remove replication throttle deadline for objects > 128MiB (#20184)

context deadline was introduced to avoid a slow transfer from blocking
replication queue(s) shared by other buckets that may not be under throttling.

This PR removes this context deadline for larger objects since they are 
anyway restricted to a limited set of workers. Otherwise, objects would 
get dequeued when the throttle limit is exceeded and cannot proceed 
within the deadline.

.github/workflows/helm-lint.yml

@@ -22,7 +22,7 @@ jobs:
         uses: actions/checkout@v4
 
       - name: Install Helm
-        uses: azure/setup-helm@v3
+        uses: azure/setup-helm@v4
 
       - name: Run helm lint
         run: |

.github/workflows/vulncheck.yml

@@ -21,8 +21,7 @@ jobs:
       - name: Set up Go
         uses: actions/setup-go@v5
         with:
-          go-version: 1.22.3
-          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]

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"
@@ -47,7 +50,7 @@ lint-fix: getdeps ## runs golangci-lint suite of linters with automatic fixes
 check: test
 test: verifiers build ## builds minio, runs linters, tests
 	@echo "Running unit tests"
-	@MINIO_API_REQUESTS_MAX=10000 CGO_ENABLED=0 go test -v -tags kqueue ./...
+	@MINIO_API_REQUESTS_MAX=10000 CGO_ENABLED=0 go test -v -tags kqueue,dev ./...
 
 test-root-disable: install-race
 	@echo "Running minio root lockdown tests"
@@ -84,13 +87,13 @@ test-race: verifiers build ## builds minio, runs linters, tests (race)
 	@echo "Running unit tests under -race"
 	@(env bash $(PWD)/buildscripts/race.sh)
 
-test-iam: build ## verify IAM (external IDP, etcd backends)
+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 -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 -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: build ## verify IAM (external LDAP IDP)
+test-iam-ldap-upgrade-import: install-race ## verify IAM (external LDAP IDP)
 	@echo "Running upgrade tests for IAM (LDAP backend)"
 	@env bash $(PWD)/buildscripts/minio-iam-ldap-upgrade-import-test.sh
 
@@ -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)
@@ -196,7 +199,7 @@ docker: build ## builds minio docker container
 
 install-race: checks build-debugging ## builds minio to $(PWD)
 	@echo "Building minio binary with -race to './minio'"
-	@GORACE=history_size=7 CGO_ENABLED=1 go build -tags kqueue -race -trimpath --ldflags "$(LDFLAGS)" -o $(PWD)/minio 1>/dev/null
+	@GORACE=history_size=7 CGO_ENABLED=1 go build -tags kqueue,dev -race -trimpath --ldflags "$(LDFLAGS)" -o $(PWD)/minio 1>/dev/null
 	@echo "Installing minio binary with -race to '$(GOPATH)/bin/minio'"
 	@mkdir -p $(GOPATH)/bin && cp -af $(PWD)/minio $(GOPATH)/bin/minio
 

buildscripts/minio-upgrade.sh

@@ -4,10 +4,22 @@ trap 'cleanup $LINENO' ERR
 
 # shellcheck disable=SC2120
 cleanup() {
-	MINIO_VERSION=dev docker-compose \
+	MINIO_VERSION=dev /tmp/gopath/bin/docker-compose \
 		-f "buildscripts/upgrade-tests/compose.yml" \
-		rm -s -f
+		down || true
+
+	MINIO_VERSION=dev /tmp/gopath/bin/docker-compose \
+		-f "buildscripts/upgrade-tests/compose.yml" \
+		rm || true
+
+	for volume in $(docker volume ls -q | grep upgrade); do
+		docker volume rm ${volume} || true
+	done
+
 	docker volume prune -f
+	docker system prune -f || true
+	docker volume prune -f || true
+	docker volume rm $(docker volume ls -q -f dangling=true) || true
 }
 
 verify_checksum_after_heal() {
@@ -60,6 +72,8 @@ __init__() {
 	go install github.com/docker/compose/v2/cmd@latest
 	mv -v /tmp/gopath/bin/cmd /tmp/gopath/bin/docker-compose
 
+	cleanup
+
 	TAG=minio/minio:dev make docker
 
 	MINIO_VERSION=RELEASE.2019-12-19T22-52-26Z docker-compose \
@@ -77,11 +91,11 @@ __init__() {
 
 	curl -s http://127.0.0.1:9000/minio-test/to-read/hosts | sha256sum
 
-	MINIO_VERSION=dev docker-compose -f "buildscripts/upgrade-tests/compose.yml" stop
+	MINIO_VERSION=dev /tmp/gopath/bin/docker-compose -f "buildscripts/upgrade-tests/compose.yml" stop
 }
 
 main() {
-	MINIO_VERSION=dev docker-compose -f "buildscripts/upgrade-tests/compose.yml" up -d --build
+	MINIO_VERSION=dev /tmp/gopath/bin/docker-compose -f "buildscripts/upgrade-tests/compose.yml" up -d --build
 
 	add_alias
 

buildscripts/upgrade-tests/compose.yml

@@ -1,5 +1,3 @@
-version: '3.7'
-
 # Settings and configurations that are common for all containers
 x-minio-common: &minio-common
   image: minio/minio:${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-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

@@ -2308,7 +2308,7 @@ 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,
@@ -2182,7 +2187,7 @@ 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()
 
@@ -2417,7 +2422,7 @@ func getServerInfo(ctx context.Context, pools, metrics bool, r *http.Request) ma
 		Mode:          string(mode),
 		Domain:        domain,
 		Region:        globalSite.Region(),
-		SQSARN:        globalEventNotifier.GetARNList(false),
+		SQSARN:        globalEventNotifier.GetARNList(),
 		DeploymentID:  globalDeploymentID(),
 		Buckets:       buckets,
 		Objects:       objects,

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-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

@@ -445,6 +445,8 @@ const (
 	ErrAdminNoAccessKey
 	ErrAdminNoSecretKey
 
+	ErrIAMNotInitialized
+
 	apiErrCodeEnd // This is used only for the testing code
 )
 
@@ -967,7 +969,7 @@ var errorCodes = errorCodeMap{
 	ErrReplicationRemoteConnectionError: {
 		Code:           "XMinioAdminReplicationRemoteConnectionError",
 		Description:    "Remote service connection error",
-		HTTPStatusCode: http.StatusNotFound,
+		HTTPStatusCode: http.StatusServiceUnavailable,
 	},
 	ErrReplicationBandwidthLimitError: {
 		Code:           "XMinioAdminReplicationBandwidthLimitError",
@@ -1305,6 +1307,11 @@ var errorCodes = errorCodeMap{
 		Description:    "Server not initialized yet, please try again.",
 		HTTPStatusCode: http.StatusServiceUnavailable,
 	},
+	ErrIAMNotInitialized: {
+		Code:           "XMinioIAMNotInitialized",
+		Description:    "IAM sub-system not initialized yet, please try again.",
+		HTTPStatusCode: http.StatusServiceUnavailable,
+	},
 	ErrBucketMetadataNotInitialized: {
 		Code:           "XMinioBucketMetadataNotInitialized",
 		Description:    "Bucket metadata not initialized yet, please try again.",

cmd/api-response.go

@@ -789,8 +789,8 @@ func generateInitiateMultipartUploadResponse(bucket, key, uploadID string) Initi
 }
 
