIR points to a directory on a filesystem. When you want to combine
+  multiple drives into a single large system, pass one directory per
+  filesystem separated by space. You may also use a '...' convention
+  to abbreviate the directory arguments. Remote directories in a
+  distributed setup are encoded as HTTP(s) URIs.
+{{if .VisibleFlags}}
+FLAGS:
+  {{range .VisibleFlags}}{{.}}
+  {{end}}{{end}}
+EXAMPLES:
+  1. Generate format.json.zip containing format.json files for all drives in a distributed MinIO server pool of 32 nodes with 32 drives each.
+     {{.Prompt}} {{.HelpName}} http://node{1...32}.example.com/mnt/export{1...32}
+
+`,
+}
+
+func fmtGenMain(ctxt *cli.Context) {
+	deploymentID := ctxt.String("deployment-id")
+	err := buildServerCtxt(ctxt, &globalServerCtxt)
+	if err != nil {
+		log.Fatalln(err)
+	}
+	handleCommonArgs(globalServerCtxt)
+	pools, _, err := createServerEndpoints(globalMinioAddr, globalServerCtxt.Layout.pools, globalServerCtxt.Layout.legacy)
+	if err != nil {
+		log.Fatalln(err)
+	}
+
+	zipFile, err := os.Create("format.json.zip")
+	if err != nil {
+		log.Fatalf("failed to create format.json.zip: %v", err)
+	}
+	defer zipFile.Close()
+	fmtZipW := zip.NewWriter(zipFile)
+	defer fmtZipW.Close()
+	for _, pool := range pools { // for each pool
+		setCount, setDriveCount := pool.SetCount, pool.DrivesPerSet
+		format := newFormatErasureV3(setCount, setDriveCount)
+		format.ID = deploymentID
+		for i := 0; i < setCount; i++ { // for each erasure set
+			for j := 0; j < setDriveCount; j++ {
+				newFormat := format.Clone()
+				newFormat.Erasure.This = format.Erasure.Sets[i][j]
+				if deploymentID != "" {
+					newFormat.ID = deploymentID
+				}
+				drive := pool.Endpoints[i*setDriveCount+j]
+				fmtBytes, err := json.Marshal(newFormat)
+				if err != nil {
+					//nolint:gocritic
+					log.Fatalf("failed to marshal format.json for %s: %v", drive.String(), err)
+				}
+				fmtJSON := filepath.Join(drive.Host, drive.Path, minioMetaBucket, "format.json")
+				embedFileInZip(fmtZipW, fmtJSON, fmtBytes, 0o600)
+			}
+		}
+	}
+}

cmd/ftp-server-driver.go

@@ -114,7 +114,7 @@ func ftpTrace(s *ftp.Context, startTime time.Time, source, objPath string, err e
 		TraceType: madmin.TraceFTP,
 		Time:      startTime,
 		NodeName:  globalLocalNodeName,
-		FuncName:  fmt.Sprintf(s.Cmd),
+		FuncName:  s.Cmd,
 		Duration:  time.Since(startTime),
 		Path:      objPath,
 		Error:     errStr,

cmd/generic-handlers.go

@@ -247,11 +247,14 @@ func guessIsRPCReq(req *http.Request) bool {
 	if req == nil {
 		return false
 	}
-	if req.Method == http.MethodGet && req.URL != nil && req.URL.Path == grid.RoutePath {
-		return true
+	if req.Method == http.MethodGet && req.URL != nil {
+		switch req.URL.Path {
+		case grid.RoutePath, grid.RouteLockPath:
+			return true
+		}
 	}
 
-	return req.Method == http.MethodPost &&
+	return (req.Method == http.MethodPost || req.Method == http.MethodGet) &&
 		strings.HasPrefix(req.URL.Path, minioReservedBucketPath+SlashSeparator)
 }
 
@@ -469,7 +472,7 @@ func setBucketForwardingMiddleware(h http.Handler) http.Handler {
 		}
 		if globalDNSConfig == nil || !globalBucketFederation ||
 			guessIsHealthCheckReq(r) || guessIsMetricsReq(r) ||
-			guessIsRPCReq(r) || guessIsLoginSTSReq(r) || isAdminReq(r) {
+			guessIsRPCReq(r) || guessIsLoginSTSReq(r) || isAdminReq(r) || isKMSReq(r) {
 			h.ServeHTTP(w, r)
 			return
 		}

cmd/generic-handlers_test.go

@@ -51,9 +51,10 @@ func TestGuessIsRPC(t *testing.T) {
 	r = &http.Request{
 		Proto:  "HTTP/1.1",
 		Method: http.MethodGet,
+		URL:    u,
 	}
-	if guessIsRPCReq(r) {
-		t.Fatal("Test shouldn't report as net/rpc for a non net/rpc request.")
+	if !guessIsRPCReq(r) {
+		t.Fatal("Test shouldn't fail for a possible net/rpc request.")
 	}
 	r = &http.Request{
 		Proto:  "HTTP/1.1",
@@ -63,6 +64,14 @@ func TestGuessIsRPC(t *testing.T) {
 	if !guessIsRPCReq(r) {
 		t.Fatal("Grid RPC path not detected")
 	}
+	r = &http.Request{
+		Proto:  "HTTP/1.1",
+		Method: http.MethodGet,
+		URL:    &url.URL{Path: grid.RouteLockPath},
+	}
+	if !guessIsRPCReq(r) {
+		t.Fatal("Grid RPC path not detected")
+	}
 }
 
 var isHTTPHeaderSizeTooLargeTests = []struct {

cmd/global-heal.go

@@ -150,6 +150,11 @@ type healEntryResult struct {
 
 // healErasureSet lists and heals all objects in a specific erasure set
 func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string, tracker *healingTracker) error {
+	bgSeq, found := globalBackgroundHealState.getHealSequenceByToken(bgHealingUUID)
+	if !found {
+		return errors.New("no local healing sequence initialized, unable to heal the drive")
+	}
+
 	scanMode := madmin.HealNormalScan
 
 	// Make sure to copy since `buckets slice`
@@ -162,9 +167,21 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 		return errServerNotInitialized
 	}
 
+	started := tracker.Started
+	if started.IsZero() || started.Equal(timeSentinel) {
+		healingLogIf(ctx, fmt.Errorf("unexpected tracker healing start time found: %v", started))
+		started = time.Time{}
+	}
+
+	// Final tracer update before quitting
+	defer func() {
+		tracker.setObject("")
+		tracker.setBucket("")
+		healingLogIf(ctx, tracker.update(ctx))
+	}()
+
 	for _, bucket := range healBuckets {
-		_, err := objAPI.HealBucket(ctx, bucket, madmin.HealOpts{ScanMode: scanMode})
-		if err != nil {
+		if err := bgSeq.healBucket(objAPI, bucket, true); err != nil {
 			// Log bucket healing error if any, we shall retry again.
 			healingLogIf(ctx, err)
 		}
@@ -190,7 +207,7 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 		numHealers = uint64(v)
 	}
 
-	healingLogEvent(ctx, fmt.Sprintf("Healing drive '%s' - use %d parallel workers.", tracker.disk.String(), numHealers))
+	healingLogEvent(ctx, "Healing drive '%s' - use %d parallel workers.", tracker.disk.String(), numHealers)
 
 	jt, _ := workers.New(int(numHealers))
 
@@ -223,7 +240,11 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 
 	// Collect updates to tracker from concurrent healEntry calls
 	results := make(chan healEntryResult, 1000)
-	defer close(results)
+	quitting := make(chan struct{})
+	defer func() {
+		close(results)
+		<-quitting
+	}()
 
 	go func() {
 		for res := range results {
@@ -237,6 +258,9 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 
 			tracker.updateProgress(res.success, res.skipped, res.bytes)
 		}
+
+		healingLogIf(ctx, tracker.update(ctx))
+		close(quitting)
 	}()
 
 	var retErr error
@@ -261,9 +285,7 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 		tracker.setBucket(bucket)
 		// Heal current bucket again in case if it is failed
 		// in the beginning of erasure set healing
-		if _, err := objAPI.HealBucket(ctx, bucket, madmin.HealOpts{
-			ScanMode: scanMode,
-		}); err != nil {
+		if err := bgSeq.healBucket(objAPI, bucket, true); err != nil {
 			// Set this such that when we return this function
 			// we let the caller retry this disk again for the
 			// buckets that failed healing.
@@ -362,6 +384,7 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 				}
 				return false
 			case results <- result:
+				bgSeq.countScanned(madmin.HealItemObject)
 				return true
 			}
 		}
@@ -412,8 +435,10 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 						return
 					}
 					result = healEntryFailure(0)
+					bgSeq.countFailed(madmin.HealItemObject)
 					healingLogIf(ctx, fmt.Errorf("unable to heal object %s/%s: %w", bucket, entry.name, err))
 				} else {
+					bgSeq.countHealed(madmin.HealItemObject)
 					result = healEntrySuccess(uint64(res.ObjectSize))
 				}
 
@@ -423,13 +448,10 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 
 			var versionNotFound int
 			for _, version := range fivs.Versions {
-				// Ignore a version with a modtime newer than healing start time.
-				if version.ModTime.After(tracker.Started) {
-					continue
-				}
-
-				// Apply lifecycle rules on the objects that are expired.
-				if filterLifecycle(bucket, version.Name, version) {
+				// Ignore healing a version if:
+				// - It is uploaded after the drive healing is started
+				// - An object that is already expired by ILM rule.
+				if !started.IsZero() && version.ModTime.After(started) || filterLifecycle(bucket, version.Name, version) {
 					versionNotFound++
 					if !send(healEntrySkipped(uint64(version.Size))) {
 						return
@@ -437,6 +459,8 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 					continue
 				}
 
+				var versionHealed bool
+
 				res, err := er.HealObject(ctx, bucket, encodedEntryName,
 					version.VersionID, madmin.HealOpts{
 						ScanMode: scanMode,
@@ -449,15 +473,24 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 						versionNotFound++
 						continue
 					}
-					// If not deleted, assume they failed.
+				} else {
+					// Look for the healing results
+					if res.After.Drives[tracker.DiskIndex].State == madmin.DriveStateOk {
+						versionHealed = true
+					}
+				}
+
+				if versionHealed {
+					bgSeq.countHealed(madmin.HealItemObject)
+					result = healEntrySuccess(uint64(version.Size))
+				} else {
+					bgSeq.countFailed(madmin.HealItemObject)
 					result = healEntryFailure(uint64(version.Size))
 					if version.VersionID != "" {
 						healingLogIf(ctx, fmt.Errorf("unable to heal object %s/%s-v(%s): %w", bucket, version.Name, version.VersionID, err))
 					} else {
 						healingLogIf(ctx, fmt.Errorf("unable to heal object %s/%s: %w", bucket, version.Name, err))
 					}
-				} else {
-					result = healEntrySuccess(uint64(res.ObjectSize))
 				}
 
 				if !send(result) {
@@ -505,7 +538,11 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 				jt.Take()
 				go healEntry(bucket, *entry)
 			},
-			finished: nil,
+			finished: func(errs []error) {
+				if countErrs(errs, nil) != len(errs) {
+					retErr = fmt.Errorf("one or more errors reported during listing: %v", errors.Join(errs...))
+				}
+			},
 		})
 		jt.Wait() // synchronize all the concurrent heal jobs
 		if err != nil {
@@ -513,7 +550,10 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 			// we let the caller retry this disk again for the
 			// buckets it failed to list.
 			retErr = err
-			healingLogIf(ctx, err)
+		}
+
+		if retErr != nil {
+			healingLogIf(ctx, fmt.Errorf("listing failed with: %v on bucket: %v", retErr, bucket))
 			continue
 		}
 
@@ -526,10 +566,6 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 			healingLogIf(ctx, tracker.update(ctx))
 		}
 	}
-
-	tracker.setObject("")
-	tracker.setBucket("")
-
 	if retErr != nil {
 		return retErr
 	}

cmd/globals.go

@@ -163,7 +163,6 @@ type serverCtxt struct {
 	MemLimit uint64
 
 	UserTimeout         time.Duration
-	ShutdownTimeout     time.Duration
 	IdleTimeout         time.Duration
 	ReadHeaderTimeout   time.Duration
 	MaxIdleConnsPerHost int
@@ -311,6 +310,7 @@ var (
 	globalBootTime = UTCNow()
 
 	globalActiveCred         auth.Credentials
+	globalNodeAuthToken      string
 	globalSiteReplicatorCred siteReplicatorCred
 
 	// Captures if root credentials are set via ENV.
@@ -384,9 +384,7 @@ var (
 	globalBackgroundHealRoutine = newHealRoutine()
 	globalBackgroundHealState   = newHealState(GlobalContext, false)
 
-	globalMRFState = mrfState{
-		opCh: make(chan partialOperation, mrfOpsQueueSize),
-	}
+	globalMRFState = newMRFState()
 
 	// If writes to FS backend should be O_SYNC.
 	globalFSOSync bool
@@ -414,10 +412,9 @@ var (
 	globalServiceFreezeCnt int32
 	globalServiceFreezeMu  sync.Mutex // Updates.
 
-	// List of local drives to this node, this is only set during server startup,
-	// and is only mutated by HealFormat. Hold globalLocalDrivesMu to access.
-	globalLocalDrives    []StorageAPI
-	globalLocalDrivesMap = make(map[string]StorageAPI)
+	// Map of local drives to this node, this is set during server startup,
+	// disk reconnect and mutated by HealFormat. Hold globalLocalDrivesMu to access.
+	globalLocalDrivesMap map[string]StorageAPI
 	globalLocalDrivesMu  sync.RWMutex
 
 	globalDriveMonitoring = env.Get("_MINIO_DRIVE_ACTIVE_MONITORING", config.EnableOn) == config.EnableOn
@@ -450,8 +447,8 @@ var (
 	// dynamic sleeper for multipart expiration routine
 	deleteMultipartCleanupSleeper = newDynamicSleeper(5, 25*time.Millisecond, false)
 
-	// Is _MINIO_DISABLE_API_FREEZE_ON_BOOT set?
-	globalDisableFreezeOnBoot bool
+	// Is MINIO_SYNC_BOOT set?
+	globalEnableSyncBoot bool
 
 	// Contains NIC interface name used for internode communication
 	globalInternodeInterface     string

cmd/grid.go

@@ -31,9 +31,15 @@ import (
 // globalGrid is the global grid manager.
 var globalGrid atomic.Pointer[grid.Manager]
 
+// globalLockGrid is the global lock grid manager.
+var globalLockGrid atomic.Pointer[grid.Manager]
+
 // globalGridStart is a channel that will block startup of grid connections until closed.
 var globalGridStart = make(chan struct{})
 