 // generates CompleteMultipartUploadResponse for given bucket, key, location and ETag.
-func generateCompleteMultipartUploadResponse(bucket, key, location string, oi ObjectInfo) CompleteMultipartUploadResponse {
-	cs := oi.decryptChecksums(0)
+func generateCompleteMultipartUploadResponse(bucket, key, location string, oi ObjectInfo, h http.Header) CompleteMultipartUploadResponse {
+	cs := oi.decryptChecksums(0, h)
 	c := CompleteMultipartUploadResponse{
 		Location: location,
 		Bucket:   bucket,
@@ -946,10 +946,20 @@ 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:
+		// 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")
+	case http.StatusTooManyRequests:
+		_, deadline := globalAPIConfig.getRequestsPool()
+		if deadline <= 0 {
+			// Set retry-after header to indicate user-agents to retry request after 10 seconds.
+			// https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Retry-After
+			w.Header().Set(xhttp.RetryAfter, "10")
+		} else {
+			w.Header().Set(xhttp.RetryAfter, strconv.Itoa(int(deadline.Seconds())))
+		}
 	}
 
 	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).
@@ -615,7 +615,7 @@ 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

@@ -88,6 +88,8 @@ type healingTracker struct {
 
 	ItemsSkipped uint64
 	BytesSkipped uint64
+
+	RetryAttempts uint64
 	// Add future tracking capabilities
 	// Be sure that they are included in toHealingDisk
 }
@@ -360,10 +362,10 @@ func initAutoHeal(ctx context.Context, objAPI ObjectLayer) {
 
 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
@@ -382,6 +384,8 @@ 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 +393,17 @@ 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 +466,27 @@ 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++
+		bugLogIf(ctx, tracker.update(ctx))
 
+		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)
+		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-1, 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")
@@ -524,7 +558,7 @@ 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

@@ -200,6 +200,12 @@ 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
+			}
 		default:
 			err = dc.Skip()
 			if err != nil {
@@ -213,9 +219,9 @@ 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 26
 	// write "ID"
-	err = en.Append(0xde, 0x0, 0x19, 0xa2, 0x49, 0x44)
+	err = en.Append(0xde, 0x0, 0x1a, 0xa2, 0x49, 0x44)
 	if err != nil {
 		return
 	}
@@ -478,15 +484,25 @@ 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
+	}
 	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 26
 	// string "ID"
-	o = append(o, 0xde, 0x0, 0x19, 0xa2, 0x49, 0x44)
+	o = append(o, 0xde, 0x0, 0x1a, 0xa2, 0x49, 0x44)
 	o = msgp.AppendString(o, z.ID)
 	// string "PoolIndex"
 	o = append(o, 0xa9, 0x50, 0x6f, 0x6f, 0x6c, 0x49, 0x6e, 0x64, 0x65, 0x78)
@@ -566,6 +582,9 @@ 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)
 	return
 }
 
@@ -763,6 +782,12 @@ 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
+			}
 		default:
 			bts, err = msgp.Skip(bts)
 			if err != nil {
@@ -785,6 +810,6 @@ 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
 	return
 }

cmd/batch-expire.go

@@ -36,6 +36,7 @@ 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 +117,7 @@ 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 +163,7 @@ 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
 	}
 
@@ -432,7 +433,7 @@ func batchObjsForDelete(ctx context.Context, r *BatchJobExpire, ri *batchJobInfo
 					})
 					if err != nil {
 						stopFn(exp, err)
-						batchLogIf(ctx, fmt.Errorf("Failed to expire %s/%s versionID=%s due to %v (attempts=%d)", toExpire[i].Bucket, toExpire[i].Name, toExpire[i].VersionID, err, attempts))
+						batchLogIf(ctx, fmt.Errorf("Failed to expire %s/%s due to %v (attempts=%d)", exp.Bucket, exp.Name, err, attempts))
 					} else {
 						stopFn(exp, err)
 						success = true
@@ -464,13 +465,13 @@ func batchObjsForDelete(ctx context.Context, r *BatchJobExpire, ri *batchJobInfo
 				copy(toDelCopy, toDel)
 				var failed int
 				errs := r.Expire(ctx, api, vc, toDel)
-				// reslice toDel in preparation for next retry
-				// attempt
+				// reslice toDel in preparation for next retry attempt
 				toDel = toDel[:0]
 				for i, err := range errs {
 					if err != nil {
 						stopFn(toDelCopy[i], err)
-						batchLogIf(ctx, fmt.Errorf("Failed to expire %s/%s versionID=%s due to %v (attempts=%d)", ri.Bucket, toDelCopy[i].ObjectName, toDelCopy[i].VersionID, err, attempts))
+						batchLogIf(ctx, fmt.Errorf("Failed to expire %s/%s versionID=%s due to %v (attempts=%d)", ri.Bucket, toDelCopy[i].ObjectName, toDelCopy[i].VersionID,
+							err, attempts))
 						failed++
 						if oi, ok := oiCache.Get(toDelCopy[i]); ok {
 							ri.trackCurrentBucketObject(r.Bucket, *oi, false, attempts)
@@ -514,7 +515,7 @@ 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
 	}
 
@@ -552,22 +553,25 @@ func (r *BatchJobExpire) Start(ctx context.Context, api ObjectLayer, job BatchJo
 	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 +588,7 @@ 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

cmd/batch-expire_gen.go

@@ -306,7 +306,7 @@ 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 +433,7 @@ 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 +544,11 @@ 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 +617,7 @@ 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 +738,7 @@ 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

@@ -20,7 +20,7 @@ package cmd
 import (
 	"testing"
 
-	"gopkg.in/yaml.v2"
+	"gopkg.in/yaml.v3"
 )
 
 func TestParseBatchJobExpire(t *testing.T) {
@@ -32,7 +32,7 @@ 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
@@ -64,7 +64,7 @@ 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)
 	}

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
 		}
@@ -534,7 +540,7 @@ 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 +562,10 @@ 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 +729,85 @@ 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 +951,12 @@ 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 +981,10 @@ 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 +1010,7 @@ 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 +1020,18 @@ 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
 		}
@@ -1071,86 +1110,84 @@ 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)
+		)
 
+		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, cancel := context.WithCancel(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
@@ -1171,7 +1208,7 @@ 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 +1235,7 @@ 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 +1243,7 @@ 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)
@@ -1221,10 +1258,10 @@ func (r *BatchJobReplicateV1) Start(ctx context.Context, api ObjectLayer, job Ba
 
 		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()
@@ -1436,10 +1473,24 @@ 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
@@ -1476,7 +1527,7 @@ func (j *BatchJobRequest) load(ctx context.Context, api ObjectLayer, name string
 
 func batchReplicationOpts(ctx context.Context, sc string, objInfo ObjectInfo) (putOpts miniogo.PutObjectOptions, err error) {
 	// TODO: support custom storage class for remote replication
-	putOpts, err = putReplicationOpts(ctx, "", objInfo)
+	putOpts, err = putReplicationOpts(ctx, "", objInfo, 0)
 	if err != nil {
 		return putOpts, err
 	}
@@ -1520,6 +1571,9 @@ 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 +1596,55 @@ 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
@@ -1633,7 +1736,7 @@ func (a adminAPIHandlers) StartBatchJob(w http.ResponseWriter, r *http.Request)
 		return
 	}
 
-	job.ID = fmt.Sprintf("%s:%d", shortuuid.New(), 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 +1824,60 @@ 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 +1890,9 @@ 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 +2143,7 @@ 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)
 				}
 			}

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

cmd/batch-replicate_gen.go

@@ -1409,13 +1409,13 @@ 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 +1489,7 @@ 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 +1499,7 @@ 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 +1567,18 @@ 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 +1627,13 @@ 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 +1702,7 @@ 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 +1,100 @@
+// 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 (
+	"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)
+	}
+}

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/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,

cmd/bucket-handlers_test.go

@@ -32,7 +32,7 @@ import (
 
 // Wrapper for calling RemoveBucket HTTP handler tests for both Erasure multiple disks and single node setup.
 func TestRemoveBucketHandler(t *testing.T) {
-	ExecObjectLayerAPITest(t, testRemoveBucketHandler, []string{"RemoveBucket"})
+	ExecObjectLayerAPITest(ExecObjectLayerAPITestArgs{t: t, objAPITest: testRemoveBucketHandler, endpoints: []string{"RemoveBucket"}})
 }
 
 func testRemoveBucketHandler(obj ObjectLayer, instanceType, bucketName string, apiRouter http.Handler,
@@ -78,7 +78,7 @@ func testRemoveBucketHandler(obj ObjectLayer, instanceType, bucketName string, a
 
 // Wrapper for calling GetBucketPolicy HTTP handler tests for both Erasure multiple disks and single node setup.
 func TestGetBucketLocationHandler(t *testing.T) {
-	ExecObjectLayerAPITest(t, testGetBucketLocationHandler, []string{"GetBucketLocation"})
+	ExecObjectLayerAPITest(ExecObjectLayerAPITestArgs{t: t, objAPITest: testGetBucketLocationHandler, endpoints: []string{"GetBucketLocation"}})
 }
 
 func testGetBucketLocationHandler(obj ObjectLayer, instanceType, bucketName string, apiRouter http.Handler,
@@ -220,7 +220,7 @@ func testGetBucketLocationHandler(obj ObjectLayer, instanceType, bucketName stri
 
 // Wrapper for calling HeadBucket HTTP handler tests for both Erasure multiple disks and single node setup.
 func TestHeadBucketHandler(t *testing.T) {
-	ExecObjectLayerAPITest(t, testHeadBucketHandler, []string{"HeadBucket"})
+	ExecObjectLayerAPITest(ExecObjectLayerAPITestArgs{t: t, objAPITest: testHeadBucketHandler, endpoints: []string{"HeadBucket"}})
 }
 
 func testHeadBucketHandler(obj ObjectLayer, instanceType, bucketName string, apiRouter http.Handler,
@@ -322,7 +322,7 @@ func testHeadBucketHandler(obj ObjectLayer, instanceType, bucketName string, api
 
 // Wrapper for calling TestListMultipartUploadsHandler tests for both Erasure multiple disks and single node setup.
 func TestListMultipartUploadsHandler(t *testing.T) {
-	ExecObjectLayerAPITest(t, testListMultipartUploadsHandler, []string{"ListMultipartUploads"})
+	ExecObjectLayerAPITest(ExecObjectLayerAPITestArgs{t: t, objAPITest: testListMultipartUploadsHandler, endpoints: []string{"ListMultipartUploads"}})
 }
 
 // testListMultipartUploadsHandler - Tests validate listing of multipart uploads.
@@ -558,7 +558,7 @@ func testListMultipartUploadsHandler(obj ObjectLayer, instanceType, bucketName s
 
 // Wrapper for calling TestListBucketsHandler tests for both Erasure multiple disks and single node setup.
 func TestListBucketsHandler(t *testing.T) {
-	ExecObjectLayerAPITest(t, testListBucketsHandler, []string{"ListBuckets"})
+	ExecObjectLayerAPITest(ExecObjectLayerAPITestArgs{t: t, objAPITest: testListBucketsHandler, endpoints: []string{"ListBuckets"}})
 }
 
 // testListBucketsHandler - Tests validate listing of buckets.
@@ -649,7 +649,7 @@ func testListBucketsHandler(obj ObjectLayer, instanceType, bucketName string, ap
 
 // Wrapper for calling DeleteMultipleObjects HTTP handler tests for both Erasure multiple disks and single node setup.
 func TestAPIDeleteMultipleObjectsHandler(t *testing.T) {
-	ExecObjectLayerAPITest(t, testAPIDeleteMultipleObjectsHandler, []string{"DeleteMultipleObjects", "PutBucketPolicy"})
+	ExecObjectLayerAPITest(ExecObjectLayerAPITestArgs{t: t, objAPITest: testAPIDeleteMultipleObjectsHandler, endpoints: []string{"DeleteMultipleObjects", "PutBucketPolicy"}})
 }
 
 func testAPIDeleteMultipleObjectsHandler(obj ObjectLayer, instanceType, bucketName string, apiRouter http.Handler,

cmd/bucket-lifecycle-handlers_test.go

@@ -29,7 +29,7 @@ import (
 
 // Test S3 Bucket lifecycle APIs with wrong credentials
 func TestBucketLifecycleWrongCredentials(t *testing.T) {
-	ExecObjectLayerAPITest(t, testBucketLifecycleHandlersWrongCredentials, []string{"GetBucketLifecycle", "PutBucketLifecycle", "DeleteBucketLifecycle"})
+	ExecObjectLayerAPITest(ExecObjectLayerAPITestArgs{t: t, objAPITest: testBucketLifecycleHandlersWrongCredentials, endpoints: []string{"GetBucketLifecycle", "PutBucketLifecycle", "DeleteBucketLifecycle"}})
 }
 
 // Test for authentication
@@ -145,7 +145,7 @@ func testBucketLifecycleHandlersWrongCredentials(obj ObjectLayer, instanceType,
 
 // Test S3 Bucket lifecycle APIs
 func TestBucketLifecycle(t *testing.T) {
-	ExecObjectLayerAPITest(t, testBucketLifecycleHandlers, []string{"GetBucketLifecycle", "PutBucketLifecycle", "DeleteBucketLifecycle"})
+	ExecObjectLayerAPITest(ExecObjectLayerAPITestArgs{t: t, objAPITest: testBucketLifecycleHandlers, endpoints: []string{"GetBucketLifecycle", "PutBucketLifecycle", "DeleteBucketLifecycle"}})
 }
 
 // Simple tests of bucket lifecycle: PUT, GET, DELETE.

cmd/bucket-lifecycle.go

@@ -583,6 +583,10 @@ func enqueueTransitionImmediate(obj ObjectInfo, src lcEventSrc) {
 	if lc, err := globalLifecycleSys.Get(obj.Bucket); err == nil {
 		switch event := lc.Eval(obj.ToLifecycleOpts()); event.Action {
 		case lifecycle.TransitionAction, lifecycle.TransitionVersionAction:
+			if obj.DeleteMarker || obj.IsDir {
+				// nothing to transition
+				return
+			}
 			globalTransitionState.queueTransitionTask(obj, event, src)
 		}
 	}
@@ -737,7 +741,7 @@ func getTransitionedObjectReader(ctx context.Context, bucket, object string, rs
 		return nil, fmt.Errorf("transition storage class not configured: %w", err)
 	}
 