+// globalLockGridStart is a channel that will block startup of lock grid connections until closed.
+var globalLockGridStart = make(chan struct{})
+
 func initGlobalGrid(ctx context.Context, eps EndpointServerPools) error {
 	hosts, local := eps.GridHosts()
 	lookupHost := globalDNSCache.LookupHost
@@ -41,21 +47,24 @@ func initGlobalGrid(ctx context.Context, eps EndpointServerPools) error {
 		// Pass Dialer for websocket grid, make sure we do not
 		// provide any DriveOPTimeout() function, as that is not
 		// useful over persistent connections.
-		Dialer:       grid.ContextDialer(xhttp.DialContextWithLookupHost(lookupHost, xhttp.NewInternodeDialContext(rest.DefaultTimeout, globalTCPOptions.ForWebsocket()))),
+		Dialer: grid.ConnectWS(
+			grid.ContextDialer(xhttp.DialContextWithLookupHost(lookupHost, xhttp.NewInternodeDialContext(rest.DefaultTimeout, globalTCPOptions.ForWebsocket()))),
+			newCachedAuthToken(),
+			&tls.Config{
+				RootCAs:          globalRootCAs,
+				CipherSuites:     fips.TLSCiphers(),
+				CurvePreferences: fips.TLSCurveIDs(),
+			}),
 		Local:        local,
 		Hosts:        hosts,
-		AddAuth:      newCachedAuthToken(),
-		AuthRequest:  storageServerRequestValidate,
+		AuthToken:    validateStorageRequestToken,
+		AuthFn:       newCachedAuthToken(),
 		BlockConnect: globalGridStart,
-		TLSConfig: &tls.Config{
-			RootCAs:          globalRootCAs,
-			CipherSuites:     fips.TLSCiphers(),
-			CurvePreferences: fips.TLSCurveIDs(),
-		},
 		// Record incoming and outgoing bytes.
-		Incoming: globalConnStats.incInternodeInputBytes,
-		Outgoing: globalConnStats.incInternodeOutputBytes,
-		TraceTo:  globalTrace,
+		Incoming:  globalConnStats.incInternodeInputBytes,
+		Outgoing:  globalConnStats.incInternodeOutputBytes,
+		TraceTo:   globalTrace,
+		RoutePath: grid.RoutePath,
 	})
 	if err != nil {
 		return err
@@ -63,3 +72,36 @@ func initGlobalGrid(ctx context.Context, eps EndpointServerPools) error {
 	globalGrid.Store(g)
 	return nil
 }
+
+func initGlobalLockGrid(ctx context.Context, eps EndpointServerPools) error {
+	hosts, local := eps.GridHosts()
+	lookupHost := globalDNSCache.LookupHost
+	g, err := grid.NewManager(ctx, grid.ManagerOptions{
+		// Pass Dialer for websocket grid, make sure we do not
+		// provide any DriveOPTimeout() function, as that is not
+		// useful over persistent connections.
+		Dialer: grid.ConnectWSWithRoutePath(
+			grid.ContextDialer(xhttp.DialContextWithLookupHost(lookupHost, xhttp.NewInternodeDialContext(rest.DefaultTimeout, globalTCPOptions.ForWebsocket()))),
+			newCachedAuthToken(),
+			&tls.Config{
+				RootCAs:          globalRootCAs,
+				CipherSuites:     fips.TLSCiphers(),
+				CurvePreferences: fips.TLSCurveIDs(),
+			}, grid.RouteLockPath),
+		Local:        local,
+		Hosts:        hosts,
+		AuthToken:    validateStorageRequestToken,
+		AuthFn:       newCachedAuthToken(),
+		BlockConnect: globalGridStart,
+		// Record incoming and outgoing bytes.
+		Incoming:  globalConnStats.incInternodeInputBytes,
+		Outgoing:  globalConnStats.incInternodeOutputBytes,
+		TraceTo:   globalTrace,
+		RoutePath: grid.RouteLockPath,
+	})
+	if err != nil {
+		return err
+	}
+	globalLockGrid.Store(g)
+	return nil
+}

cmd/handler-api.go

@@ -39,14 +39,14 @@ import (
 type apiConfig struct {
 	mu sync.RWMutex
 
-	requestsDeadline      time.Duration
-	requestsPool          chan struct{}
-	clusterDeadline       time.Duration
-	listQuorum            string
-	corsAllowOrigins      []string
-	replicationPriority   string
-	replicationMaxWorkers int
-	transitionWorkers     int
+	requestsPool           chan struct{}
+	clusterDeadline        time.Duration
+	listQuorum             string
+	corsAllowOrigins       []string
+	replicationPriority    string
+	replicationMaxWorkers  int
+	replicationMaxLWorkers int
+	transitionWorkers      int
 
 	staleUploadsExpiry          time.Duration
 	staleUploadsCleanupInterval time.Duration
@@ -163,18 +163,18 @@ func (t *apiConfig) init(cfg api.Config, setDriveCounts []int, legacy bool) {
 		// but this shouldn't last long.
 		t.requestsPool = make(chan struct{}, apiRequestsMaxPerNode)
 	}
-	t.requestsDeadline = cfg.RequestsDeadline
 	listQuorum := cfg.ListQuorum
 	if listQuorum == "" {
 		listQuorum = "strict"
 	}
 	t.listQuorum = listQuorum
-	if globalReplicationPool != nil &&
-		(cfg.ReplicationPriority != t.replicationPriority || cfg.ReplicationMaxWorkers != t.replicationMaxWorkers) {
-		globalReplicationPool.ResizeWorkerPriority(cfg.ReplicationPriority, cfg.ReplicationMaxWorkers)
+	if r := globalReplicationPool.GetNonBlocking(); r != nil &&
+		(cfg.ReplicationPriority != t.replicationPriority || cfg.ReplicationMaxWorkers != t.replicationMaxWorkers || cfg.ReplicationMaxLWorkers != t.replicationMaxLWorkers) {
+		r.ResizeWorkerPriority(cfg.ReplicationPriority, cfg.ReplicationMaxWorkers, cfg.ReplicationMaxLWorkers)
 	}
 	t.replicationPriority = cfg.ReplicationPriority
 	t.replicationMaxWorkers = cfg.ReplicationMaxWorkers
+	t.replicationMaxLWorkers = cfg.ReplicationMaxLWorkers
 
 	// N B api.transition_workers will be deprecated
 	if globalTransitionState != nil {
@@ -288,19 +288,15 @@ func (t *apiConfig) getRequestsPoolCapacity() int {
 	return cap(t.requestsPool)
 }
 
-func (t *apiConfig) getRequestsPool() (chan struct{}, time.Duration) {
+func (t *apiConfig) getRequestsPool() chan struct{} {
 	t.mu.RLock()
 	defer t.mu.RUnlock()
 
 	if t.requestsPool == nil {
-		return nil, 10 * time.Second
+		return nil
 	}
 
-	if t.requestsDeadline <= 0 {
-		return t.requestsPool, 10 * time.Second
-	}
-
-	return t.requestsPool, t.requestsDeadline
+	return t.requestsPool
 }
 
 // maxClients throttles the S3 API calls
@@ -323,27 +319,19 @@ func maxClients(f http.HandlerFunc) http.HandlerFunc {
 		}
 
 		globalHTTPStats.addRequestsInQueue(1)
-		pool, deadline := globalAPIConfig.getRequestsPool()
+		pool := globalAPIConfig.getRequestsPool()
 		if pool == nil {
 			globalHTTPStats.addRequestsInQueue(-1)
 			f.ServeHTTP(w, r)
 			return
 		}
 
-		// No deadline to wait, there is nothing to queue
-		// perform the API call immediately.
-		if deadline <= 0 {
-			globalHTTPStats.addRequestsInQueue(-1)
-			f.ServeHTTP(w, r)
-			return
-		}
-
 		if tc, ok := r.Context().Value(mcontext.ContextTraceKey).(*mcontext.TraceCtxt); ok {
 			tc.FuncName = "s3.MaxClients"
 		}
 
-		deadlineTimer := time.NewTimer(deadline)
-		defer deadlineTimer.Stop()
+		w.Header().Set("X-RateLimit-Limit", strconv.Itoa(cap(pool)))
+		w.Header().Set("X-RateLimit-Remaining", strconv.Itoa(cap(pool)-len(pool)))
 
 		ctx := r.Context()
 		select {
@@ -355,7 +343,13 @@ func maxClients(f http.HandlerFunc) http.HandlerFunc {
 				return
 			}
 			f.ServeHTTP(w, r)
-		case <-deadlineTimer.C:
+		case <-r.Context().Done():
+			globalHTTPStats.addRequestsInQueue(-1)
+			// When the client disconnects before getting the S3 handler
+			// status code response, set the status code to 499 so this request
+			// will be properly audited and traced.
+			w.WriteHeader(499)
+		default:
 			globalHTTPStats.addRequestsInQueue(-1)
 			if contextCanceled(ctx) {
 				w.WriteHeader(499)
@@ -365,12 +359,7 @@ func maxClients(f http.HandlerFunc) http.HandlerFunc {
 			writeErrorResponse(ctx, w,
 				errorCodes.ToAPIErr(ErrTooManyRequests),
 				r.URL)
-		case <-r.Context().Done():
-			globalHTTPStats.addRequestsInQueue(-1)
-			// When the client disconnects before getting the S3 handler
-			// status code response, set the status code to 499 so this request
-			// will be properly audited and traced.
-			w.WriteHeader(499)
+
 		}
 	}
 }
@@ -381,14 +370,16 @@ func (t *apiConfig) getReplicationOpts() replicationPoolOpts {
 
 	if t.replicationPriority == "" {
 		return replicationPoolOpts{
-			Priority:   "auto",
-			MaxWorkers: WorkerMaxLimit,
+			Priority:    "auto",
+			MaxWorkers:  WorkerMaxLimit,
+			MaxLWorkers: LargeWorkerCount,
 		}
 	}
 
 	return replicationPoolOpts{
-		Priority:   t.replicationPriority,
-		MaxWorkers: t.replicationMaxWorkers,
+		Priority:    t.replicationPriority,
+		MaxWorkers:  t.replicationMaxWorkers,
+		MaxLWorkers: t.replicationMaxLWorkers,
 	}
 }
 

cmd/healthcheck-handler.go

@@ -29,14 +29,35 @@ import (
 
 const unavailable = "offline"
 
-// ClusterCheckHandler returns if the server is ready for requests.
-func ClusterCheckHandler(w http.ResponseWriter, r *http.Request) {
-	ctx := newContext(r, w, "ClusterCheckHandler")
-
+func checkHealth(w http.ResponseWriter) ObjectLayer {
 	objLayer := newObjectLayerFn()
 	if objLayer == nil {
 		w.Header().Set(xhttp.MinIOServerStatus, unavailable)
 		writeResponse(w, http.StatusServiceUnavailable, nil, mimeNone)
+		return nil
+	}
+
+	if !globalBucketMetadataSys.Initialized() {
+		w.Header().Set(xhttp.MinIOServerStatus, "bucket-metadata-offline")
+		writeResponse(w, http.StatusServiceUnavailable, nil, mimeNone)
+		return nil
+	}
+
+	if !globalIAMSys.Initialized() {
+		w.Header().Set(xhttp.MinIOServerStatus, "iam-offline")
+		writeResponse(w, http.StatusServiceUnavailable, nil, mimeNone)
+		return nil
+	}
+
+	return objLayer
+}
+
+// ClusterCheckHandler returns if the server is ready for requests.
+func ClusterCheckHandler(w http.ResponseWriter, r *http.Request) {
+	ctx := newContext(r, w, "ClusterCheckHandler")
+
+	objLayer := checkHealth(w)
+	if objLayer == nil {
 		return
 	}
 
@@ -72,10 +93,8 @@ func ClusterCheckHandler(w http.ResponseWriter, r *http.Request) {
 func ClusterReadCheckHandler(w http.ResponseWriter, r *http.Request) {
 	ctx := newContext(r, w, "ClusterReadCheckHandler")
 
-	objLayer := newObjectLayerFn()
+	objLayer := checkHealth(w)
 	if objLayer == nil {
-		w.Header().Set(xhttp.MinIOServerStatus, unavailable)
-		writeResponse(w, http.StatusServiceUnavailable, nil, mimeNone)
 		return
 	}
 

cmd/http-stats.go

@@ -27,10 +27,6 @@ import (
 	"github.com/prometheus/client_golang/prometheus"
 )
 
-const (
-	apiGetObject = "GetObject"
-)
-
 // connStats - Network statistics
 // Count total input/output transferred bytes during
 // the server's life.
@@ -132,7 +128,7 @@ func (bh *bucketHTTPStats) updateHTTPStats(bucket, api string, w *xhttp.Response
 		return
 	}
 
-	if w != nil && api == apiGetObject {
+	if w != nil {
 		// Increment the prometheus http request response histogram with API, Bucket
 		bucketHTTPRequestsDuration.With(prometheus.Labels{
 			"api":    api,
@@ -437,9 +433,7 @@ func (st *HTTPStats) updateStats(api string, w *xhttp.ResponseRecorder) {
 	st.totalS3Requests.Inc(api)
 
 	// Increment the prometheus http request response histogram with appropriate label
-	if api == apiGetObject {
-		httpRequestsDuration.With(prometheus.Labels{"api": api}).Observe(w.TimeToFirstByte.Seconds())
-	}
+	httpRequestsDuration.With(prometheus.Labels{"api": api}).Observe(w.TimeToFirstByte.Seconds())
 
 	code := w.StatusCode
 

cmd/iam-etcd-store.go

@@ -249,6 +249,18 @@ func (ies *IAMEtcdStore) addUser(ctx context.Context, user string, userType IAMU
 	return nil
 }
 
+func (ies *IAMEtcdStore) loadSecretKey(ctx context.Context, user string, userType IAMUserType) (string, error) {
+	var u UserIdentity
+	err := ies.loadIAMConfig(ctx, &u, getUserIdentityPath(user, userType))
+	if err != nil {
+		if errors.Is(err, errConfigNotFound) {
+			return "", errNoSuchUser
+		}
+		return "", err
+	}
+	return u.Credentials.SecretKey, nil
+}
+
 func (ies *IAMEtcdStore) loadUser(ctx context.Context, user string, userType IAMUserType, m map[string]UserIdentity) error {
 	var u UserIdentity
 	err := ies.loadIAMConfig(ctx, &u, getUserIdentityPath(user, userType))

cmd/iam-object-store.go

@@ -34,6 +34,7 @@ import (
 	xioutil "github.com/minio/minio/internal/ioutil"
 	"github.com/minio/minio/internal/kms"
 	"github.com/minio/minio/internal/logger"
+	"github.com/minio/pkg/v3/sync/errgroup"
 	"github.com/puzpuzpuz/xsync/v3"
 )
 
@@ -182,19 +183,20 @@ func (iamOS *IAMObjectStore) loadPolicyDocWithRetry(ctx context.Context, policy
 	}
 }
 
-func (iamOS *IAMObjectStore) loadPolicyDoc(ctx context.Context, policy string, m map[string]PolicyDoc) error {
+func (iamOS *IAMObjectStore) loadPolicy(ctx context.Context, policy string) (PolicyDoc, error) {
+	var p PolicyDoc
+
 	data, objInfo, err := iamOS.loadIAMConfigBytesWithMetadata(ctx, getPolicyDocPath(policy))
 	if err != nil {
 		if err == errConfigNotFound {
-			return errNoSuchPolicy
+			return p, errNoSuchPolicy
 		}
-		return err
+		return p, err
 	}
 
-	var p PolicyDoc
 	err = p.parseJSON(data)
 	if err != nil {
-		return err
+		return p, err
 	}
 
 	if p.Version == 0 {
@@ -205,6 +207,14 @@ func (iamOS *IAMObjectStore) loadPolicyDoc(ctx context.Context, policy string, m
 		p.UpdateDate = objInfo.ModTime
 	}
 
+	return p, nil
+}
+
+func (iamOS *IAMObjectStore) loadPolicyDoc(ctx context.Context, policy string, m map[string]PolicyDoc) error {
+	p, err := iamOS.loadPolicy(ctx, policy)
+	if err != nil {
+		return err
+	}
 	m[policy] = p
 	return nil
 }
@@ -225,21 +235,33 @@ func (iamOS *IAMObjectStore) loadPolicyDocs(ctx context.Context, m map[string]Po
 	return nil
 }
 
-func (iamOS *IAMObjectStore) loadUser(ctx context.Context, user string, userType IAMUserType, m map[string]UserIdentity) error {
+func (iamOS *IAMObjectStore) loadSecretKey(ctx context.Context, user string, userType IAMUserType) (string, error) {
+	var u UserIdentity
+	err := iamOS.loadIAMConfig(ctx, &u, getUserIdentityPath(user, userType))
+	if err != nil {
+		if errors.Is(err, errConfigNotFound) {
+			return "", errNoSuchUser
+		}
+		return "", err
+	}
+	return u.Credentials.SecretKey, nil
+}
+
+func (iamOS *IAMObjectStore) loadUserIdentity(ctx context.Context, user string, userType IAMUserType) (UserIdentity, error) {
 	var u UserIdentity
 	err := iamOS.loadIAMConfig(ctx, &u, getUserIdentityPath(user, userType))
 	if err != nil {
 		if err == errConfigNotFound {
-			return errNoSuchUser
+			return u, errNoSuchUser
 		}
-		return err
+		return u, err
 	}
 
 	if u.Credentials.IsExpired() {
 		// Delete expired identity - ignoring errors here.
 		iamOS.deleteIAMConfig(ctx, getUserIdentityPath(user, userType))
 		iamOS.deleteIAMConfig(ctx, getMappedPolicyPath(user, userType, false))
-		return nil
+		return u, errNoSuchUser
 	}
 
 	if u.Credentials.AccessKey == "" {
@@ -255,7 +277,7 @@ func (iamOS *IAMObjectStore) loadUser(ctx context.Context, user string, userType
 				iamOS.deleteIAMConfig(ctx, getUserIdentityPath(user, userType))
 				iamOS.deleteIAMConfig(ctx, getMappedPolicyPath(user, userType, false))
 			}
-			return nil
+			return u, errNoSuchUser
 