-	fn, off, length, err := NewGetObjectReader(rs, oi, opts)
+	fn, off, length, err := NewGetObjectReader(rs, oi, opts, h)
 	if err != nil {
 		return nil, ErrorRespToObjectError(err, bucket, object)
 	}

cmd/bucket-listobjects-handlers.go

@@ -202,7 +202,7 @@ func (api objectAPIHandlers) listObjectsV2Handler(ctx context.Context, w http.Re
 
 	if r.Header.Get(xMinIOExtract) == "true" && strings.Contains(prefix, archivePattern) {
 		// Initiate a list objects operation inside a zip file based in the input params
-		listObjectsV2Info, err = listObjectsV2InArchive(ctx, objectAPI, bucket, prefix, token, delimiter, maxKeys, fetchOwner, startAfter)
+		listObjectsV2Info, err = listObjectsV2InArchive(ctx, objectAPI, bucket, prefix, token, delimiter, maxKeys, startAfter, r.Header)
 	} else {
 		// Initiate a list objects operation based on the input params.
 		// On success would return back ListObjectsInfo object to be
@@ -231,7 +231,7 @@ func parseRequestToken(token string) (subToken string, nodeIndex int) {
 	if token == "" {
 		return token, -1
 	}
-	i := strings.Index(token, ":")
+	i := strings.Index(token, getKeySeparator())
 	if i < 0 {
 		return token, -1
 	}

cmd/bucket-metadata-sys.go

@@ -392,9 +392,7 @@ 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 +411,7 @@ 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
 }
@@ -471,7 +467,7 @@ 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 +480,7 @@ 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, failedBuckets map[string]struct{}) {
 	g := errgroup.WithNErrs(len(buckets))
 	bucketMetas := make([]BucketMetadata, len(buckets))
 	for index := range buckets {
@@ -494,8 +490,8 @@ 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 +504,7 @@ 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)
 		}
 	}
 
@@ -519,7 +515,7 @@ 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()
 
@@ -528,7 +524,7 @@ func (sys *BucketMetadataSys) concurrentLoad(ctx context.Context, buckets []Buck
 			if failedBuckets == nil {
 				failedBuckets = make(map[string]struct{})
 			}
-			failedBuckets[buckets[i].Name] = struct{}{}
+			failedBuckets[buckets[i]] = struct{}{}
 			continue
 		}
 		globalEventNotifier.set(buckets[i], meta)   // set notification targets
@@ -548,7 +544,7 @@ 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
@@ -579,8 +575,8 @@ func (sys *BucketMetadataSys) refreshBucketsMetadataLoop(ctx context.Context, fa
 
 				// Initialize the failed buckets
 				if _, ok := failedBuckets[buckets[i].Name]; ok {
-					globalEventNotifier.set(buckets[i], meta)
-					globalBucketTargetSys.set(buckets[i], meta)
+					globalEventNotifier.set(buckets[i].Name, meta)
+					globalBucketTargetSys.set(buckets[i].Name, meta)
 					delete(failedBuckets, buckets[i].Name)
 				}
 
@@ -600,8 +596,8 @@ 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.
+func (sys *BucketMetadataSys) init(ctx context.Context, buckets []string) {
+	count := globalEndpoints.ESCount() * 10
 	failedBuckets := make(map[string]struct{})
 	for {
 		if len(buckets) < count {

cmd/bucket-policy-handlers_test.go

@@ -107,7 +107,7 @@ func getAnonWriteOnlyObjectPolicy(bucketName, prefix string) *policy.BucketPolic
 
 // Wrapper for calling Create Bucket and ensure we get one and only one success.
 func TestCreateBucket(t *testing.T) {
-	ExecObjectLayerAPITest(t, testCreateBucket, []string{"MakeBucket"})
+	ExecObjectLayerAPITest(ExecObjectLayerAPITestArgs{t: t, objAPITest: testCreateBucket, endpoints: []string{"MakeBucket"}})
 }
 
 // testCreateBucket - Test for calling Create Bucket and ensure we get one and only one success.
@@ -154,7 +154,7 @@ func testCreateBucket(obj ObjectLayer, instanceType, bucketName string, apiRoute
 
 // Wrapper for calling Put Bucket Policy HTTP handler tests for both Erasure multiple disks and single node setup.
 func TestPutBucketPolicyHandler(t *testing.T) {
-	ExecObjectLayerAPITest(t, testPutBucketPolicyHandler, []string{"PutBucketPolicy"})
+	ExecObjectLayerAPITest(ExecObjectLayerAPITestArgs{t: t, objAPITest: testPutBucketPolicyHandler, endpoints: []string{"PutBucketPolicy"}})
 }
 
 // testPutBucketPolicyHandler - Test for Bucket policy end point.
@@ -373,7 +373,7 @@ func testPutBucketPolicyHandler(obj ObjectLayer, instanceType, bucketName string
 
 // Wrapper for calling Get Bucket Policy HTTP handler tests for both Erasure multiple disks and single node setup.
 func TestGetBucketPolicyHandler(t *testing.T) {
-	ExecObjectLayerAPITest(t, testGetBucketPolicyHandler, []string{"PutBucketPolicy", "GetBucketPolicy"})
+	ExecObjectLayerAPITest(ExecObjectLayerAPITestArgs{t: t, objAPITest: testGetBucketPolicyHandler, endpoints: []string{"PutBucketPolicy", "GetBucketPolicy"}})
 }
 
 // testGetBucketPolicyHandler - Test for end point which fetches the access policy json of the given bucket.
@@ -577,7 +577,7 @@ func testGetBucketPolicyHandler(obj ObjectLayer, instanceType, bucketName string
 
 // Wrapper for calling Delete Bucket Policy HTTP handler tests for both Erasure multiple disks and single node setup.
 func TestDeleteBucketPolicyHandler(t *testing.T) {
-	ExecObjectLayerAPITest(t, testDeleteBucketPolicyHandler, []string{"PutBucketPolicy", "DeleteBucketPolicy"})
+	ExecObjectLayerAPITest(ExecObjectLayerAPITestArgs{t: t, objAPITest: testDeleteBucketPolicyHandler, endpoints: []string{"PutBucketPolicy", "DeleteBucketPolicy"}})
 }
 
 // testDeleteBucketPolicyHandler - Test for Delete bucket policy end point.

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
@@ -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-utils.go

@@ -534,7 +534,7 @@ func getHealReplicateObjectInfo(oi ObjectInfo, rcfg replicationConfig) Replicate
 	rstate.ReplicateDecisionStr = dsc.String()
 	asz, _ := oi.GetActualSize()
 
-	return ReplicateObjectInfo{
+	r := ReplicateObjectInfo{
 		Name:                       oi.Name,
 		Size:                       oi.Size,
 		ActualSize:                 asz,
@@ -558,6 +558,10 @@ func getHealReplicateObjectInfo(oi ObjectInfo, rcfg replicationConfig) Replicate
 		SSEC:                 crypto.SSEC.IsEncrypted(oi.UserDefined),
 		UserTags:             oi.UserTags,
 	}
+	if r.SSEC {
+		r.Checksum = oi.Checksum
+	}
+	return r
 }
 
 // ReplicationState - returns replication state using other internal replication metadata in ObjectInfo

cmd/bucket-replication.go

@@ -19,6 +19,7 @@ package cmd
 
 import (
 	"context"
+	"encoding/base64"
 	"encoding/binary"
 	"errors"
 	"fmt"
@@ -74,8 +75,9 @@ const (
 	ObjectLockRetentionTimestamp = "objectlock-retention-timestamp"
 	// ObjectLockLegalHoldTimestamp - the last time a legal hold metadata modification happened on this cluster for this object version
 	ObjectLockLegalHoldTimestamp = "objectlock-legalhold-timestamp"
-	// ReplicationWorkerMultiplier is suggested worker multiplier if traffic exceeds replication worker capacity
-	ReplicationWorkerMultiplier = 1.5
+
+	// ReplicationSsecChecksumHeader - the encrypted checksum of the SSE-C encrypted object.
+	ReplicationSsecChecksumHeader = "X-Minio-Replication-Ssec-Crc"
 )
 