 		}
 		u.Credentials.Claims = jwtClaims.Map()
@@ -265,6 +287,35 @@ func (iamOS *IAMObjectStore) loadUser(ctx context.Context, user string, userType
 		u.Credentials.Description = u.Credentials.Comment
 	}
 
+	return u, nil
+}
+
+func (iamOS *IAMObjectStore) loadUserConcurrent(ctx context.Context, userType IAMUserType, users ...string) ([]UserIdentity, error) {
+	userIdentities := make([]UserIdentity, len(users))
+	g := errgroup.WithNErrs(len(users))
+
+	for index := range users {
+		index := index
+		g.Go(func() error {
+			userName := path.Dir(users[index])
+			user, err := iamOS.loadUserIdentity(ctx, userName, userType)
+			if err != nil && !errors.Is(err, errNoSuchUser) {
+				return fmt.Errorf("unable to load the user `%s`: %w", userName, err)
+			}
+			userIdentities[index] = user
+			return nil
+		}, index)
+	}
+
+	err := errors.Join(g.Wait()...)
+	return userIdentities, err
+}
+
+func (iamOS *IAMObjectStore) loadUser(ctx context.Context, user string, userType IAMUserType, m map[string]UserIdentity) error {
+	u, err := iamOS.loadUserIdentity(ctx, user, userType)
+	if err != nil {
+		return err
+	}
 	m[user] = u
 	return nil
 }
@@ -346,16 +397,44 @@ func (iamOS *IAMObjectStore) loadMappedPolicyWithRetry(ctx context.Context, name
 	}
 }
 
-func (iamOS *IAMObjectStore) loadMappedPolicy(ctx context.Context, name string, userType IAMUserType, isGroup bool, m *xsync.MapOf[string, MappedPolicy]) error {
+func (iamOS *IAMObjectStore) loadMappedPolicyInternal(ctx context.Context, name string, userType IAMUserType, isGroup bool) (MappedPolicy, error) {
 	var p MappedPolicy
 	err := iamOS.loadIAMConfig(ctx, &p, getMappedPolicyPath(name, userType, isGroup))
 	if err != nil {
 		if err == errConfigNotFound {
-			return errNoSuchPolicy
+			return p, errNoSuchPolicy
 		}
-		return err
+		return p, err
+	}
+	return p, nil
+}
+
+func (iamOS *IAMObjectStore) loadMappedPolicyConcurrent(ctx context.Context, userType IAMUserType, isGroup bool, users ...string) ([]MappedPolicy, error) {
+	mappedPolicies := make([]MappedPolicy, len(users))
+	g := errgroup.WithNErrs(len(users))
+
+	for index := range users {
+		index := index
+		g.Go(func() error {
+			userName := strings.TrimSuffix(users[index], ".json")
+			userMP, err := iamOS.loadMappedPolicyInternal(ctx, userName, userType, isGroup)
+			if err != nil && !errors.Is(err, errNoSuchPolicy) {
+				return fmt.Errorf("unable to load the user policy map `%s`: %w", userName, err)
+			}
+			mappedPolicies[index] = userMP
+			return nil
+		}, index)
 	}
 
+	err := errors.Join(g.Wait()...)
+	return mappedPolicies, err
+}
+
+func (iamOS *IAMObjectStore) loadMappedPolicy(ctx context.Context, name string, userType IAMUserType, isGroup bool, m *xsync.MapOf[string, MappedPolicy]) error {
+	p, err := iamOS.loadMappedPolicyInternal(ctx, name, userType, isGroup)
+	if err != nil {
+		return err
+	}
 	m.Store(name, p)
 	return nil
 }
@@ -439,82 +518,230 @@ func (iamOS *IAMObjectStore) listAllIAMConfigItems(ctx context.Context) (res map
 	return res, nil
 }
 
+const (
+	maxIAMLoadOpTime = 5 * time.Second
+)
+
+func (iamOS *IAMObjectStore) loadPolicyDocConcurrent(ctx context.Context, policies ...string) ([]PolicyDoc, error) {
+	policyDocs := make([]PolicyDoc, len(policies))
+	g := errgroup.WithNErrs(len(policies))
+
+	for index := range policies {
+		index := index
+		g.Go(func() error {
+			policyName := path.Dir(policies[index])
+			policyDoc, err := iamOS.loadPolicy(ctx, policyName)
+			if err != nil && !errors.Is(err, errNoSuchPolicy) {
+				return fmt.Errorf("unable to load the policy doc `%s`: %w", policyName, err)
+			}
+			policyDocs[index] = policyDoc
+			return nil
+		}, index)
+	}
+
+	err := errors.Join(g.Wait()...)
+	return policyDocs, err
+}
+
 // Assumes cache is locked by caller.
-func (iamOS *IAMObjectStore) loadAllFromObjStore(ctx context.Context, cache *iamCache) error {
+func (iamOS *IAMObjectStore) loadAllFromObjStore(ctx context.Context, cache *iamCache, firstTime bool) error {
+	bootstrapTraceMsgFirstTime := func(s string) {
+		if firstTime {
+			bootstrapTraceMsg(s)
+		}
+	}
+
 	if iamOS.objAPI == nil {
 		return errServerNotInitialized
 	}
 
-	bootstrapTraceMsg("loading all IAM items")
+	bootstrapTraceMsgFirstTime("loading all IAM items")
 
+	setDefaultCannedPolicies(cache.iamPolicyDocsMap)
+
+	listStartTime := UTCNow()
 	listedConfigItems, err := iamOS.listAllIAMConfigItems(ctx)
 	if err != nil {
 		return fmt.Errorf("unable to list IAM data: %w", err)
 	}
+	if took := time.Since(listStartTime); took > maxIAMLoadOpTime {
+		var s strings.Builder
+		for k, v := range listedConfigItems {
+			s.WriteString(fmt.Sprintf("    %s: %d items\n", k, len(v)))
+		}
+		logger.Info("listAllIAMConfigItems took %.2fs with contents:\n%s", took.Seconds(), s.String())
+	}
 
 	// Loads things in the same order as `LoadIAMCache()`
 
-	bootstrapTraceMsg("loading policy documents")
+	bootstrapTraceMsgFirstTime("loading policy documents")
 
+	policyLoadStartTime := UTCNow()
 	policiesList := listedConfigItems[policiesListKey]
-	for _, item := range policiesList {
-		policyName := path.Dir(item)
-		if err := iamOS.loadPolicyDoc(ctx, policyName, cache.iamPolicyDocsMap); err != nil && !errors.Is(err, errNoSuchPolicy) {
-			return fmt.Errorf("unable to load the policy doc `%s`: %w", policyName, err)
+	count := 32 // number of parallel IAM loaders
+	for {
+		if len(policiesList) < count {
+			policyDocs, err := iamOS.loadPolicyDocConcurrent(ctx, policiesList...)
+			if err != nil {
+				return err
+			}
+			for index := range policiesList {
+				if policyDocs[index].Policy.Version != "" {
+					policyName := path.Dir(policiesList[index])
+					cache.iamPolicyDocsMap[policyName] = policyDocs[index]
+				}
+			}
+			break
 		}
-	}
-	setDefaultCannedPolicies(cache.iamPolicyDocsMap)
 
-	if iamOS.usersSysType == MinIOUsersSysType {
-		bootstrapTraceMsg("loading regular IAM users")
-		regUsersList := listedConfigItems[usersListKey]
-		for _, item := range regUsersList {
-			userName := path.Dir(item)
-			if err := iamOS.loadUser(ctx, userName, regUser, cache.iamUsersMap); err != nil && err != errNoSuchUser {
-				return fmt.Errorf("unable to load the user `%s`: %w", userName, err)
+		policyDocs, err := iamOS.loadPolicyDocConcurrent(ctx, policiesList[:count]...)
+		if err != nil {
+			return err
+		}
+
+		for index := range policiesList[:count] {
+			if policyDocs[index].Policy.Version != "" {
+				policyName := path.Dir(policiesList[index])
+				cache.iamPolicyDocsMap[policyName] = policyDocs[index]
 			}
 		}
 
-		bootstrapTraceMsg("loading regular IAM groups")
+		policiesList = policiesList[count:]
+	}
+
+	if took := time.Since(policyLoadStartTime); took > maxIAMLoadOpTime {
+		logger.Info("Policy docs load took %.2fs (for %d items)", took.Seconds(), len(policiesList))
+	}
+
+	if iamOS.usersSysType == MinIOUsersSysType {
+		bootstrapTraceMsgFirstTime("loading regular IAM users")
+		regUsersLoadStartTime := UTCNow()
+		regUsersList := listedConfigItems[usersListKey]
+
+		for {
+			if len(regUsersList) < count {
+				users, err := iamOS.loadUserConcurrent(ctx, regUser, regUsersList...)
+				if err != nil {
+					return err
+				}
+				for index := range regUsersList {
+					if users[index].Credentials.AccessKey != "" {
+						userName := path.Dir(regUsersList[index])
+						cache.iamUsersMap[userName] = users[index]
+					}
+				}
+				break
+			}
+
+			users, err := iamOS.loadUserConcurrent(ctx, regUser, regUsersList[:count]...)
+			if err != nil {
+				return err
+			}
+
+			for index := range regUsersList[:count] {
+				if users[index].Credentials.AccessKey != "" {
+					userName := path.Dir(regUsersList[index])
+					cache.iamUsersMap[userName] = users[index]
+				}
+			}
+
+			regUsersList = regUsersList[count:]
+		}
+
+		if took := time.Since(regUsersLoadStartTime); took > maxIAMLoadOpTime {
+			actualLoaded := len(cache.iamUsersMap)
+			logger.Info("Reg. users load took %.2fs (for %d items with %d expired items)", took.Seconds(),
+				len(regUsersList), len(regUsersList)-actualLoaded)
+		}
+
+		bootstrapTraceMsgFirstTime("loading regular IAM groups")
+		groupsLoadStartTime := UTCNow()
 		groupsList := listedConfigItems[groupsListKey]
 		for _, item := range groupsList {
 			group := path.Dir(item)
 			if err := iamOS.loadGroup(ctx, group, cache.iamGroupsMap); err != nil && err != errNoSuchGroup {
-				return fmt.Errorf("unable to load the group `%s`: %w", group, err)
+				return fmt.Errorf("unable to load the group: %w", err)
 			}
 		}
-	}
-
-	bootstrapTraceMsg("loading user policy mapping")
-	userPolicyMappingsList := listedConfigItems[policyDBUsersListKey]
-	for _, item := range userPolicyMappingsList {
-		userName := strings.TrimSuffix(item, ".json")
-		if err := iamOS.loadMappedPolicy(ctx, userName, regUser, false, cache.iamUserPolicyMap); err != nil && !errors.Is(err, errNoSuchPolicy) {
-			return fmt.Errorf("unable to load the policy mapping for the user `%s`: %w", userName, err)
+		if took := time.Since(groupsLoadStartTime); took > maxIAMLoadOpTime {
+			logger.Info("Groups load took %.2fs (for %d items)", took.Seconds(), len(groupsList))
 		}
 	}
 
-	bootstrapTraceMsg("loading group policy mapping")
+	bootstrapTraceMsgFirstTime("loading user policy mapping")
+	userPolicyMappingLoadStartTime := UTCNow()
+	userPolicyMappingsList := listedConfigItems[policyDBUsersListKey]
+	for {
+		if len(userPolicyMappingsList) < count {
+			mappedPolicies, err := iamOS.loadMappedPolicyConcurrent(ctx, regUser, false, userPolicyMappingsList...)
+			if err != nil {
+				return err
+			}
+
+			for index := range userPolicyMappingsList {
+				if mappedPolicies[index].Policies != "" {
+					userName := strings.TrimSuffix(userPolicyMappingsList[index], ".json")
+					cache.iamUserPolicyMap.Store(userName, mappedPolicies[index])
+				}
+			}
+
+			break
+		}
+
+		mappedPolicies, err := iamOS.loadMappedPolicyConcurrent(ctx, regUser, false, userPolicyMappingsList[:count]...)
+		if err != nil {
+			return err
+		}
+
+		for index := range userPolicyMappingsList[:count] {
+			if mappedPolicies[index].Policies != "" {
+				userName := strings.TrimSuffix(userPolicyMappingsList[index], ".json")
+				cache.iamUserPolicyMap.Store(userName, mappedPolicies[index])
+			}
+		}
+
+		userPolicyMappingsList = userPolicyMappingsList[count:]
+	}
+
+	if took := time.Since(userPolicyMappingLoadStartTime); took > maxIAMLoadOpTime {
+		logger.Info("User policy mappings load took %.2fs (for %d items)", took.Seconds(), len(userPolicyMappingsList))
+	}
+
+	bootstrapTraceMsgFirstTime("loading group policy mapping")
+	groupPolicyMappingLoadStartTime := UTCNow()
 	groupPolicyMappingsList := listedConfigItems[policyDBGroupsListKey]
 	for _, item := range groupPolicyMappingsList {
 		groupName := strings.TrimSuffix(item, ".json")
 		if err := iamOS.loadMappedPolicy(ctx, groupName, regUser, true, cache.iamGroupPolicyMap); err != nil && !errors.Is(err, errNoSuchPolicy) {
-			return fmt.Errorf("unable to load the policy mapping for the group `%s`: %w", groupName, err)
+			return fmt.Errorf("unable to load the policy mapping for the group: %w", err)
 		}
 	}
+	if took := time.Since(groupPolicyMappingLoadStartTime); took > maxIAMLoadOpTime {
+		logger.Info("Group policy mappings load took %.2fs (for %d items)", took.Seconds(), len(groupPolicyMappingsList))
+	}
 
-	bootstrapTraceMsg("loading service accounts")
+	bootstrapTraceMsgFirstTime("loading service accounts")
+	svcAccLoadStartTime := UTCNow()
 	svcAccList := listedConfigItems[svcAccListKey]
 	svcUsersMap := make(map[string]UserIdentity, len(svcAccList))
 	for _, item := range svcAccList {
 		userName := path.Dir(item)
 		if err := iamOS.loadUser(ctx, userName, svcUser, svcUsersMap); err != nil && err != errNoSuchUser {
-			return fmt.Errorf("unable to load the service account `%s`: %w", userName, err)
+			return fmt.Errorf("unable to load the service account: %w", err)
 		}
 	}
+	if took := time.Since(svcAccLoadStartTime); took > maxIAMLoadOpTime {
+		logger.Info("Service accounts load took %.2fs (for %d items with %d expired items)", took.Seconds(),
+			len(svcAccList), len(svcAccList)-len(svcUsersMap))
+	}
+
+	bootstrapTraceMsg("loading STS account policy mapping")
+	stsPolicyMappingLoadStartTime := UTCNow()
+	var stsPolicyMappingsCount int
 	for _, svcAcc := range svcUsersMap {
 		svcParent := svcAcc.Credentials.ParentUser
 		if _, ok := cache.iamUsersMap[svcParent]; !ok {
+			stsPolicyMappingsCount++
 			// If a service account's parent user is not in iamUsersMap, the
 			// parent is an STS account. Such accounts may have a policy mapped
 			// on the parent user, so we load them. This is not needed for the
@@ -529,10 +756,14 @@ func (iamOS *IAMObjectStore) loadAllFromObjStore(ctx context.Context, cache *iam
 			// OIDC/AssumeRoleWithCustomToken/AssumeRoleWithCertificate).
 			err := iamOS.loadMappedPolicy(ctx, svcParent, stsUser, false, cache.iamSTSPolicyMap)
 			if err != nil && !errors.Is(err, errNoSuchPolicy) {
-				return fmt.Errorf("unable to load the policy mapping for the STS user `%s`: %w", svcParent, err)
+				return fmt.Errorf("unable to load the policy mapping for the STS user: %w", err)
 			}
 		}
 	}
+	if took := time.Since(stsPolicyMappingLoadStartTime); took > maxIAMLoadOpTime {
+		logger.Info("STS policy mappings load took %.2fs (for %d items)", took.Seconds(), stsPolicyMappingsCount)
+	}
+
 	// Copy svcUsersMap to cache.iamUsersMap
 	for k, v := range svcUsersMap {
 		cache.iamUsersMap[k] = v
@@ -550,21 +781,16 @@ func (iamOS *IAMObjectStore) loadAllFromObjStore(ctx context.Context, cache *iam
 	for _, item := range listedConfigItems[stsListKey] {
 		userName := path.Dir(item)
 		// loadUser() will delete expired user during the load.
-		err := iamOS.loadUser(ctx, userName, stsUser, stsAccountsFromStore)
-		if err != nil && !errors.Is(err, errNoSuchUser) {
-			return fmt.Errorf("unable to load user during STS purge: %w (%s)", err, item)
-		}
-
+		iamLogIf(ctx, iamOS.loadUser(ctx, userName, stsUser, stsAccountsFromStore))
+		// No need to return errors for failed expiration of STS users
 	}
+
 	// Loading the STS policy mappings from disk ensures that stale entries
 	// (removed during loadUser() in the loop above) are removed from memory.
 	for _, item := range listedConfigItems[policyDBSTSUsersListKey] {
 		stsName := strings.TrimSuffix(item, ".json")
-		err := iamOS.loadMappedPolicy(ctx, stsName, stsUser, false, stsAccPoliciesFromStore)
-		if err != nil && !errors.Is(err, errNoSuchPolicy) {
-			return fmt.Errorf("unable to load policies during STS purge: %w (%s)", err, item)
-		}
-
+		iamLogIf(ctx, iamOS.loadMappedPolicy(ctx, stsName, stsUser, false, stsAccPoliciesFromStore))
+		// No need to return errors for failed expiration of STS users
 	}
 