 // gets replication config associated to a given bucket name.
@@ -89,7 +91,10 @@ 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)
@@ -111,7 +116,7 @@ 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)
@@ -131,23 +136,29 @@ 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
 		}
 	}
 
@@ -763,11 +774,13 @@ func (m caseInsensitiveMap) Lookup(key string) (string, bool) {
 	return "", false
 }
 
-func putReplicationOpts(ctx context.Context, sc string, objInfo ObjectInfo) (putOpts minio.PutObjectOptions, err error) {
+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)
+
 	for k, v := range objInfo.UserDefined {
 		// In case of SSE-C objects copy the allowed internal headers as well
-		if !crypto.SSEC.IsEncrypted(objInfo.UserDefined) || !slices.Contains(maps.Keys(validSSEReplicationHeaders), k) {
+		if !isSSEC || !slices.Contains(maps.Keys(validSSEReplicationHeaders), k) {
 			if stringsHasPrefixFold(k, ReservedMetadataPrefixLower) {
 				continue
 			}
@@ -782,6 +795,21 @@ func putReplicationOpts(ctx context.Context, sc string, objInfo ObjectInfo) (put
 		}
 	}
 
+	if len(objInfo.Checksum) > 0 {
+		// Add encrypted CRC to metadata for SSE-C objects.
+		if isSSEC {
+			meta[ReplicationSsecChecksumHeader] = base64.StdEncoding.EncodeToString(objInfo.Checksum)
+		} else {
+			for _, pi := range objInfo.Parts {
+				if pi.Number == partNum {
+					for k, v := range pi.Checksums {
+						meta[k] = v
+					}
+				}
+			}
+		}
+	}
+
 	if sc == "" && (objInfo.StorageClass == storageclass.STANDARD || objInfo.StorageClass == storageclass.RRS) {
 		sc = objInfo.StorageClass
 	}
@@ -1201,17 +1229,23 @@ func (ri ReplicateObjectInfo) replicateObject(ctx context.Context, objectAPI Obj
 	// make sure we have the latest metadata for metrics calculation
 	rinfo.PrevReplicationStatus = objInfo.TargetReplicationStatus(tgt.ARN)
 
-	size, err := objInfo.GetActualSize()
-	if err != nil {
-		replLogIf(ctx, err)
-		sendEvent(eventArgs{
-			EventName:  event.ObjectReplicationNotTracked,
-			BucketName: bucket,
-			Object:     objInfo,
-			UserAgent:  "Internal: [Replication]",
-			Host:       globalLocalNodeName,
-		})
-		return
+	// Set the encrypted size for SSE-C objects
+	var size int64
+	if crypto.SSEC.IsEncrypted(objInfo.UserDefined) {
+		size = objInfo.Size
+	} else {
+		size, err = objInfo.GetActualSize()
+		if err != nil {
+			replLogIf(ctx, err)
+			sendEvent(eventArgs{
+				EventName:  event.ObjectReplicationNotTracked,
+				BucketName: bucket,
+				Object:     objInfo,
+				UserAgent:  "Internal: [Replication]",
+				Host:       globalLocalNodeName,
+			})
+			return
+		}
 	}
 
 	if tgt.Bucket == "" {
@@ -1239,7 +1273,7 @@ func (ri ReplicateObjectInfo) replicateObject(ctx context.Context, objectAPI Obj
 	// use core client to avoid doing multipart on PUT
 	c := &minio.Core{Client: tgt.Client}
 
-	putOpts, err := putReplicationOpts(ctx, tgt.StorageClass, objInfo)
+	putOpts, err := putReplicationOpts(ctx, tgt.StorageClass, objInfo, 0)
 	if err != nil {
 		replLogIf(ctx, fmt.Errorf("failure setting options for replication bucket:%s err:%w", bucket, err))
 		sendEvent(eventArgs{
@@ -1265,7 +1299,7 @@ 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()
@@ -1377,7 +1411,8 @@ func (ri ReplicateObjectInfo) replicateAll(ctx context.Context, objectAPI Object
 	}
 
 	// Set the encrypted size for SSE-C objects
-	if crypto.SSEC.IsEncrypted(objInfo.UserDefined) {
+	isSSEC := crypto.SSEC.IsEncrypted(objInfo.UserDefined)
+	if isSSEC {
 		size = objInfo.Size
 	}
 
@@ -1437,6 +1472,13 @@ func (ri ReplicateObjectInfo) replicateAll(ctx context.Context, objectAPI Object
 			return
 		}
 	} else {
+		// SSEC objects will refuse HeadObject without the decryption key.
+		// Ignore the error, since we know the object exists and versioning prevents overwriting existing versions.
+		if isSSEC && strings.Contains(cerr.Error(), errorCodes[ErrSSEEncryptedObject].Description) {
+			rinfo.ReplicationStatus = replication.Completed
+			rinfo.ReplicationAction = replicateNone
+			goto applyAction
+		}
 		// if target returns error other than NoSuchKey, defer replication attempt
 		if minio.IsNetworkOrHostDown(cerr, true) && !globalBucketTargetSys.isOffline(tgt.EndpointURL()) {
 			globalBucketTargetSys.markOffline(tgt.EndpointURL())
@@ -1464,6 +1506,7 @@ func (ri ReplicateObjectInfo) replicateAll(ctx context.Context, objectAPI Object
 			return
 		}
 	}
+applyAction:
 	rinfo.ReplicationStatus = replication.Completed
 	rinfo.Size = size
 	rinfo.ReplicationAction = rAction
@@ -1506,7 +1549,7 @@ func (ri ReplicateObjectInfo) replicateAll(ctx context.Context, objectAPI Object
 		}
 	} else {
 		var putOpts minio.PutObjectOptions
-		putOpts, err = putReplicationOpts(ctx, tgt.StorageClass, objInfo)
+		putOpts, err = putReplicationOpts(ctx, tgt.StorageClass, objInfo, 0)
 		if err != nil {
 			replLogIf(ctx, fmt.Errorf("failed to set replicate options for object %s/%s(%s) (target %s) err:%w", bucket, objInfo.Name, objInfo.VersionID, tgt.EndpointURL(), err))
 			sendEvent(eventArgs{
@@ -1531,7 +1574,7 @@ func (ri ReplicateObjectInfo) replicateAll(ctx context.Context, objectAPI Object
 			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()
@@ -1597,13 +1640,14 @@ func replicateObjectWithMultipart(ctx context.Context, c *minio.Core, bucket, ob
 	}()
 
 	var (
-		hr    *hash.Reader
-		pInfo minio.ObjectPart
+		hr     *hash.Reader
+		pInfo  minio.ObjectPart
+		isSSEC = crypto.SSEC.IsEncrypted(objInfo.UserDefined)
 	)
 
 	var objectSize int64
 	for _, partInfo := range objInfo.Parts {
-		if crypto.SSEC.IsEncrypted(objInfo.UserDefined) {
+		if isSSEC {
 			hr, err = hash.NewReader(ctx, io.LimitReader(r, partInfo.Size), partInfo.Size, "", "", partInfo.ActualSize)
 		} else {
 			hr, err = hash.NewReader(ctx, io.LimitReader(r, partInfo.ActualSize), partInfo.ActualSize, "", "", partInfo.ActualSize)
@@ -1614,12 +1658,17 @@ func replicateObjectWithMultipart(ctx context.Context, c *minio.Core, bucket, ob
 
 		cHeader := http.Header{}
 		cHeader.Add(xhttp.MinIOSourceReplicationRequest, "true")
+		if !isSSEC {
+			for k, v := range partInfo.Checksums {
+				cHeader.Add(k, v)
+			}
+		}
 		popts := minio.PutObjectPartOptions{
 			SSE:          opts.ServerSideEncryption,
 			CustomHeader: cHeader,
 		}
 