 	took := time.Since(purgeStart).Seconds()
@@ -576,7 +802,11 @@ func (iamOS *IAMObjectStore) loadAllFromObjStore(ctx context.Context, cache *iam
 	// Store the newly populated map in the iam cache. This takes care of
 	// removing stale entries from the existing map.
 	cache.iamSTSAccountsMap = stsAccountsFromStore
-	cache.iamSTSPolicyMap = stsAccPoliciesFromStore
+
+	stsAccPoliciesFromStore.Range(func(k string, v MappedPolicy) bool {
+		cache.iamSTSPolicyMap.Store(k, v)
+		return true
+	})
 
 	return nil
 }

cmd/iam.go

@@ -46,9 +46,9 @@ import (
 	"github.com/minio/minio/internal/config/policy/opa"
 	polplugin "github.com/minio/minio/internal/config/policy/plugin"
 	xhttp "github.com/minio/minio/internal/http"
-	xioutil "github.com/minio/minio/internal/ioutil"
 	"github.com/minio/minio/internal/jwt"
 	"github.com/minio/minio/internal/logger"
+	"github.com/minio/pkg/v3/env"
 	"github.com/minio/pkg/v3/ldap"
 	"github.com/minio/pkg/v3/policy"
 	etcd "go.etcd.io/etcd/client/v3"
@@ -178,9 +178,16 @@ func (sys *IAMSys) initStore(objAPI ObjectLayer, etcdClient *etcd.Client) {
 	}
 
 	if etcdClient == nil {
-		sys.store = &IAMStoreSys{newIAMObjectStore(objAPI, sys.usersSysType), &singleflight.Group{}}
+		var group *singleflight.Group
+		if env.Get("_MINIO_IAM_SINGLE_FLIGHT", config.EnableOn) == config.EnableOn {
+			group = &singleflight.Group{}
+		}
+		sys.store = &IAMStoreSys{
+			IAMStorageAPI: newIAMObjectStore(objAPI, sys.usersSysType),
+			group:         group,
+		}
 	} else {
-		sys.store = &IAMStoreSys{newIAMEtcdStore(etcdClient, sys.usersSysType), &singleflight.Group{}}
+		sys.store = &IAMStoreSys{IAMStorageAPI: newIAMEtcdStore(etcdClient, sys.usersSysType)}
 	}
 }
 
@@ -211,18 +218,21 @@ func (sys *IAMSys) Load(ctx context.Context, firstTime bool) error {
 	if !globalSiteReplicatorCred.IsValid() {
 		sa, _, err := sys.getServiceAccount(ctx, siteReplicatorSvcAcc)
 		if err == nil {
-			globalSiteReplicatorCred.Set(sa.Credentials)
+			globalSiteReplicatorCred.Set(sa.Credentials.SecretKey)
 		}
 	}
 
 	if firstTime {
 		bootstrapTraceMsg(fmt.Sprintf("globalIAMSys.Load(): (duration: %s)", loadDuration))
+		if globalIsDistErasure {
+			logger.Info("IAM load(startup) finished. (duration: %s)", loadDuration)
+		}
 	}
 
 	select {
 	case <-sys.configLoaded:
 	default:
-		xioutil.SafeClose(sys.configLoaded)
+		close(sys.configLoaded)
 	}
 	return nil
 }
@@ -304,8 +314,9 @@ func (sys *IAMSys) Init(ctx context.Context, objAPI ObjectLayer, etcdClient *etc
 		// Migrate IAM configuration, if necessary.
 		if err := saveIAMFormat(retryCtx, sys.store); err != nil {
 			if configRetriableErrors(err) {
-				logger.Info("Waiting for all MinIO IAM sub-system to be initialized.. possible cause (%v)", err)
-				time.Sleep(time.Duration(r.Float64() * float64(time.Second)))
+				retryInterval := time.Duration(r.Float64() * float64(time.Second))
+				logger.Info("Waiting for all MinIO IAM sub-system to be initialized.. possible cause (%v) (retrying in %s)", err, retryInterval)
+				time.Sleep(retryInterval)
 				continue
 			}
 			iamLogIf(ctx, fmt.Errorf("IAM sub-system is partially initialized, unable to write the IAM format: %w", err), logger.WarningKind)
@@ -315,24 +326,9 @@ func (sys *IAMSys) Init(ctx context.Context, objAPI ObjectLayer, etcdClient *etc
 		break
 	}
 
-	// Load IAM data from storage.
-	for {
-		if err := sys.Load(retryCtx, true); err != nil {
-			if configRetriableErrors(err) {
-				logger.Info("Waiting for all MinIO IAM sub-system to be initialized.. possible cause (%v)", err)
-				time.Sleep(time.Duration(r.Float64() * float64(time.Second)))
-				continue
-			}
-			if err != nil {
-				iamLogIf(ctx, fmt.Errorf("Unable to initialize IAM sub-system, some users may not be available: %w", err), logger.WarningKind)
-			}
-		}
-		break
-	}
-
-	refreshInterval := sys.iamRefreshInterval
-
-	go sys.periodicRoutines(ctx, refreshInterval)
+	cache := sys.store.lock()
+	setDefaultCannedPolicies(cache.iamPolicyDocsMap)
+	sys.store.unlock()
 
 	// Load RoleARNs
 	sys.rolesMap = make(map[arn.ARN]string)
@@ -348,6 +344,26 @@ func (sys *IAMSys) Init(ctx context.Context, objAPI ObjectLayer, etcdClient *etc
 		sys.validateAndAddRolePolicyMappings(ctx, riMap)
 	}
 
+	// Load IAM data from storage.
+	for {
+		if err := sys.Load(retryCtx, true); err != nil {
+			if configRetriableErrors(err) {
+				retryInterval := time.Duration(r.Float64() * float64(time.Second))
+				logger.Info("Waiting for all MinIO IAM sub-system to be initialized.. possible cause (%v) (retrying in %s)", err, retryInterval)
+				time.Sleep(retryInterval)
+				continue
+			}
+			if err != nil {
+				iamLogIf(ctx, fmt.Errorf("Unable to initialize IAM sub-system, some users may not be available: %w", err), logger.WarningKind)
+			}
+		}
+		break
+	}
+
+	refreshInterval := sys.iamRefreshInterval
+
+	go sys.periodicRoutines(ctx, refreshInterval)
+
 	sys.printIAMRoles()
 
 	bootstrapTraceMsg("finishing IAM loading")
@@ -396,12 +412,12 @@ func (sys *IAMSys) periodicRoutines(ctx context.Context, baseInterval time.Durat
 			// Load all IAM items (except STS creds) periodically.
 			refreshStart := time.Now()
 			if err := sys.Load(ctx, false); err != nil {
-				iamLogIf(ctx, fmt.Errorf("Failure in periodic refresh for IAM (took %.2fs): %v", time.Since(refreshStart).Seconds(), err), logger.WarningKind)
+				iamLogIf(ctx, fmt.Errorf("Failure in periodic refresh for IAM (duration: %s): %v", time.Since(refreshStart), err), logger.WarningKind)
 			} else {
 				took := time.Since(refreshStart).Seconds()
 				if took > maxDurationSecondsForLog {
 					// Log if we took a lot of time to load.
-					logger.Info("IAM refresh took %.2fs", took)
+					logger.Info("IAM refresh took (duration: %.2fs)", took)
 				}
 			}
 
@@ -436,7 +452,7 @@ func (sys *IAMSys) validateAndAddRolePolicyMappings(ctx context.Context, m map[a
 	// running server by creating the policies after start up.
 	for arn, rolePolicies := range m {
 		specifiedPoliciesSet := newMappedPolicy(rolePolicies).policySet()
-		validPolicies, _ := sys.store.FilterPolicies(rolePolicies, "")
+		validPolicies, _ := sys.store.MergePolicies(rolePolicies)
 		knownPoliciesSet := newMappedPolicy(validPolicies).policySet()
 		unknownPoliciesSet := specifiedPoliciesSet.Difference(knownPoliciesSet)
 		if len(unknownPoliciesSet) > 0 {
@@ -672,7 +688,7 @@ func (sys *IAMSys) CurrentPolicies(policyName string) string {
 		return ""
 	}
 
-	policies, _ := sys.store.FilterPolicies(policyName, "")
+	policies, _ := sys.store.MergePolicies(policyName)
 	return policies
 }
 
@@ -786,11 +802,15 @@ func (sys *IAMSys) ListLDAPUsers(ctx context.Context) (map[string]madmin.UserInf
 
 	select {
 	case <-sys.configLoaded:
-		ldapUsers := make(map[string]madmin.UserInfo)
-		for user, policy := range sys.store.GetUsersWithMappedPolicies() {
+		stsMap, err := sys.store.GetAllSTSUserMappings(sys.LDAPConfig.IsLDAPUserDN)
+		if err != nil {
+			return nil, err
+		}
+		ldapUsers := make(map[string]madmin.UserInfo, len(stsMap))
+		for user, policy := range stsMap {
 			ldapUsers[user] = madmin.UserInfo{
 				PolicyName: policy,
-				Status:     madmin.AccountEnabled,
+				Status:     statusEnabled,
 			}
 		}
 		return ldapUsers, nil
@@ -799,6 +819,57 @@ func (sys *IAMSys) ListLDAPUsers(ctx context.Context) (map[string]madmin.UserInf
 	}
 }
 
+type cleanEntitiesQuery struct {
+	Users    map[string]set.StringSet
+	Groups   set.StringSet
+	Policies set.StringSet
+}
+
+// createCleanEntitiesQuery - maps users to their groups and normalizes user or group DNs if ldap.
+func (sys *IAMSys) createCleanEntitiesQuery(q madmin.PolicyEntitiesQuery, ldap bool) cleanEntitiesQuery {
+	cleanQ := cleanEntitiesQuery{
+		Users:    make(map[string]set.StringSet),
+		Groups:   set.CreateStringSet(q.Groups...),
+		Policies: set.CreateStringSet(q.Policy...),
+	}
+
+	if ldap {
+		// Validate and normalize users, then fetch and normalize their groups
+		// Also include unvalidated users for backward compatibility.
+		for _, user := range q.Users {
+			lookupRes, actualGroups, _ := sys.LDAPConfig.GetValidatedDNWithGroups(user)
+			if lookupRes != nil {
+				groupSet := set.CreateStringSet(actualGroups...)
+
+				// duplicates can be overwritten, fetched groups should be identical.
+				cleanQ.Users[lookupRes.NormDN] = groupSet
+			}
+			// Search for non-normalized DN as well for backward compatibility.
+			if _, ok := cleanQ.Users[user]; !ok {
+				cleanQ.Users[user] = nil
+			}
+		}
+
+		// Validate and normalize groups.
+		for _, group := range q.Groups {
+			lookupRes, underDN, _ := sys.LDAPConfig.GetValidatedGroupDN(nil, group)
+			if lookupRes != nil && underDN {
+				cleanQ.Groups.Add(lookupRes.NormDN)
+			}
+		}
+	} else {
+		for _, user := range q.Users {
+			info, err := sys.store.GetUserInfo(user)
+			var groupSet set.StringSet
+			if err == nil {
+				groupSet = set.CreateStringSet(info.MemberOf...)
+			}
+			cleanQ.Users[user] = groupSet
+		}
+	}
+	return cleanQ
+}
+
 // QueryLDAPPolicyEntities - queries policy associations for LDAP users/groups/policies.
 func (sys *IAMSys) QueryLDAPPolicyEntities(ctx context.Context, q madmin.PolicyEntitiesQuery) (*madmin.PolicyEntitiesResult, error) {
 	if !sys.Initialized() {
@@ -811,7 +882,8 @@ func (sys *IAMSys) QueryLDAPPolicyEntities(ctx context.Context, q madmin.PolicyE
 
 	select {
 	case <-sys.configLoaded:
-		pe := sys.store.ListPolicyMappings(q, sys.LDAPConfig.IsLDAPUserDN, sys.LDAPConfig.IsLDAPGroupDN)
+		cleanQuery := sys.createCleanEntitiesQuery(q, true)
+		pe := sys.store.ListPolicyMappings(cleanQuery, sys.LDAPConfig.IsLDAPUserDN, sys.LDAPConfig.IsLDAPGroupDN, sys.LDAPConfig.DecodeDN)
 		pe.Timestamp = UTCNow()
 		return &pe, nil
 	case <-ctx.Done():
@@ -885,6 +957,7 @@ func (sys *IAMSys) QueryPolicyEntities(ctx context.Context, q madmin.PolicyEntit
 
 	select {
 	case <-sys.configLoaded:
+		cleanQuery := sys.createCleanEntitiesQuery(q, false)
 		var userPredicate, groupPredicate func(string) bool
 		if sys.LDAPConfig.Enabled() {
 			userPredicate = func(s string) bool {
@@ -894,7 +967,7 @@ func (sys *IAMSys) QueryPolicyEntities(ctx context.Context, q madmin.PolicyEntit
 				return !sys.LDAPConfig.IsLDAPGroupDN(s)
 			}
 		}
-		pe := sys.store.ListPolicyMappings(q, userPredicate, groupPredicate)
+		pe := sys.store.ListPolicyMappings(cleanQuery, userPredicate, groupPredicate, nil)
 		pe.Timestamp = UTCNow()
 		return &pe, nil
 	case <-ctx.Done():
@@ -1148,9 +1221,18 @@ func (sys *IAMSys) getServiceAccount(ctx context.Context, accessKey string) (Use
 
 // GetTemporaryAccount - wrapper method to get information about a temporary account
 func (sys *IAMSys) GetTemporaryAccount(ctx context.Context, accessKey string) (auth.Credentials, *policy.Policy, error) {
+	if !sys.Initialized() {
+		return auth.Credentials{}, nil, errServerNotInitialized
+	}
 	tmpAcc, embeddedPolicy, err := sys.getTempAccount(ctx, accessKey)
 	if err != nil {
-		return auth.Credentials{}, nil, err
+		if err == errNoSuchTempAccount {
+			sys.store.LoadUser(ctx, accessKey)
+			tmpAcc, embeddedPolicy, err = sys.getTempAccount(ctx, accessKey)
+		}
+		if err != nil {
+			return auth.Credentials{}, nil, err
+		}
 	}
 	// Hide secret & session keys
 	tmpAcc.Credentials.SecretKey = ""
@@ -1378,11 +1460,13 @@ func (sys *IAMSys) updateGroupMembershipsForLDAP(ctx context.Context) {
 	// 1. Collect all LDAP users with active creds.
 	allCreds := sys.store.GetSTSAndServiceAccounts()
 	// List of unique LDAP (parent) user DNs that have active creds
-	var parentUsers []string
-	// Map of LDAP user to list of active credential objects
+	var parentUserActualDNList []string
+	// Map of LDAP user (internal representation) to list of active credential objects
 	parentUserToCredsMap := make(map[string][]auth.Credentials)
 	// DN to ldap username mapping for each LDAP user
-	parentUserToLDAPUsernameMap := make(map[string]string)
+	actualDNToLDAPUsernameMap := make(map[string]string)
+	// External (actual) LDAP DN to internal normalized representation
+	actualDNToParentUserMap := make(map[string]string)
 	for _, cred := range allCreds {
 		// Expired credentials don't need parent user updates.
 		if cred.IsExpired() {
@@ -1425,25 +1509,28 @@ func (sys *IAMSys) updateGroupMembershipsForLDAP(ctx context.Context) {
 				continue
 			}
 