-		if crypto.SSEC.IsEncrypted(objInfo.UserDefined) {
+		if isSSEC {
 			objectSize += partInfo.Size
 			pInfo, err = c.PutObjectPart(ctx, bucket, object, uploadID, partInfo.Number, hr, partInfo.Size, popts)
 		} else {
@@ -1629,20 +1678,30 @@ func replicateObjectWithMultipart(ctx context.Context, c *minio.Core, bucket, ob
 		if err != nil {
 			return err
 		}
-		if !crypto.SSEC.IsEncrypted(objInfo.UserDefined) && pInfo.Size != partInfo.ActualSize {
+		if !isSSEC && pInfo.Size != partInfo.ActualSize {
 			return fmt.Errorf("Part size mismatch: got %d, want %d", pInfo.Size, partInfo.ActualSize)
 		}
 		uploadedParts = append(uploadedParts, minio.CompletePart{
-			PartNumber: pInfo.PartNumber,
-			ETag:       pInfo.ETag,
+			PartNumber:     pInfo.PartNumber,
+			ETag:           pInfo.ETag,
+			ChecksumCRC32:  pInfo.ChecksumCRC32,
+			ChecksumCRC32C: pInfo.ChecksumCRC32C,
+			ChecksumSHA1:   pInfo.ChecksumSHA1,
+			ChecksumSHA256: pInfo.ChecksumSHA256,
 		})
 	}
+	userMeta := map[string]string{
+		validSSEReplicationHeaders[ReservedMetadataPrefix+"Actual-Object-Size"]: objInfo.UserDefined[ReservedMetadataPrefix+"actual-size"],
+	}
+	if isSSEC && objInfo.UserDefined[ReplicationSsecChecksumHeader] != "" {
+		userMeta[ReplicationSsecChecksumHeader] = objInfo.UserDefined[ReplicationSsecChecksumHeader]
+	}
 
 	// really big value but its okay on heavily loaded systems. This is just tail end timeout.
 	cctx, ccancel := context.WithTimeout(ctx, 10*time.Minute)
 	defer ccancel()
 	_, err = c.CompleteMultipartUpload(cctx, bucket, object, uploadID, uploadedParts, minio.PutObjectOptions{
-		UserMetadata: map[string]string{validSSEReplicationHeaders[ReservedMetadataPrefix+"Actual-Object-Size"]: objInfo.UserDefined[ReservedMetadataPrefix+"actual-size"]},
+		UserMetadata: userMeta,
 		Internal: minio.AdvancedPutOptions{
 			SourceMTime: objInfo.ModTime,
 			SourceETag:  objInfo.ETag,
@@ -1736,15 +1795,18 @@ var (
 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
+
+	mu       sync.RWMutex
+	mrfMU    sync.Mutex
+	resyncer *replicationResyncer
 
 	// workers:
 	workers    []chan ReplicationWorkerOperation
@@ -1815,9 +1877,13 @@ 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(),
@@ -1827,9 +1893,10 @@ func NewReplicationPool(ctx context.Context, o ObjectLayer, opts replicationPool
 		objLayer:        o,
 		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)
@@ -1908,23 +1975,8 @@ 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():
@@ -1935,11 +1987,23 @@ func (p *ReplicationPool) AddLargeWorker(input <-chan ReplicationWorkerOperation
 			}
 			switch v := oi.(type) {
 			case ReplicateObjectInfo:
+				if opTracker != nil {
+					atomic.AddInt32(opTracker, 1)
+				}
 				globalReplicationStats.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)
+				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")
 			}
@@ -1947,6 +2011,30 @@ 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))
@@ -1957,6 +2045,11 @@ 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
@@ -1982,7 +2075,7 @@ 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 {
@@ -2009,11 +2102,15 @@ 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
@@ -2057,9 +2154,18 @@ 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.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
 	}
@@ -2162,8 +2268,9 @@ 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) {
@@ -2179,12 +2286,12 @@ type proxyResult struct {
 
 // get Reader from replication target if active-active replication is in place and
 // this node returns a 404
-func proxyGetToReplicationTarget(ctx context.Context, bucket, object string, rs *HTTPRangeSpec, _ http.Header, opts ObjectOptions, proxyTargets *madmin.BucketTargets) (gr *GetObjectReader, proxy proxyResult, err error) {
+func proxyGetToReplicationTarget(ctx context.Context, bucket, object string, rs *HTTPRangeSpec, h http.Header, opts ObjectOptions, proxyTargets *madmin.BucketTargets) (gr *GetObjectReader, proxy proxyResult, err error) {
 	tgt, oi, proxy := proxyHeadToRepTarget(ctx, bucket, object, rs, opts, proxyTargets)
 	if !proxy.Proxy {
 		return nil, proxy, nil
 	}
-	fn, _, _, err := NewGetObjectReader(nil, oi, opts)
+	fn, _, _, err := NewGetObjectReader(nil, oi, opts, h)
 	if err != nil {
 		return nil, proxy, err
 	}
@@ -2369,7 +2476,9 @@ func scheduleReplication(ctx context.Context, oi ObjectInfo, o ObjectLayer, dsc
 		SSEC:                 crypto.SSEC.IsEncrypted(oi.UserDefined),
 		UserTags:             oi.UserTags,
 	}
-
+	if ri.SSEC {
+		ri.Checksum = oi.Checksum
+	}
 	if dsc.Synchronous() {
 		replicateObject(ctx, ri, o)
 	} else {
@@ -3017,7 +3126,7 @@ 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
 	}
@@ -3026,7 +3135,7 @@ 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 {
@@ -3044,13 +3153,13 @@ 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 {
@@ -3436,7 +3545,7 @@ func (p *ReplicationPool) persistToDrive(ctx context.Context, v MRFReplicateEntr
 	}
 
 	globalLocalDrivesMu.RLock()
-	localDrives := cloneDrives(globalLocalDrives)
+	localDrives := cloneDrives(globalLocalDrivesMap)
 	globalLocalDrivesMu.RUnlock()
 
 	for _, localDrive := range localDrives {
@@ -3503,7 +3612,7 @@ func (p *ReplicationPool) loadMRF() (mrfRec MRFReplicateEntries, err error) {
 	}
 
 	globalLocalDrivesMu.RLock()
-	localDrives := cloneDrives(globalLocalDrives)
+	localDrives := cloneDrives(globalLocalDrivesMap)
 	globalLocalDrivesMu.RUnlock()
 
 	for _, localDrive := range localDrives {
@@ -3624,3 +3733,12 @@ 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-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) {
@@ -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.
@@ -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++
@@ -1233,7 +1262,7 @@ func evalActionFromLifecycle(ctx context.Context, lc lifecycle.Lifecycle, lr loc
 }
 
 func applyTransitionRule(event lifecycle.Event, src lcEventSrc, obj ObjectInfo) bool {
-	if obj.DeleteMarker {
+	if obj.DeleteMarker || obj.IsDir {
 		return false
 	}
 	globalTransitionState.queueTransitionTask(obj, event, src)
@@ -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 == "" {