-			ldapUsername, ok := jwtClaims.Lookup(ldapUserN)
-			if !ok {
+			ldapUsername, okUserN := jwtClaims.Lookup(ldapUserN)
+			ldapActualDN, okDN := jwtClaims.Lookup(ldapActualUser)
+			if !okUserN || !okDN {
 				// skip this cred - we dont have the
 				// username info needed
 				continue
 			}
 
 			// Collect each new cred.ParentUser into parentUsers
-			parentUsers = append(parentUsers, cred.ParentUser)
+			parentUserActualDNList = append(parentUserActualDNList, ldapActualDN)
 
 			// Update the ldapUsernameMap
-			parentUserToLDAPUsernameMap[cred.ParentUser] = ldapUsername
+			actualDNToLDAPUsernameMap[ldapActualDN] = ldapUsername
+
+			// Update the actualDNToParentUserMap
+			actualDNToParentUserMap[ldapActualDN] = cred.ParentUser
 		}
 		parentUserToCredsMap[cred.ParentUser] = append(parentUserToCredsMap[cred.ParentUser], cred)
-
 	}
 
 	// 2. Query LDAP server for groups of the LDAP users collected.
-	updatedGroups, err := sys.LDAPConfig.LookupGroupMemberships(parentUsers, parentUserToLDAPUsernameMap)
+	updatedGroups, err := sys.LDAPConfig.LookupGroupMemberships(parentUserActualDNList, actualDNToLDAPUsernameMap)
 	if err != nil {
 		// Log and return on error - perhaps it'll work the next time.
 		iamLogIf(GlobalContext, err)
@@ -1451,8 +1538,9 @@ func (sys *IAMSys) updateGroupMembershipsForLDAP(ctx context.Context) {
 	}
 
 	// 3. Update creds for those users whose groups are changed
-	for _, parentUser := range parentUsers {
-		currGroupsSet := updatedGroups[parentUser]
+	for _, parentActualDN := range parentUserActualDNList {
+		currGroupsSet := updatedGroups[parentActualDN]
+		parentUser := actualDNToParentUserMap[parentActualDN]
 		currGroups := currGroupsSet.ToSlice()
 		for _, cred := range parentUserToCredsMap[parentUser] {
 			gSet := set.CreateStringSet(cred.Groups...)
@@ -1510,11 +1598,11 @@ func (sys *IAMSys) NormalizeLDAPAccessKeypairs(ctx context.Context, accessKeyMap
 		// server and is under a configured base DN.
 		validatedParent, isUnderBaseDN, err := sys.LDAPConfig.GetValidatedUserDN(conn, parent)
 		if err != nil {
-			collectedErrors = append(collectedErrors, fmt.Errorf("could not validate `%s` exists in LDAP directory: %w", parent, err))
+			collectedErrors = append(collectedErrors, fmt.Errorf("could not validate parent exists in LDAP directory: %w", err))
 			continue
 		}
 		if validatedParent == nil || !isUnderBaseDN {
-			err := fmt.Errorf("DN `%s` was not found in the LDAP directory", parent)
+			err := fmt.Errorf("DN parent was not found in the LDAP directory")
 			collectedErrors = append(collectedErrors, err)
 			continue
 		}
@@ -1529,11 +1617,11 @@ func (sys *IAMSys) NormalizeLDAPAccessKeypairs(ctx context.Context, accessKeyMap
 			// configured base DN.
 			validatedGroup, _, err := sys.LDAPConfig.GetValidatedGroupDN(conn, group)
 			if err != nil {
-				collectedErrors = append(collectedErrors, fmt.Errorf("could not validate `%s` exists in LDAP directory: %w", group, err))
+				collectedErrors = append(collectedErrors, fmt.Errorf("could not validate group exists in LDAP directory: %w", err))
 				continue
 			}
 			if validatedGroup == nil {
-				err := fmt.Errorf("DN `%s` was not found in the LDAP directory", group)
+				err := fmt.Errorf("DN group was not found in the LDAP directory")
 				collectedErrors = append(collectedErrors, err)
 				continue
 			}
@@ -1623,7 +1711,7 @@ func (sys *IAMSys) NormalizeLDAPMappingImport(ctx context.Context, isGroup bool,
 			continue
 		}
 		if validatedDN == nil || !underBaseDN {
-			err := fmt.Errorf("DN `%s` was not found in the LDAP directory", k)
+			err := fmt.Errorf("DN was not found in the LDAP directory")
 			collectedErrors = append(collectedErrors, err)
 			continue
 		}
@@ -1966,7 +2054,7 @@ func (sys *IAMSys) PolicyDBUpdateLDAP(ctx context.Context, isAttach bool,
 		if dnResult == nil {
 			// dn not found - still attempt to detach if provided user is a DN.
 			if !isAttach && sys.LDAPConfig.IsLDAPUserDN(r.User) {
-				dn = r.User
+				dn = sys.LDAPConfig.QuickNormalizeDN(r.User)
 			} else {
 				err = errNoSuchUser
 				return
@@ -1983,7 +2071,7 @@ func (sys *IAMSys) PolicyDBUpdateLDAP(ctx context.Context, isAttach bool,
 		}
 		if dnResult == nil || !underBaseDN {
 			if !isAttach {
-				dn = r.Group
+				dn = sys.LDAPConfig.QuickNormalizeDN(r.Group)
 			} else {
 				err = errNoSuchGroup
 				return
@@ -2118,7 +2206,7 @@ func (sys *IAMSys) IsAllowedServiceAccount(args policy.Args, parentUser string)
 	var combinedPolicy policy.Policy
 	// Policies were found, evaluate all of them.
 	if !isOwnerDerived {
-		availablePoliciesStr, c := sys.store.FilterPolicies(strings.Join(svcPolicies, ","), "")
+		availablePoliciesStr, c := sys.store.MergePolicies(strings.Join(svcPolicies, ","))
 		if availablePoliciesStr == "" {
 			return false
 		}
@@ -2210,22 +2298,16 @@ func (sys *IAMSys) IsAllowedSTS(args policy.Args, parentUser string) bool {
 	// 2. Combine the mapped policies into a single combined policy.
 
 	var combinedPolicy policy.Policy
+	// Policies were found, evaluate all of them.
 	if !isOwnerDerived {
-		var err error
-		combinedPolicy, err = sys.store.GetPolicy(strings.Join(policies, ","))
-		if errors.Is(err, errNoSuchPolicy) {
-			for _, pname := range policies {
-				_, err := sys.store.GetPolicy(pname)
-				if errors.Is(err, errNoSuchPolicy) {
-					// all policies presented in the claim should exist
-					iamLogIf(GlobalContext, fmt.Errorf("expected policy (%s) missing from the JWT claim %s, rejecting the request", pname, iamPolicyClaimNameOpenID()))
-					return false
-				}
-			}
-			iamLogIf(GlobalContext, fmt.Errorf("all policies were unexpectedly present!"))
+		availablePoliciesStr, c := sys.store.MergePolicies(strings.Join(policies, ","))
+		if availablePoliciesStr == "" {
+			// all policies presented in the claim should exist
+			iamLogIf(GlobalContext, fmt.Errorf("expected policy (%s) missing from the JWT claim %s, rejecting the request", policies, iamPolicyClaimNameOpenID()))
+
 			return false
 		}
-
+		combinedPolicy = c
 	}
 
 	// 3. If an inline session-policy is present, evaluate it.
@@ -2346,7 +2428,7 @@ func isAllowedBySessionPolicy(args policy.Args) (hasSessionPolicy bool, isAllowe
 
 // GetCombinedPolicy returns a combined policy combining all policies
 func (sys *IAMSys) GetCombinedPolicy(policies ...string) policy.Policy {
-	_, policy := sys.store.FilterPolicies(strings.Join(policies, ","), "")
+	_, policy := sys.store.MergePolicies(strings.Join(policies, ","))
 	return policy
 }
 

cmd/jwt.go

@@ -24,10 +24,8 @@ import (
 
 	jwtgo "github.com/golang-jwt/jwt/v4"
 	jwtreq "github.com/golang-jwt/jwt/v4/request"
-	"github.com/hashicorp/golang-lru/v2/expirable"
 	"github.com/minio/minio/internal/auth"
 	xjwt "github.com/minio/minio/internal/jwt"
-	"github.com/minio/minio/internal/logger"
 	"github.com/minio/pkg/v3/policy"
 )
 
@@ -37,8 +35,8 @@ const (
 	// Default JWT token for web handlers is one day.
 	defaultJWTExpiry = 24 * time.Hour
 
-	// Inter-node JWT token expiry is 15 minutes.
-	defaultInterNodeJWTExpiry = 15 * time.Minute
+	// Inter-node JWT token expiry is 100 years approx.
+	defaultInterNodeJWTExpiry = 100 * 365 * 24 * time.Hour
 )
 
 var (
@@ -50,17 +48,10 @@ var (
 	errMalformedAuth      = errors.New("Malformed authentication input")
 )
 
-type cacheKey struct {
-	accessKey, secretKey, audience string
-}
-
-var cacheLRU = expirable.NewLRU[cacheKey, string](1000, nil, 15*time.Second)
-
-func authenticateNode(accessKey, secretKey, audience string) (string, error) {
+func authenticateNode(accessKey, secretKey string) (string, error) {
 	claims := xjwt.NewStandardClaims()
 	claims.SetExpiry(UTCNow().Add(defaultInterNodeJWTExpiry))
 	claims.SetAccessKey(accessKey)
-	claims.SetAudience(audience)
 
 	jwt := jwtgo.NewWithClaims(jwtgo.SigningMethodHS512, claims)
 	return jwt.SignedString([]byte(secretKey))
@@ -141,27 +132,9 @@ func metricsRequestAuthenticate(req *http.Request) (*xjwt.MapClaims, []string, b
 	return claims, groups, owner, nil
 }
 
-// newCachedAuthToken returns a token that is cached up to 15 seconds.
-// If globalActiveCred is updated it is reflected at once.
-func newCachedAuthToken() func(audience string) string {
-	fn := func(accessKey, secretKey, audience string) (s string, err error) {
-		k := cacheKey{accessKey: accessKey, secretKey: secretKey, audience: audience}
-
-		var ok bool
-		s, ok = cacheLRU.Get(k)
-		if !ok {
-			s, err = authenticateNode(accessKey, secretKey, audience)
-			if err != nil {
-				return "", err
-			}
-			cacheLRU.Add(k, s)
-		}
-		return s, nil
-	}
-	return func(audience string) string {
-		cred := globalActiveCred
-		token, err := fn(cred.AccessKey, cred.SecretKey, audience)
-		logger.CriticalIf(GlobalContext, err)
-		return token
+// newCachedAuthToken returns the cached token.
+func newCachedAuthToken() func() string {
+	return func() string {
+		return globalNodeAuthToken
 	}
 }

cmd/jwt_test.go

@@ -107,7 +107,7 @@ func BenchmarkParseJWTStandardClaims(b *testing.B) {
 	}
 
 	creds := globalActiveCred
-	token, err := authenticateNode(creds.AccessKey, creds.SecretKey, "")
+	token, err := authenticateNode(creds.AccessKey, creds.SecretKey)
 	if err != nil {
 		b.Fatal(err)
 	}
@@ -138,7 +138,7 @@ func BenchmarkParseJWTMapClaims(b *testing.B) {
 	}
 
 	creds := globalActiveCred
-	token, err := authenticateNode(creds.AccessKey, creds.SecretKey, "")
+	token, err := authenticateNode(creds.AccessKey, creds.SecretKey)
 	if err != nil {
 		b.Fatal(err)
 	}
@@ -176,7 +176,7 @@ func BenchmarkAuthenticateNode(b *testing.B) {
 		b.ResetTimer()
 		b.ReportAllocs()
 		for i := 0; i < b.N; i++ {
-			fn(creds.AccessKey, creds.SecretKey, "aud")
+			fn(creds.AccessKey, creds.SecretKey)
 		}
 	})
 	b.Run("cached", func(b *testing.B) {
@@ -184,7 +184,7 @@ func BenchmarkAuthenticateNode(b *testing.B) {
 		b.ResetTimer()
 		b.ReportAllocs()
 		for i := 0; i < b.N; i++ {
-			fn("aud")
+			fn()
 		}
 	})
 }

cmd/kms-handlers.go

@@ -22,8 +22,8 @@ import (
 	"encoding/json"
 	"net/http"
 
-	"github.com/minio/kms-go/kes"
 	"github.com/minio/madmin-go/v3"
+	"github.com/minio/minio/internal/auth"
 	"github.com/minio/minio/internal/kms"
 	"github.com/minio/minio/internal/logger"
 	"github.com/minio/pkg/v3/policy"
@@ -56,7 +56,7 @@ func (a kmsAPIHandlers) KMSStatusHandler(w http.ResponseWriter, r *http.Request)
 	writeSuccessResponseJSON(w, resp)
 }
 
-// KMSMetricsHandler - POST /minio/kms/v1/metrics
+// KMSMetricsHandler - GET /minio/kms/v1/metrics
 func (a kmsAPIHandlers) KMSMetricsHandler(w http.ResponseWriter, r *http.Request) {
 	ctx := newContext(r, w, "KMSMetrics")
 	defer logger.AuditLog(ctx, w, r, mustGetClaimsFromToken(r))
@@ -83,7 +83,7 @@ func (a kmsAPIHandlers) KMSMetricsHandler(w http.ResponseWriter, r *http.Request
 	}
 }
 
-// KMSAPIsHandler - POST /minio/kms/v1/apis
+// KMSAPIsHandler - GET /minio/kms/v1/apis
 func (a kmsAPIHandlers) KMSAPIsHandler(w http.ResponseWriter, r *http.Request) {
 	ctx := newContext(r, w, "KMSAPIs")
 	defer logger.AuditLog(ctx, w, r, mustGetClaimsFromToken(r))
@@ -114,7 +114,7 @@ type versionResponse struct {
 	Version string `json:"version"`
 }
 
-// KMSVersionHandler - POST /minio/kms/v1/version
+// KMSVersionHandler - GET /minio/kms/v1/version
 func (a kmsAPIHandlers) KMSVersionHandler(w http.ResponseWriter, r *http.Request) {
 	ctx := newContext(r, w, "KMSVersion")
 	defer logger.AuditLog(ctx, w, r, mustGetClaimsFromToken(r))
@@ -159,7 +159,20 @@ func (a kmsAPIHandlers) KMSCreateKeyHandler(w http.ResponseWriter, r *http.Reque
 		return
 	}
 
-	if err := GlobalKMS.CreateKey(ctx, &kms.CreateKeyRequest{Name: r.Form.Get("key-id")}); err != nil {
+	keyID := r.Form.Get("key-id")
+
+	// Ensure policy allows the user to create this key name
+	cred, owner, s3Err := validateAdminSignature(ctx, r, "")
+	if s3Err != ErrNone {
+		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(s3Err), r.URL)
+		return
+	}
+	if !checkKMSActionAllowed(r, owner, cred, policy.KMSCreateKeyAction, keyID) {
+		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrAccessDenied), r.URL)
+		return
+	}
+
+	if err := GlobalKMS.CreateKey(ctx, &kms.CreateKeyRequest{Name: keyID}); err != nil {
 		writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
 		return
 	}
@@ -171,6 +184,9 @@ func (a kmsAPIHandlers) KMSListKeysHandler(w http.ResponseWriter, r *http.Reques
 	ctx := newContext(r, w, "KMSListKeys")
 	defer logger.AuditLog(ctx, w, r, mustGetClaimsFromToken(r))
 
+	// This only checks if the action (kms:ListKeys) is allowed, it does not check
+	// each key name against the policy's Resources. We check that below, once
+	// we have the list of key names from the KMS.
 	objectAPI, _ := validateAdminReq(ctx, w, r, policy.KMSListKeysAction)
 	if objectAPI == nil {
 		return
@@ -180,7 +196,7 @@ func (a kmsAPIHandlers) KMSListKeysHandler(w http.ResponseWriter, r *http.Reques
 		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrKMSNotConfigured), r.URL)
 		return
 	}
-	names, _, err := GlobalKMS.ListKeyNames(ctx, &kms.ListRequest{
+	allKeys, _, err := GlobalKMS.ListKeys(ctx, &kms.ListRequest{
 		Prefix: r.Form.Get("pattern"),
 	})
 	if err != nil {
@@ -188,13 +204,25 @@ func (a kmsAPIHandlers) KMSListKeysHandler(w http.ResponseWriter, r *http.Reques
 		return
 	}
 
-	values := make([]kes.KeyInfo, 0, len(names))
-	for _, name := range names {
-		values = append(values, kes.KeyInfo{
-			Name: name,
-		})
+	// Get the cred and owner for checking authz below.
+	cred, owner, s3Err := validateAdminSignature(ctx, r, "")
+	if s3Err != ErrNone {
+		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(s3Err), r.URL)
+		return
 	}
-	if res, err := json.Marshal(values); err != nil {
+
+	// Now we have all the key names, for each of them, check whether the policy grants permission for
+	// the user to list it. Filter in place to leave only allowed keys.
+	n := 0
+	for _, k := range allKeys {
+		if checkKMSActionAllowed(r, owner, cred, policy.KMSListKeysAction, k.Name) {
+			allKeys[n] = k
+			n++
+		}
+	}
+	allKeys = allKeys[:n]
+
+	if res, err := json.Marshal(allKeys); err != nil {
 		writeCustomErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrInternalError), err.Error(), r.URL)
 	} else {
 		writeSuccessResponseJSON(w, res)
@@ -224,6 +252,17 @@ func (a kmsAPIHandlers) KMSKeyStatusHandler(w http.ResponseWriter, r *http.Reque
 		KeyID: keyID,
 	}
 
+	// Ensure policy allows the user to get this key's status
+	cred, owner, s3Err := validateAdminSignature(ctx, r, "")
+	if s3Err != ErrNone {
+		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(s3Err), r.URL)
+		return
+	}
+	if !checkKMSActionAllowed(r, owner, cred, policy.KMSKeyStatusAction, keyID) {
+		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrAccessDenied), r.URL)
+		return
+	}
+
 	kmsContext := kms.Context{"MinIO admin API": "KMSKeyStatusHandler"} // Context for a test key operation
 	// 1. Generate a new key using the KMS.
 	key, err := GlobalKMS.GenerateKey(ctx, &kms.GenerateKeyRequest{Name: keyID, AssociatedData: kmsContext})
@@ -274,3 +313,16 @@ func (a kmsAPIHandlers) KMSKeyStatusHandler(w http.ResponseWriter, r *http.Reque
 	}
 	writeSuccessResponseJSON(w, resp)
 }
+
+// checkKMSActionAllowed checks for authorization for a specific action on a resource.
+func checkKMSActionAllowed(r *http.Request, owner bool, cred auth.Credentials, action policy.KMSAction, resource string) bool {
+	return globalIAMSys.IsAllowed(policy.Args{
+		AccountName:     cred.AccessKey,
+		Groups:          cred.Groups,
+		Action:          policy.Action(action),
+		ConditionValues: getConditionValues(r, "", cred),
+		IsOwner:         owner,
+		Claims:          cred.Claims,
+		BucketName:      resource, // overloading BucketName as that's what the policy engine uses to assemble a Resource.
+	})
+}