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/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
@@ -1025,7 +1030,9 @@ func DecryptObjectInfo(info *ObjectInfo, r *http.Request) (encrypted bool, err e
 	if encrypted {
 		if crypto.SSEC.IsEncrypted(info.UserDefined) {
 			if !(crypto.SSEC.IsRequested(headers) || crypto.SSECopy.IsRequested(headers)) {
-				return encrypted, errEncryptedObject
+				if r.Header.Get(xhttp.MinIOSourceReplicationRequest) != "true" {
+					return encrypted, errEncryptedObject
+				}
 			}
 		}
 
@@ -1077,13 +1084,16 @@ func metadataEncrypter(key crypto.ObjectKey) objectMetaEncryptFn {
 }
 
 // metadataDecrypter reverses metadataEncrypter.
-func (o *ObjectInfo) metadataDecrypter() objectMetaDecryptFn {
+func (o *ObjectInfo) metadataDecrypter(h http.Header) objectMetaDecryptFn {
 	return func(baseKey string, input []byte) ([]byte, error) {
 		if len(input) == 0 {
 			return input, nil
 		}
-
-		key, err := decryptObjectMeta(nil, o.Bucket, o.Name, o.UserDefined)
+		var key []byte
+		if k, err := crypto.SSEC.ParseHTTP(h); err == nil {
+			key = k[:]
+		}
+		key, err := decryptObjectMeta(key, o.Bucket, o.Name, o.UserDefined)
 		if err != nil {
 			return nil, err
 		}
@@ -1095,13 +1105,13 @@ func (o *ObjectInfo) metadataDecrypter() objectMetaDecryptFn {
 
 // decryptChecksums 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() {
+func (o *ObjectInfo) decryptPartsChecksums(h http.Header) {
 	data := o.Checksum
 	if len(data) == 0 {
 		return
 	}
 	if _, encrypted := crypto.IsEncrypted(o.UserDefined); encrypted {
-		decrypted, err := o.metadataDecrypter()("object-checksum", data)
+		decrypted, err := o.metadataDecrypter(h)("object-checksum", data)
 		if err != nil {
 			encLogIf(GlobalContext, err)
 			return
@@ -1157,15 +1167,25 @@ func (o *ObjectInfo) metadataEncryptFn(headers http.Header) (objectMetaEncryptFn
 
 // decryptChecksums 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) decryptChecksums(part int) map[string]string {
+func (o *ObjectInfo) decryptChecksums(part int, h http.Header) map[string]string {
 	data := o.Checksum
 	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()("object-checksum", data)
+		decrypted, err := o.metadataDecrypter(h)("object-checksum", data)
 		if err != nil {
-			encLogIf(GlobalContext, err)
+			if err != crypto.ErrSecretKeyMismatch {
+				encLogIf(GlobalContext, err)
+			}
 			return nil
 		}
 		data = decrypted

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-healing-common.go

@@ -20,6 +20,7 @@ package cmd
 import (
 	"bytes"
 	"context"
+	"slices"
 	"time"
 
 	"github.com/minio/madmin-go/v3"
@@ -253,6 +254,33 @@ func listOnlineDisks(disks []StorageAPI, partsMetadata []FileInfo, errs []error,
 	return onlineDisks, modTime, ""
 }
 
+// Convert verify or check parts returned error to integer representation
+func convPartErrToInt(err error) int {
+	err = unwrapAll(err)
+	switch err {
+	case nil:
+		return checkPartSuccess
+	case errFileNotFound, errFileVersionNotFound:
+		return checkPartFileNotFound
+	case errFileCorrupt:
+		return checkPartFileCorrupt
+	case errVolumeNotFound:
+		return checkPartVolumeNotFound
+	case errDiskNotFound:
+		return checkPartDiskNotFound
+	default:
+		return checkPartUnknown
+	}
+}
+
+func partNeedsHealing(partErrs []int) bool {
+	return slices.IndexFunc(partErrs, func(i int) bool { return i != checkPartSuccess && i != checkPartUnknown }) > -1
+}
+
+func hasPartErr(partErrs []int) bool {
+	return slices.IndexFunc(partErrs, func(i int) bool { return i != checkPartSuccess }) > -1
+}
+
 // disksWithAllParts - This function needs to be called with
 // []StorageAPI returned by listOnlineDisks. Returns,
 //
@@ -262,10 +290,19 @@ func listOnlineDisks(disks []StorageAPI, partsMetadata []FileInfo, errs []error,
 //     a not-found error or a hash-mismatch error.
 func disksWithAllParts(ctx context.Context, onlineDisks []StorageAPI, partsMetadata []FileInfo,
 	errs []error, latestMeta FileInfo, bucket, object string,
-	scanMode madmin.HealScanMode) ([]StorageAPI, []error, time.Time,
-) {
-	availableDisks := make([]StorageAPI, len(onlineDisks))
-	dataErrs := make([]error, len(onlineDisks))
+	scanMode madmin.HealScanMode,
+) (availableDisks []StorageAPI, dataErrsByDisk map[int][]int, dataErrsByPart map[int][]int) {
+	availableDisks = make([]StorageAPI, len(onlineDisks))
+
+	dataErrsByDisk = make(map[int][]int, len(onlineDisks))
+	for i := range onlineDisks {
+		dataErrsByDisk[i] = make([]int, len(latestMeta.Parts))
+	}
+
+	dataErrsByPart = make(map[int][]int, len(latestMeta.Parts))
+	for i := range latestMeta.Parts {
+		dataErrsByPart[i] = make([]int, len(onlineDisks))
+	}
 
 	inconsistent := 0
 	for i, meta := range partsMetadata {
@@ -295,19 +332,21 @@ func disksWithAllParts(ctx context.Context, onlineDisks []StorageAPI, partsMetad
 		erasureDistributionReliable = false
 	}
 
+	metaErrs := make([]error, len(errs))
+
 	for i, onlineDisk := range onlineDisks {
 		if errs[i] != nil {
-			dataErrs[i] = errs[i]
+			metaErrs[i] = errs[i]
 			continue
 		}
 		if onlineDisk == OfflineDisk {
-			dataErrs[i] = errDiskNotFound
+			metaErrs[i] = errDiskNotFound
 			continue
 		}
 
 		meta := partsMetadata[i]
 		if !meta.ModTime.Equal(latestMeta.ModTime) || meta.DataDir != latestMeta.DataDir {
-			dataErrs[i] = errFileCorrupt
+			metaErrs[i] = errFileCorrupt
 			partsMetadata[i] = FileInfo{}
 			continue
 		}
@@ -315,7 +354,7 @@ func disksWithAllParts(ctx context.Context, onlineDisks []StorageAPI, partsMetad
 		if erasureDistributionReliable {
 			if !meta.IsValid() {
 				partsMetadata[i] = FileInfo{}
-				dataErrs[i] = errFileCorrupt
+				metaErrs[i] = errFileCorrupt
 				continue
 			}
 
@@ -325,46 +364,79 @@ func disksWithAllParts(ctx context.Context, onlineDisks []StorageAPI, partsMetad
 					// attempt a fix if possible, assuming other entries
 					// might have the right erasure distribution.
 					partsMetadata[i] = FileInfo{}
-					dataErrs[i] = errFileCorrupt
+					metaErrs[i] = errFileCorrupt
 					continue
 				}
 			}
 		}
+	}
+
+	// Copy meta errors to part errors
+	for i, err := range metaErrs {
+		if err != nil {
+			partErr := convPartErrToInt(err)
+			for p := range latestMeta.Parts {
+				dataErrsByPart[p][i] = partErr
+			}
+		}
+	}
+
+	for i, onlineDisk := range onlineDisks {
+		if metaErrs[i] != nil {
+			continue
+		}
+		meta := partsMetadata[i]
+
+		if meta.Deleted || meta.IsRemote() {
+			continue
+		}
 