cmd/kms-handlers_test.go

@@ -0,0 +1,848 @@
+// Copyright (c) 2015-2024 MinIO, Inc.
+//
+// This file is part of MinIO Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"net/http"
+	"net/http/httptest"
+	"net/url"
+	"strings"
+	"testing"
+
+	"github.com/minio/madmin-go/v3"
+	"github.com/minio/minio/internal/kms"
+	"github.com/minio/pkg/v3/policy"
+)
+
+const (
+	// KMS API paths
+	// For example: /minio/kms/v1/key/list?pattern=*
+	kmsURL           = kmsPathPrefix + kmsAPIVersionPrefix
+	kmsStatusPath    = kmsURL + "/status"
+	kmsMetricsPath   = kmsURL + "/metrics"
+	kmsAPIsPath      = kmsURL + "/apis"
+	kmsVersionPath   = kmsURL + "/version"
+	kmsKeyCreatePath = kmsURL + "/key/create"
+	kmsKeyListPath   = kmsURL + "/key/list"
+	kmsKeyStatusPath = kmsURL + "/key/status"
+
+	// Admin API paths
+	// For example: /minio/admin/v3/kms/status
+	adminURL              = adminPathPrefix + adminAPIVersionPrefix
+	kmsAdminStatusPath    = adminURL + "/kms/status"
+	kmsAdminKeyStatusPath = adminURL + "/kms/key/status"
+	kmsAdminKeyCreate     = adminURL + "/kms/key/create"
+)
+
+const (
+	userAccessKey = "miniofakeuseraccesskey"
+	userSecretKey = "miniofakeusersecret"
+)
+
+type kmsTestCase struct {
+	name   string
+	method string
+	path   string
+	query  map[string]string
+
+	// User credentials and policy for request
+	policy string
+	asRoot bool
+
+	// Wanted in response.
+	wantStatusCode int
+	wantKeyNames   []string
+	wantResp       []string
+}
+
+func TestKMSHandlersCreateKey(t *testing.T) {
+	adminTestBed, tearDown := setupKMSTest(t, true)
+	defer tearDown()
+
+	tests := []kmsTestCase{
+		// Create key test
+		{
+			name:   "create key as user with no policy want forbidden",
+			method: http.MethodPost,
+			path:   kmsKeyCreatePath,
+			query:  map[string]string{"key-id": "new-test-key"},
+			asRoot: false,
+
+			wantStatusCode: http.StatusForbidden,
+			wantResp:       []string{"AccessDenied"},
+		},
+		{
+			name:   "create key as user with no resources specified want success",
+			method: http.MethodPost,
+			path:   kmsKeyCreatePath,
+			query:  map[string]string{"key-id": "new-test-key"},
+			asRoot: false,
+
+			policy: `{"Effect": "Allow",
+				"Action": ["kms:CreateKey"] }`,
+
+			wantStatusCode: http.StatusOK,
+		},
+		{
+			name:   "create key as user set policy to allow want success",
+			method: http.MethodPost,
+			path:   kmsKeyCreatePath,
+			query:  map[string]string{"key-id": "second-new-test-key"},
+			asRoot: false,
+
+			policy: `{"Effect": "Allow",
+				"Action": ["kms:CreateKey"],
+				"Resource": ["arn:minio:kms:::second-new-test-*"] }`,
+
+			wantStatusCode: http.StatusOK,
+		},
+		{
+			name:   "create key as user set policy to non matching resource want forbidden",
+			method: http.MethodPost,
+			path:   kmsKeyCreatePath,
+			query:  map[string]string{"key-id": "third-new-test-key"},
+			asRoot: false,
+
+			policy: `{"Effect": "Allow",
+				"Action": ["kms:CreateKey"],
+				"Resource": ["arn:minio:kms:::non-matching-key-name"] }`,
+
+			wantStatusCode: http.StatusForbidden,
+			wantResp:       []string{"AccessDenied"},
+		},
+	}
+	for testNum, test := range tests {
+		t.Run(fmt.Sprintf("%d %s", testNum+1, test.name), func(t *testing.T) {
+			execKMSTest(t, test, adminTestBed)
+		})
+	}
+}
+
+func TestKMSHandlersKeyStatus(t *testing.T) {
+	adminTestBed, tearDown := setupKMSTest(t, true)
+	defer tearDown()
+
+	tests := []kmsTestCase{
+		{
+			name:   "create a first key root user",
+			method: http.MethodPost,
+			path:   kmsKeyCreatePath,
+			query:  map[string]string{"key-id": "abc-test-key"},
+			asRoot: true,
+
+			wantStatusCode: http.StatusOK,
+		},
+		{
+			name:   "key status as root want success",
+			method: http.MethodGet,
+			path:   kmsKeyStatusPath,
+			query:  map[string]string{"key-id": "abc-test-key"},
+			asRoot: true,
+
+			wantStatusCode: http.StatusOK,
+			wantResp:       []string{"abc-test-key"},
+		},
+		{
+			name:   "key status as user no policy want forbidden",
+			method: http.MethodGet,
+			path:   kmsKeyStatusPath,
+			query:  map[string]string{"key-id": "abc-test-key"},
+			asRoot: false,
+
+			wantStatusCode: http.StatusForbidden,
+			wantResp:       []string{"AccessDenied"},
+		},
+		{
+			name:   "key status as user legacy no resources specified want success",
+			method: http.MethodGet,
+			path:   kmsKeyStatusPath,
+			query:  map[string]string{"key-id": "abc-test-key"},
+			asRoot: false,
+
+			policy: `{"Effect": "Allow",
+					"Action": ["kms:KeyStatus"] }`,
+
+			wantStatusCode: http.StatusOK,
+			wantResp:       []string{"abc-test-key"},
+		},
+		{
+			name:   "key status as user set policy to allow only one key",
+			method: http.MethodGet,
+			path:   kmsKeyStatusPath,
+			query:  map[string]string{"key-id": "abc-test-key"},
+			asRoot: false,
+
+			policy: `{"Effect": "Allow",
+					"Action": ["kms:KeyStatus"],
+					"Resource": ["arn:minio:kms:::abc-test-*"] }`,
+
+			wantStatusCode: http.StatusOK,
+			wantResp:       []string{"abc-test-key"},
+		},
+		{
+			name:   "key status as user set policy to allow non-matching key",
+			method: http.MethodGet,
+			path:   kmsKeyStatusPath,
+			query:  map[string]string{"key-id": "abc-test-key"},
+			asRoot: false,
+
+			policy: `{"Effect": "Allow",
+					"Action": ["kms:KeyStatus"],
+					"Resource": ["arn:minio:kms:::xyz-test-key"] }`,
+
+			wantStatusCode: http.StatusForbidden,
+			wantResp:       []string{"AccessDenied"},
+		},
+	}
+	for testNum, test := range tests {
+		t.Run(fmt.Sprintf("%d %s", testNum+1, test.name), func(t *testing.T) {
+			execKMSTest(t, test, adminTestBed)
+		})
+	}
+}
+
+func TestKMSHandlersAPIs(t *testing.T) {
+	adminTestBed, tearDown := setupKMSTest(t, true)
+	defer tearDown()
+
+	tests := []kmsTestCase{
+		// Version test
+		{
+			name:   "version as root want success",
+			method: http.MethodGet,
+			path:   kmsVersionPath,
+			asRoot: true,
+
+			wantStatusCode: http.StatusOK,
+			wantResp:       []string{"version"},
+		},
+		{
+			name:   "version as user with no policy want forbidden",
+			method: http.MethodGet,
+			path:   kmsVersionPath,
+			asRoot: false,
+
+			wantStatusCode: http.StatusForbidden,
+			wantResp:       []string{"AccessDenied"},
+		},
+		{
+			name:   "version as user with policy ignores resource want success",
+			method: http.MethodGet,
+			path:   kmsVersionPath,
+			asRoot: false,
+
+			policy: `{"Effect": "Allow",
+				"Action": ["kms:Version"],
+				"Resource": ["arn:minio:kms:::does-not-matter-it-is-ignored"] }`,
+
+			wantStatusCode: http.StatusOK,
+			wantResp:       []string{"version"},
+		},
+
+		// APIs test
+		{
+			name:   "apis as root want success",
+			method: http.MethodGet,
+			path:   kmsAPIsPath,
+			asRoot: true,
+
+			wantStatusCode: http.StatusOK,
+			wantResp:       []string{"stub/path"},
+		},
+		{
+			name:   "apis as user with no policy want forbidden",
+			method: http.MethodGet,
+			path:   kmsAPIsPath,
+			asRoot: false,
+
+			wantStatusCode: http.StatusForbidden,
+			wantResp:       []string{"AccessDenied"},
+		},
+		{
+			name:   "apis as user with policy ignores resource want success",
+			method: http.MethodGet,
+			path:   kmsAPIsPath,
+			asRoot: false,
+
+			policy: `{"Effect": "Allow",
+				"Action": ["kms:API"],
+				"Resource": ["arn:minio:kms:::does-not-matter-it-is-ignored"] }`,
+
+			wantStatusCode: http.StatusOK,
+			wantResp:       []string{"stub/path"},
+		},
+
+		// Metrics test
+		{
+			name:   "metrics as root want success",
+			method: http.MethodGet,
+			path:   kmsMetricsPath,
+			asRoot: true,
+
+			wantStatusCode: http.StatusOK,
+			wantResp:       []string{"kms"},
+		},
+		{
+			name:   "metrics as user with no policy want forbidden",
+			method: http.MethodGet,
+			path:   kmsMetricsPath,
+			asRoot: false,
+
+			wantStatusCode: http.StatusForbidden,
+			wantResp:       []string{"AccessDenied"},
+		},
+		{
+			name:   "metrics as user with policy ignores resource want success",
+			method: http.MethodGet,
+			path:   kmsMetricsPath,
+			asRoot: false,
+
+			policy: `{"Effect": "Allow",
+				"Action": ["kms:Metrics"],
+				"Resource": ["arn:minio:kms:::does-not-matter-it-is-ignored"] }`,
+
+			wantStatusCode: http.StatusOK,
+			wantResp:       []string{"kms"},
+		},
+
+		// Status tests
+		{
+			name:   "status as root want success",
+			method: http.MethodGet,
+			path:   kmsStatusPath,
+			asRoot: true,
+
+			wantStatusCode: http.StatusOK,
+			wantResp:       []string{"MinIO builtin"},
+		},
+		{
+			name:   "status as user with no policy want forbidden",
+			method: http.MethodGet,
+			path:   kmsStatusPath,
+			asRoot: false,
+
+			wantStatusCode: http.StatusForbidden,
+			wantResp:       []string{"AccessDenied"},
+		},
+		{
+			name:   "status as user with policy ignores resource want success",
+			method: http.MethodGet,
+			path:   kmsStatusPath,
+			asRoot: false,
+
+			policy: `{"Effect": "Allow",
+				"Action": ["kms:Status"],
+				"Resource": ["arn:minio:kms:::does-not-matter-it-is-ignored"]}`,
+
+			wantStatusCode: http.StatusOK,
+			wantResp:       []string{"MinIO builtin"},
+		},
+	}
+	for testNum, test := range tests {
+		t.Run(fmt.Sprintf("%d %s", testNum+1, test.name), func(t *testing.T) {
+			execKMSTest(t, test, adminTestBed)
+		})
+	}
+}
+
+func TestKMSHandlersListKeys(t *testing.T) {
+	adminTestBed, tearDown := setupKMSTest(t, true)
+	defer tearDown()
+
+	tests := []kmsTestCase{
+		{
+			name:   "create a first key root user",
+			method: http.MethodPost,
+			path:   kmsKeyCreatePath,
+			query:  map[string]string{"key-id": "abc-test-key"},
+			asRoot: true,
+
+			wantStatusCode: http.StatusOK,
+		},
+		{
+			name:   "create a second key root user",
+			method: http.MethodPost,
+			path:   kmsKeyCreatePath,
+			query:  map[string]string{"key-id": "xyz-test-key"},
+			asRoot: true,
+
+			wantStatusCode: http.StatusOK,
+		},
+
+		// List keys tests
+		{
+			name:   "list keys as root want all to be returned",
+			method: http.MethodGet,
+			path:   kmsKeyListPath,
+			query:  map[string]string{"pattern": "*"},
+			asRoot: true,
+
+			wantStatusCode: http.StatusOK,
+			wantKeyNames:   []string{"default-test-key", "abc-test-key", "xyz-test-key"},
+		},
+		{
+			name:   "list keys as user with no policy want forbidden",
+			method: http.MethodGet,
+			path:   kmsKeyListPath,
+			query:  map[string]string{"pattern": "*"},
+			asRoot: false,
+
+			wantStatusCode: http.StatusForbidden,
+			wantResp:       []string{"AccessDenied"},
+		},
+		{
+			name:   "list keys as user with no resources specified want success",
+			method: http.MethodGet,
+			path:   kmsKeyListPath,
+			query:  map[string]string{"pattern": "*"},
+			asRoot: false,
+
+			policy: `{"Effect": "Allow",
+				"Action": ["kms:ListKeys"]
+			}`,
+
+			wantStatusCode: http.StatusOK,
+			wantKeyNames:   []string{"default-test-key", "abc-test-key", "xyz-test-key"},
+		},
+		{
+			name:   "list keys as user set policy resource to allow only one key",
+			method: http.MethodGet,
+			path:   kmsKeyListPath,
+			query:  map[string]string{"pattern": "*"},
+			asRoot: false,
+
+			policy: `{"Effect": "Allow",
+					"Action": ["kms:ListKeys"],
+					"Resource": ["arn:minio:kms:::abc*"]}`,
+
+			wantStatusCode: http.StatusOK,
+			wantKeyNames:   []string{"abc-test-key"},
+		},
+		{
+			name:   "list keys as user set policy to allow only one key, use pattern that includes correct key",
+			method: http.MethodGet,
+			path:   kmsKeyListPath,
+			query:  map[string]string{"pattern": "abc*"},
+
+			policy: `{"Effect": "Allow",
+					"Action": ["kms:ListKeys"],
+					"Resource": ["arn:minio:kms:::abc*"]}`,
+
+			wantStatusCode: http.StatusOK,
+			wantKeyNames:   []string{"abc-test-key"},
+		},
+		{
+			name:   "list keys as user set policy to allow only one key, use pattern that excludes correct key",
+			method: http.MethodGet,
+			path:   kmsKeyListPath,
+			query:  map[string]string{"pattern": "xyz*"},
+			asRoot: false,
+
+			policy: `{"Effect": "Allow",
+					"Action": ["kms:ListKeys"],
+					"Resource": ["arn:minio:kms:::abc*"]}`,
+
+			wantStatusCode: http.StatusOK,
+			wantKeyNames:   []string{},
+		},
+		{
+			name:   "list keys as user set policy that has no matching key resources",
+			method: http.MethodGet,
+			path:   kmsKeyListPath,
+			query:  map[string]string{"pattern": "*"},
+			asRoot: false,
+
+			policy: `{"Effect": "Allow",
+					"Action": ["kms:ListKeys"],
+					"Resource": ["arn:minio:kms:::nonematch*"]}`,
+
+			wantStatusCode: http.StatusOK,
+			wantKeyNames:   []string{},
+		},
+		{
+			name:   "list keys as user set policy that allows listing but denies specific keys",
+			method: http.MethodGet,
+			path:   kmsKeyListPath,
+			query:  map[string]string{"pattern": "*"},
+			asRoot: false,
+
+			// It looks like this should allow listing any key that isn't "default-test-key", however
+			// the policy engine matches all Deny statements first, without regard to Resources (for KMS).
+			// This is for backwards compatibility where historically KMS statements ignored Resources.
+			policy: `{
+						"Effect": "Allow",
+						"Action": ["kms:ListKeys"]
+					},{
+						"Effect": "Deny",
+						"Action": ["kms:ListKeys"],
+						"Resource": ["arn:minio:kms:::default-test-key"]
+					}`,
+
+			wantStatusCode: http.StatusForbidden,
+			wantResp:       []string{"AccessDenied"},
+		},
+	}
+
+	for testNum, test := range tests {
+		t.Run(fmt.Sprintf("%d %s", testNum+1, test.name), func(t *testing.T) {
+			execKMSTest(t, test, adminTestBed)
+		})
+	}
+}
+
+func TestKMSHandlerAdminAPI(t *testing.T) {
+	adminTestBed, tearDown := setupKMSTest(t, true)
+	defer tearDown()
+
+	tests := []kmsTestCase{
+		// Create key tests
+		{
+			name:   "create a key root user",
+			method: http.MethodPost,
+			path:   kmsAdminKeyCreate,
+			query:  map[string]string{"key-id": "abc-test-key"},
+			asRoot: true,
+
+			wantStatusCode: http.StatusOK,
+		},
+		{
+			name:   "create key as user with no policy want forbidden",
+			method: http.MethodPost,
+			path:   kmsAdminKeyCreate,
+			query:  map[string]string{"key-id": "new-test-key"},
+			asRoot: false,
+
+			wantStatusCode: http.StatusForbidden,
+			wantResp:       []string{"AccessDenied"},
+		},
+		{
+			name:   "create key as user with no resources specified want success",
+			method: http.MethodPost,
+			path:   kmsAdminKeyCreate,
+			query:  map[string]string{"key-id": "new-test-key"},
+			asRoot: false,
+
+			policy: `{"Effect": "Allow",
+				"Action": ["admin:KMSCreateKey"] }`,
+
+			wantStatusCode: http.StatusOK,
+		},
+		{
+			name:   "create key as user set policy to non matching resource want success",
+			method: http.MethodPost,
+			path:   kmsAdminKeyCreate,
+			query:  map[string]string{"key-id": "third-new-test-key"},
+			asRoot: false,
+
+			// Admin actions ignore Resources
+			policy: `{"Effect": "Allow",
+				"Action": ["admin:KMSCreateKey"],
+				"Resource": ["arn:minio:kms:::this-is-disregarded"] }`,
+
+			wantStatusCode: http.StatusOK,
+		},
+
+		// Status tests
+		{
+			name:   "status as root want success",
+			method: http.MethodPost,
+			path:   kmsAdminStatusPath,
+			asRoot: true,
+
+			wantStatusCode: http.StatusOK,
+			wantResp:       []string{"MinIO builtin"},
+		},
+		{
+			name:   "status as user with no policy want forbidden",
+			method: http.MethodPost,
+			path:   kmsAdminStatusPath,
+			asRoot: false,
+
+			wantStatusCode: http.StatusForbidden,
+			wantResp:       []string{"AccessDenied"},
+		},
+		{
+			name:   "status as user with policy ignores resource want success",
+			method: http.MethodPost,
+			path:   kmsAdminStatusPath,
+			asRoot: false,
+
+			policy: `{"Effect": "Allow",
+				"Action": ["admin:KMSKeyStatus"],
+				"Resource": ["arn:minio:kms:::does-not-matter-it-is-ignored"] }`,
+
+			wantStatusCode: http.StatusOK,
+			wantResp:       []string{"MinIO builtin"},
+		},
+
+		// Key status tests
+		{
+			name:   "key status as root want success",
+			method: http.MethodGet,
+			path:   kmsAdminKeyStatusPath,
+			asRoot: true,
+
+			wantStatusCode: http.StatusOK,
+			wantResp:       []string{"key-id"},
+		},
+		{
+			name:   "key status as user with no policy want forbidden",
+			method: http.MethodGet,
+			path:   kmsAdminKeyStatusPath,
+			asRoot: false,
+
+			wantStatusCode: http.StatusForbidden,
+			wantResp:       []string{"AccessDenied"},
+		},
+		{
+			name:   "key status as user with policy ignores resource want success",
+			method: http.MethodGet,
+			path:   kmsAdminKeyStatusPath,
+			asRoot: false,
+
+			policy: `{"Effect": "Allow",
+				"Action": ["admin:KMSKeyStatus"],
+				"Resource": ["arn:minio:kms:::does-not-matter-it-is-ignored"] }`,
+
+			wantStatusCode: http.StatusOK,
+			wantResp:       []string{"key-id"},
+		},
+	}
+
+	for testNum, test := range tests {
+		t.Run(fmt.Sprintf("%d %s", testNum+1, test.name), func(t *testing.T) {
+			execKMSTest(t, test, adminTestBed)
+		})
+	}
+}
+
+// execKMSTest runs a single test case for KMS handlers
+func execKMSTest(t *testing.T, test kmsTestCase, adminTestBed *adminErasureTestBed) {
+	var accessKey, secretKey string
+	if test.asRoot {
+		accessKey, secretKey = globalActiveCred.AccessKey, globalActiveCred.SecretKey
+	} else {
+		setupKMSUser(t, userAccessKey, userSecretKey, test.policy)
+		accessKey = userAccessKey
+		secretKey = userSecretKey
+	}
+
+	req := buildKMSRequest(t, test.method, test.path, accessKey, secretKey, test.query)
+	rec := httptest.NewRecorder()
+	adminTestBed.router.ServeHTTP(rec, req)
+
+	t.Logf("HTTP req: %s, resp code: %d, resp body: %s", req.URL.String(), rec.Code, rec.Body.String())
+
+	// Check status code
+	if rec.Code != test.wantStatusCode {
+		t.Errorf("want status code %d, got %d", test.wantStatusCode, rec.Code)
+	}
+
+	// Check returned key list is correct
+	if test.wantKeyNames != nil {
+		keys := []madmin.KMSKeyInfo{}
+		err := json.Unmarshal(rec.Body.Bytes(), &keys)
+		if err != nil {
+			t.Fatal(err)
+		}
+		if len(keys) != len(test.wantKeyNames) {
+			t.Fatalf("want %d keys, got %d", len(test.wantKeyNames), len(keys))
+		}
+
+		for i, want := range keys {
+			if want.CreatedBy != kms.StubCreatedBy {
+				t.Fatalf("want key created by %s, got %s", kms.StubCreatedBy, want.CreatedBy)
+			}
+			if want.CreatedAt != kms.StubCreatedAt {
+				t.Fatalf("want key created at %s, got %s", kms.StubCreatedAt, want.CreatedAt)
+			}
+			if test.wantKeyNames[i] != want.Name {
+				t.Fatalf("want key name %s, got %s", test.wantKeyNames[i], want.Name)
+			}
+		}
+	}
+
+	// Check generic text in the response
+	if test.wantResp != nil {
+		for _, want := range test.wantResp {
+			if !strings.Contains(rec.Body.String(), want) {
+				t.Fatalf("want response to contain %s, got %s", want, rec.Body.String())
+			}
+		}
+	}
+}
+
+// TestKMSHandlerNotConfiguredOrInvalidCreds tests KMS handlers for situations where KMS is not configured
+// or invalid credentials are provided.
+func TestKMSHandlerNotConfiguredOrInvalidCreds(t *testing.T) {
+	adminTestBed, tearDown := setupKMSTest(t, false)
+	defer tearDown()
+
+	tests := []struct {
+		name   string
+		method string
+		path   string
+		query  map[string]string
+	}{
+		{
+			name:   "GET status",
+			method: http.MethodGet,
+			path:   kmsStatusPath,
+		},
+		{
+			name:   "GET metrics",
+			method: http.MethodGet,
+			path:   kmsMetricsPath,
+		},
+		{
+			name:   "GET apis",
+			method: http.MethodGet,
+			path:   kmsAPIsPath,
+		},
+		{
+			name:   "GET version",
+			method: http.MethodGet,
+			path:   kmsVersionPath,
+		},
+		{
+			name:   "POST key create",
+			method: http.MethodPost,
+			path:   kmsKeyCreatePath,
+			query:  map[string]string{"key-id": "master-key-id"},
+		},
+		{
+			name:   "GET key list",
+			method: http.MethodGet,
+			path:   kmsKeyListPath,
+			query:  map[string]string{"pattern": "*"},
+		},
+		{
+			name:   "GET key status",
+			method: http.MethodGet,
+			path:   kmsKeyStatusPath,
+			query:  map[string]string{"key-id": "master-key-id"},
+		},
+	}
+
+	// Test when the GlobalKMS is not configured
+	for _, test := range tests {
+		t.Run(test.name+" not configured", func(t *testing.T) {
+			req := buildKMSRequest(t, test.method, test.path, "", "", test.query)
+			rec := httptest.NewRecorder()
+			adminTestBed.router.ServeHTTP(rec, req)
+			if rec.Code != http.StatusNotImplemented {
+				t.Errorf("want status code %d, got %d", http.StatusNotImplemented, rec.Code)
+			}
+		})
+	}
+
+	// Test when the GlobalKMS is configured but the credentials are invalid
+	GlobalKMS = kms.NewStub("default-test-key")
+	for _, test := range tests {
+		t.Run(test.name+" invalid credentials", func(t *testing.T) {
+			req := buildKMSRequest(t, test.method, test.path, userAccessKey, userSecretKey, test.query)
+			rec := httptest.NewRecorder()
+			adminTestBed.router.ServeHTTP(rec, req)
+			if rec.Code != http.StatusForbidden {
+				t.Errorf("want status code %d, got %d", http.StatusForbidden, rec.Code)
+			}
+		})
+	}
+}
+
+func setupKMSTest(t *testing.T, enableKMS bool) (*adminErasureTestBed, func()) {
+	adminTestBed, err := prepareAdminErasureTestBed(context.Background())
+	if err != nil {
+		t.Fatal(err)
+	}
+	registerKMSRouter(adminTestBed.router)
+
+	if enableKMS {
+		GlobalKMS = kms.NewStub("default-test-key")
+	}
+
+	tearDown := func() {
+		adminTestBed.TearDown()
+		GlobalKMS = nil
+	}
+	return adminTestBed, tearDown
+}
+
+func buildKMSRequest(t *testing.T, method, path, accessKey, secretKey string, query map[string]string) *http.Request {
+	if len(query) > 0 {
+		queryVal := url.Values{}
+		for k, v := range query {
+			queryVal.Add(k, v)
+		}
+		path = path + "?" + queryVal.Encode()
+	}
+
+	if accessKey == "" && secretKey == "" {
+		accessKey = globalActiveCred.AccessKey
+		secretKey = globalActiveCred.SecretKey
+	}
+
+	req, err := newTestSignedRequestV4(method, path, 0, nil, accessKey, secretKey, nil)
+	if err != nil {
+		t.Fatal(err)
+	}
+	return req
+}
+
+// setupKMSUser is a test helper that creates a new user with the provided access key and secret key
+// and applies the given policy to the user.
+func setupKMSUser(t *testing.T, accessKey, secretKey, p string) {
+	ctx := context.Background()
+	createUserParams := madmin.AddOrUpdateUserReq{
+		SecretKey: secretKey,
+		Status:    madmin.AccountEnabled,
+	}
+	_, err := globalIAMSys.CreateUser(ctx, accessKey, createUserParams)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	testKMSPolicyName := "testKMSPolicy"
+	if p != "" {
+		p = `{"Version":"2012-10-17","Statement":[` + p + `]}`
+		policyData, err := policy.ParseConfig(strings.NewReader(p))
+		if err != nil {
+			t.Fatal(err)
+		}
+		_, err = globalIAMSys.SetPolicy(ctx, testKMSPolicyName, *policyData)
+		if err != nil {
+			t.Fatal(err)
+		}
+		_, err = globalIAMSys.PolicyDBSet(ctx, accessKey, testKMSPolicyName, regUser, false)
+		if err != nil {
+			t.Fatal(err)
+		}
+	} else {
+		err = globalIAMSys.DeletePolicy(ctx, testKMSPolicyName, false)
+		if err != nil {
+			t.Fatal(err)
+		}
+		_, err = globalIAMSys.PolicyDBSet(ctx, accessKey, "", regUser, false)
+		if err != nil {
+			t.Fatal(err)
+		}
+	}
+}

cmd/license-update.go

@@ -1,117 +0,0 @@
-// Copyright (c) 2015-2023 MinIO, Inc.
-//
-// This file is part of MinIO Object Storage stack
-//
-// This program is free software: you can redistribute it and/or modify
-// it under the terms of the GNU Affero General Public License as published by
-// the Free Software Foundation, either version 3 of the License, or
-// (at your option) any later version.
-//
-// This program is distributed in the hope that it will be useful
-// but WITHOUT ANY WARRANTY; without even the implied warranty of
-// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-// GNU Affero General Public License for more details.
-//
-// You should have received a copy of the GNU Affero General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
-
-package cmd
-
-import (
-	"context"
-	"fmt"
-	"math/rand"
-	"time"
-
-	"github.com/tidwall/gjson"
-)
-
-const (
-	licUpdateCycle = 24 * time.Hour * 30
-	licRenewPath   = "/api/cluster/renew-license"
-)
-
-// initlicenseUpdateJob start the periodic license update job in the background.
-func initLicenseUpdateJob(ctx context.Context, objAPI ObjectLayer) {
-	go func() {
-		r := rand.New(rand.NewSource(time.Now().UnixNano()))
-		// Leader node (that successfully acquires the lock inside licenceUpdaterLoop)
-		// will keep performing the license update. If the leader goes down for some
-		// reason, the lock will be released and another node will acquire it and
-		// take over because of this loop.
-		for {
-			licenceUpdaterLoop(ctx, objAPI)
-
-			// license update stopped for some reason.
-			// sleep for some time and try again.
-			duration := time.Duration(r.Float64() * float64(time.Hour))
-			if duration < time.Second {
-				// Make sure to sleep at least a second to avoid high CPU ticks.
-				duration = time.Second
-			}
-			time.Sleep(duration)
-		}
-	}()
-}
-
-func licenceUpdaterLoop(ctx context.Context, objAPI ObjectLayer) {
-	ctx, cancel := globalLeaderLock.GetLock(ctx)
-	defer cancel()
-
-	licenseUpdateTimer := time.NewTimer(licUpdateCycle)
-	defer licenseUpdateTimer.Stop()
-
-	for {
-		select {
-		case <-ctx.Done():
-			return
-		case <-licenseUpdateTimer.C:
-
-			if globalSubnetConfig.Registered() {
-				performLicenseUpdate(ctx, objAPI)
-			}
-
-			// Reset the timer for next cycle.
-			licenseUpdateTimer.Reset(licUpdateCycle)
-		}
-	}
-}
-
-func performLicenseUpdate(ctx context.Context, objectAPI ObjectLayer) {
-	// the subnet license renewal api renews the license only
-	// if required e.g. when it is expiring soon
-	url := globalSubnetConfig.BaseURL + licRenewPath
-
-	resp, err := globalSubnetConfig.Post(url, nil)
-	if err != nil {
-		subnetLogIf(ctx, fmt.Errorf("error from %s: %w", url, err))
-		return
-	}
-
-	r := gjson.Parse(resp).Get("license_v2")
-	if r.Index == 0 {
-		internalLogIf(ctx, fmt.Errorf("license not found in response from %s", url))
-		return
-	}
-
-	lic := r.String()
-	if lic == globalSubnetConfig.License {
-		// license hasn't changed.
-		return
-	}
-
-	kv := "subnet license=" + lic
-	result, err := setConfigKV(ctx, objectAPI, []byte(kv))
-	if err != nil {
-		internalLogIf(ctx, fmt.Errorf("error setting subnet license config: %w", err))
-		return
-	}
-
-	if result.Dynamic {
-		if err := applyDynamicConfigForSubSys(GlobalContext, objectAPI, result.Cfg, result.SubSys); err != nil {
-			subnetLogIf(ctx, fmt.Errorf("error applying subnet dynamic config: %w", err))
-			return
-		}
-		globalNotificationSys.SignalConfigReload(result.SubSys)
-	}
-}

cmd/local-locker.go

@@ -31,16 +31,16 @@ import (
 
 // lockRequesterInfo stores various info from the client for each lock that is requested.
 type lockRequesterInfo struct {
-	Name            string    // name of the resource lock was requested for
-	Writer          bool      // Bool whether write or read lock.
-	UID             string    // UID to uniquely identify request of client.
-	Timestamp       time.Time // Timestamp set at the time of initialization.
-	TimeLastRefresh time.Time // Timestamp for last lock refresh.
-	Source          string    // Contains line, function and filename requesting the lock.
-	Group           bool      // indicates if it was a group lock.
-	Owner           string    // Owner represents the UUID of the owner who originally requested the lock.
-	Quorum          int       // Quorum represents the quorum required for this lock to be active.
-	idx             int       `msg:"-"` // index of the lock in the lockMap.
+	Name            string // name of the resource lock was requested for
+	Writer          bool   // Bool whether write or read lock.
+	UID             string // UID to uniquely identify request of client.
+	Timestamp       int64  // Timestamp set at the time of initialization.
+	TimeLastRefresh int64  // Timestamp for last lock refresh.
+	Source          string // Contains line, function and filename requesting the lock.
+	Group           bool   // indicates if it was a group lock.
+	Owner           string // Owner represents the UUID of the owner who originally requested the lock.
+	Quorum          int    // Quorum represents the quorum required for this lock to be active.
+	idx             int    `msg:"-"` // index of the lock in the lockMap.
 }
 
 // isWriteLock returns whether the lock is a write or read lock.
@@ -87,6 +87,7 @@ func (l *localLocker) Lock(ctx context.Context, args dsync.LockArgs) (reply bool
 
 	// No locks held on the all resources, so claim write
 	// lock on all resources at once.
+	now := UTCNow()
 	for i, resource := range args.Resources {
 		l.lockMap[resource] = []lockRequesterInfo{
 			{
@@ -95,10 +96,10 @@ func (l *localLocker) Lock(ctx context.Context, args dsync.LockArgs) (reply bool
 				Source:          args.Source,
 				Owner:           args.Owner,
 				UID:             args.UID,
-				Timestamp:       UTCNow(),
-				TimeLastRefresh: UTCNow(),
+				Timestamp:       now.UnixNano(),
+				TimeLastRefresh: now.UnixNano(),
 				Group:           len(args.Resources) > 1,
-				Quorum:          args.Quorum,
+				Quorum:          *args.Quorum,
 				idx:             i,
 			},
 		}
@@ -108,7 +109,7 @@ func (l *localLocker) Lock(ctx context.Context, args dsync.LockArgs) (reply bool
 }
 
 func formatUUID(s string, idx int) string {
-	return s + strconv.Itoa(idx)
+	return concat(s, strconv.Itoa(idx))
 }
 
 func (l *localLocker) Unlock(_ context.Context, args dsync.LockArgs) (reply bool, err error) {
@@ -167,17 +168,19 @@ func (l *localLocker) RLock(ctx context.Context, args dsync.LockArgs) (reply boo
 	l.mutex.Lock()
 	defer l.mutex.Unlock()
 	resource := args.Resources[0]
+	now := UTCNow()
 	lrInfo := lockRequesterInfo{
 		Name:            resource,
 		Writer:          false,
 		Source:          args.Source,
 		Owner:           args.Owner,
 		UID:             args.UID,
-		Timestamp:       UTCNow(),
-		TimeLastRefresh: UTCNow(),
-		Quorum:          args.Quorum,
+		Timestamp:       now.UnixNano(),
+		TimeLastRefresh: now.UnixNano(),
+		Quorum:          *args.Quorum,
 	}
-	if lri, ok := l.lockMap[resource]; ok {
+	lri, ok := l.lockMap[resource]
+	if ok {
 		if reply = !isWriteLock(lri); reply {
 			// Unless there is a write lock
 			l.lockMap[resource] = append(l.lockMap[resource], lrInfo)
@@ -351,9 +354,10 @@ func (l *localLocker) Refresh(ctx context.Context, args dsync.LockArgs) (refresh
 				delete(l.lockUID, formatUUID(args.UID, idx))
 				return idx > 0, nil
 			}
+			now := UTCNow()
 			for i := range lris {
 				if lris[i].UID == args.UID {
-					lris[i].TimeLastRefresh = UTCNow()
+					lris[i].TimeLastRefresh = now.UnixNano()
 				}
 			}
 			idx++
@@ -376,7 +380,7 @@ func (l *localLocker) expireOldLocks(interval time.Duration) {
 		modified := false
 		for i := 0; i < len(lris); {
 			lri := &lris[i]
-			if time.Since(lri.TimeLastRefresh) > interval {
+			if time.Since(time.Unix(0, lri.TimeLastRefresh)) > interval {
 				delete(l.lockUID, formatUUID(lri.UID, lri.idx))
 				if len(lris) == 1 {
 					// Remove the write lock.

cmd/local-locker_gen.go

@@ -204,13 +204,13 @@ func (z *lockRequesterInfo) DecodeMsg(dc *msgp.Reader) (err error) {
 				return
 			}
 		case "Timestamp":
-			z.Timestamp, err = dc.ReadTime()
+			z.Timestamp, err = dc.ReadInt64()
 			if err != nil {
 				err = msgp.WrapError(err, "Timestamp")
 				return
 			}
 		case "TimeLastRefresh":
-			z.TimeLastRefresh, err = dc.ReadTime()
+			z.TimeLastRefresh, err = dc.ReadInt64()
 			if err != nil {
 				err = msgp.WrapError(err, "TimeLastRefresh")
 				return
@@ -288,7 +288,7 @@ func (z *lockRequesterInfo) EncodeMsg(en *msgp.Writer) (err error) {
 	if err != nil {
 		return
 	}
-	err = en.WriteTime(z.Timestamp)
+	err = en.WriteInt64(z.Timestamp)
 	if err != nil {
 		err = msgp.WrapError(err, "Timestamp")
 		return
@@ -298,7 +298,7 @@ func (z *lockRequesterInfo) EncodeMsg(en *msgp.Writer) (err error) {
 	if err != nil {
 		return
 	}
-	err = en.WriteTime(z.TimeLastRefresh)
+	err = en.WriteInt64(z.TimeLastRefresh)
 	if err != nil {
 		err = msgp.WrapError(err, "TimeLastRefresh")
 		return
@@ -361,10 +361,10 @@ func (z *lockRequesterInfo) MarshalMsg(b []byte) (o []byte, err error) {
 	o = msgp.AppendString(o, z.UID)
 	// string "Timestamp"
 	o = append(o, 0xa9, 0x54, 0x69, 0x6d, 0x65, 0x73, 0x74, 0x61, 0x6d, 0x70)
-	o = msgp.AppendTime(o, z.Timestamp)
+	o = msgp.AppendInt64(o, z.Timestamp)
 	// string "TimeLastRefresh"
 	o = append(o, 0xaf, 0x54, 0x69, 0x6d, 0x65, 0x4c, 0x61, 0x73, 0x74, 0x52, 0x65, 0x66, 0x72, 0x65, 0x73, 0x68)
-	o = msgp.AppendTime(o, z.TimeLastRefresh)
+	o = msgp.AppendInt64(o, z.TimeLastRefresh)
 	// string "Source"
 	o = append(o, 0xa6, 0x53, 0x6f, 0x75, 0x72, 0x63, 0x65)
 	o = msgp.AppendString(o, z.Source)
@@ -417,13 +417,13 @@ func (z *lockRequesterInfo) UnmarshalMsg(bts []byte) (o []byte, err error) {
 				return
 			}
 		case "Timestamp":
-			z.Timestamp, bts, err = msgp.ReadTimeBytes(bts)
+			z.Timestamp, bts, err = msgp.ReadInt64Bytes(bts)
 			if err != nil {
 				err = msgp.WrapError(err, "Timestamp")
 				return
 			}
 		case "TimeLastRefresh":
-			z.TimeLastRefresh, bts, err = msgp.ReadTimeBytes(bts)
+			z.TimeLastRefresh, bts, err = msgp.ReadInt64Bytes(bts)
 			if err != nil {
 				err = msgp.WrapError(err, "TimeLastRefresh")
 				return
@@ -466,7 +466,7 @@ func (z *lockRequesterInfo) UnmarshalMsg(bts []byte) (o []byte, err error) {
 
 // Msgsize returns an upper bound estimate of the number of bytes occupied by the serialized message
 func (z *lockRequesterInfo) Msgsize() (s int) {
-	s = 1 + 5 + msgp.StringPrefixSize + len(z.Name) + 7 + msgp.BoolSize + 4 + msgp.StringPrefixSize + len(z.UID) + 10 + msgp.TimeSize + 16 + msgp.TimeSize + 7 + msgp.StringPrefixSize + len(z.Source) + 6 + msgp.BoolSize + 6 + msgp.StringPrefixSize + len(z.Owner) + 7 + msgp.IntSize
+	s = 1 + 5 + msgp.StringPrefixSize + len(z.Name) + 7 + msgp.BoolSize + 4 + msgp.StringPrefixSize + len(z.UID) + 10 + msgp.Int64Size + 16 + msgp.Int64Size + 7 + msgp.StringPrefixSize + len(z.Source) + 6 + msgp.BoolSize + 6 + msgp.StringPrefixSize + len(z.Owner) + 7 + msgp.IntSize
 	return
 }
 

cmd/local-locker_test.go

@@ -32,6 +32,7 @@ import (
 func TestLocalLockerExpire(t *testing.T) {
 	wResources := make([]string, 1000)
 	rResources := make([]string, 1000)
+	quorum := 0
 	l := newLocker()
 	ctx := context.Background()
 	for i := range wResources {
@@ -40,7 +41,7 @@ func TestLocalLockerExpire(t *testing.T) {
 			Resources: []string{mustGetUUID()},
 			Source:    t.Name(),
 			Owner:     "owner",
-			Quorum:    0,
+			Quorum:    &quorum,
 		}
 		ok, err := l.Lock(ctx, arg)
 		if err != nil {
@@ -58,14 +59,14 @@ func TestLocalLockerExpire(t *testing.T) {
 			Resources: []string{name},
 			Source:    t.Name(),
 			Owner:     "owner",
-			Quorum:    0,
+			Quorum:    &quorum,
 		}
 		ok, err := l.RLock(ctx, arg)
 		if err != nil {
 			t.Fatal(err)
 		}
 		if !ok {
-			t.Fatal("did not get write lock")
+			t.Fatal("did not get read lock")
 		}
 		// RLock twice
 		ok, err = l.RLock(ctx, arg)
@@ -112,6 +113,7 @@ func TestLocalLockerUnlock(t *testing.T) {
 	rUIDs := make([]string, 0, n*2)
 	l := newLocker()
 	ctx := context.Background()
+	quorum := 0
 	for i := range wResources {
 		names := [m]string{}
 		for j := range names {
@@ -123,7 +125,7 @@ func TestLocalLockerUnlock(t *testing.T) {
 			Resources: names[:],
 			Source:    t.Name(),
 			Owner:     "owner",
-			Quorum:    0,
+			Quorum:    &quorum,
 		}
 		ok, err := l.Lock(ctx, arg)
 		if err != nil {
@@ -144,7 +146,7 @@ func TestLocalLockerUnlock(t *testing.T) {
 			Resources: []string{name},
 			Source:    t.Name(),
 			Owner:     "owner",
-			Quorum:    0,
+			Quorum:    &quorum,
 		}
 		ok, err := l.RLock(ctx, arg)
 		if err != nil {
@@ -183,7 +185,7 @@ func TestLocalLockerUnlock(t *testing.T) {
 			Resources: []string{name},
 			Source:    t.Name(),
 			Owner:     "owner",
-			Quorum:    0,
+			Quorum:    &quorum,
 		}
 		ok, err := l.RUnlock(ctx, arg)
 		if err != nil {
@@ -203,6 +205,7 @@ func TestLocalLockerUnlock(t *testing.T) {
 	if len(l.lockUID) != len(rResources)+len(wResources)*m {
 		t.Fatalf("lockUID len, got %d, want %d + %d", len(l.lockUID), len(rResources), len(wResources)*m)
 	}
+
 	// RUnlock again, different uids
 	for i, name := range rResources {
 		arg := dsync.LockArgs{
@@ -210,7 +213,7 @@ func TestLocalLockerUnlock(t *testing.T) {
 			Resources: []string{name},
 			Source:    "minio",
 			Owner:     "owner",
-			Quorum:    0,
+			Quorum:    &quorum,
 		}
 		ok, err := l.RUnlock(ctx, arg)
 		if err != nil {
@@ -238,7 +241,7 @@ func TestLocalLockerUnlock(t *testing.T) {
 			Resources: names[:],
 			Source:    "minio",
 			Owner:     "owner",
-			Quorum:    0,
+			Quorum:    &quorum,
 		}
 		ok, err := l.Unlock(ctx, arg)
 		if err != nil {
@@ -261,6 +264,7 @@ func TestLocalLockerUnlock(t *testing.T) {
 
 func Test_localLocker_expireOldLocksExpire(t *testing.T) {
 	rng := rand.New(rand.NewSource(0))
+	quorum := 0
 	// Numbers of unique locks
 	for _, locks := range []int{100, 1000, 1e6} {
 		if testing.Short() && locks > 100 {
@@ -289,7 +293,7 @@ func Test_localLocker_expireOldLocksExpire(t *testing.T) {
 								Resources: res,
 								Source:    hex.EncodeToString(tmp[:8]),
 								Owner:     hex.EncodeToString(tmp[8:]),
-								Quorum:    0,
+								Quorum:    &quorum,
 							})
 							if !ok || err != nil {
 								t.Fatal("failed:", err, ok)
@@ -311,7 +315,7 @@ func Test_localLocker_expireOldLocksExpire(t *testing.T) {
 					for _, v := range l.lockMap {
 						for i := range v {
 							if rng.Intn(2) == 0 {
-								v[i].TimeLastRefresh = expired
+								v[i].TimeLastRefresh = expired.UnixNano()
 							}
 						}
 					}
@@ -347,6 +351,7 @@ func Test_localLocker_expireOldLocksExpire(t *testing.T) {
 
 func Test_localLocker_RUnlock(t *testing.T) {
 	rng := rand.New(rand.NewSource(0))
+	quorum := 0
 	// Numbers of unique locks
 	for _, locks := range []int{1, 100, 1000, 1e6} {
 		if testing.Short() && locks > 100 {
@@ -375,7 +380,7 @@ func Test_localLocker_RUnlock(t *testing.T) {
 								Resources: res,
 								Source:    hex.EncodeToString(tmp[:8]),
 								Owner:     hex.EncodeToString(tmp[8:]),
-								Quorum:    0,
+								Quorum:    &quorum,
 							})
 							if !ok || err != nil {
 								t.Fatal("failed:", err, ok)