 		// Always check data, if we got it.
 		if (len(meta.Data) > 0 || meta.Size == 0) && len(meta.Parts) > 0 {
 			checksumInfo := meta.Erasure.GetChecksumInfo(meta.Parts[0].Number)
-			dataErrs[i] = bitrotVerify(bytes.NewReader(meta.Data),
+			verifyErr := bitrotVerify(bytes.NewReader(meta.Data),
 				int64(len(meta.Data)),
 				meta.Erasure.ShardFileSize(meta.Size),
 				checksumInfo.Algorithm,
 				checksumInfo.Hash, meta.Erasure.ShardSize())
-			if dataErrs[i] == nil {
-				// All parts verified, mark it as all data available.
-				availableDisks[i] = onlineDisk
-			} else {
-				// upon errors just make that disk's fileinfo invalid
-				partsMetadata[i] = FileInfo{}
-			}
+			dataErrsByPart[0][i] = convPartErrToInt(verifyErr)
 			continue
 		}
 
+		var (
+			verifyErr  error
+			verifyResp *CheckPartsResp
+		)
+
 		meta.DataDir = latestMeta.DataDir
 		switch scanMode {
 		case madmin.HealDeepScan:
 			// disk has a valid xl.meta but may not have all the
 			// parts. This is considered an outdated disk, since
 			// it needs healing too.
-			if !meta.Deleted && !meta.IsRemote() {
-				dataErrs[i] = onlineDisk.VerifyFile(ctx, bucket, object, meta)
-			}
-		case madmin.HealNormalScan:
-			if !meta.Deleted && !meta.IsRemote() {
-				dataErrs[i] = onlineDisk.CheckParts(ctx, bucket, object, meta)
-			}
+			verifyResp, verifyErr = onlineDisk.VerifyFile(ctx, bucket, object, meta)
+		default:
+			verifyResp, verifyErr = onlineDisk.CheckParts(ctx, bucket, object, meta)
 		}
 
-		if dataErrs[i] == nil {
+		for p := range latestMeta.Parts {
+			if verifyErr != nil {
+				dataErrsByPart[p][i] = convPartErrToInt(verifyErr)
+			} else {
+				dataErrsByPart[p][i] = verifyResp.Results[p]
+			}
+		}
+	}
+
+	// Build dataErrs by disk from dataErrs by part
+	for part, disks := range dataErrsByPart {
+		for disk := range disks {
+			dataErrsByDisk[disk][part] = dataErrsByPart[part][disk]
+		}
+	}
+
+	for i, onlineDisk := range onlineDisks {
+		if metaErrs[i] == nil && !hasPartErr(dataErrsByDisk[i]) {
 			// All parts verified, mark it as all data available.
 			availableDisks[i] = onlineDisk
 		} else {
@@ -373,5 +445,5 @@ func disksWithAllParts(ctx context.Context, onlineDisks []StorageAPI, partsMetad
 		}
 	}
 
-	return availableDisks, dataErrs, timeSentinel
+	return
 }

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)
@@ -308,9 +313,8 @@ func TestListOnlineDisks(t *testing.T) {
 				t.Fatalf("Expected modTime to be equal to %v but was found to be %v",
 					test.expectedTime, modTime)
 			}
-			availableDisks, newErrs, _ := disksWithAllParts(ctx, onlineDisks, partsMetadata,
+			availableDisks, _, _ := disksWithAllParts(ctx, onlineDisks, partsMetadata,
 				test.errs, fi, bucket, object, madmin.HealDeepScan)
-			test.errs = newErrs
 
 			if test._tamperBackend != noTamper {
 				if tamperedIndex != -1 && availableDisks[tamperedIndex] != nil {
@@ -491,9 +495,8 @@ func TestListOnlineDisksSmallObjects(t *testing.T) {
 					test.expectedTime, modTime)
 			}
 
-			availableDisks, newErrs, _ := disksWithAllParts(ctx, onlineDisks, partsMetadata,
+			availableDisks, _, _ := disksWithAllParts(ctx, onlineDisks, partsMetadata,
 				test.errs, fi, bucket, object, madmin.HealDeepScan)
-			test.errs = newErrs
 
 			if test._tamperBackend != noTamper {
 				if tamperedIndex != -1 && availableDisks[tamperedIndex] != nil {
@@ -554,7 +557,7 @@ func TestDisksWithAllParts(t *testing.T) {
 
 	erasureDisks, _, _ = listOnlineDisks(erasureDisks, partsMetadata, errs, readQuorum)
 
-	filteredDisks, errs, _ := disksWithAllParts(ctx, erasureDisks, partsMetadata,
+	filteredDisks, _, dataErrsPerDisk := disksWithAllParts(ctx, erasureDisks, partsMetadata,
 		errs, fi, bucket, object, madmin.HealDeepScan)
 
 	if len(filteredDisks) != len(erasureDisks) {
@@ -562,8 +565,8 @@ func TestDisksWithAllParts(t *testing.T) {
 	}
 
 	for diskIndex, disk := range filteredDisks {
-		if errs[diskIndex] != nil {
-			t.Errorf("Unexpected error %s", errs[diskIndex])
+		if partNeedsHealing(dataErrsPerDisk[diskIndex]) {
+			t.Errorf("Unexpected error: %v", dataErrsPerDisk[diskIndex])
 		}
 
 		if disk == nil {
@@ -634,7 +637,7 @@ func TestDisksWithAllParts(t *testing.T) {
 	}
 
 	errs = make([]error, len(erasureDisks))
-	filteredDisks, errs, _ = disksWithAllParts(ctx, erasureDisks, partsMetadata,
+	filteredDisks, dataErrsPerDisk, _ = disksWithAllParts(ctx, erasureDisks, partsMetadata,
 		errs, fi, bucket, object, madmin.HealDeepScan)
 
 	if len(filteredDisks) != len(erasureDisks) {
@@ -646,15 +649,15 @@ func TestDisksWithAllParts(t *testing.T) {
 			if disk != nil {
 				t.Errorf("Drive not filtered as expected, drive: %d", diskIndex)
 			}
-			if errs[diskIndex] == nil {
-				t.Errorf("Expected error not received, driveIndex: %d", diskIndex)
+			if !partNeedsHealing(dataErrsPerDisk[diskIndex]) {
+				t.Errorf("Disk expected to be healed, driveIndex: %d", diskIndex)
 			}
 		} else {
 			if disk == nil {
 				t.Errorf("Drive erroneously filtered, driveIndex: %d", diskIndex)
 			}
-			if errs[diskIndex] != nil {
-				t.Errorf("Unexpected error, %s, driveIndex: %d", errs[diskIndex], diskIndex)
+			if partNeedsHealing(dataErrsPerDisk[diskIndex]) {
+				t.Errorf("Disk not expected to be healed, driveIndex: %d", diskIndex)
 			}
 
 		}

cmd/erasure-healing.go

@@ -32,6 +32,7 @@ import (
 	"github.com/minio/minio/internal/grid"
 	"github.com/minio/minio/internal/logger"
 	"github.com/minio/pkg/v3/sync/errgroup"
+	"golang.org/x/exp/slices"
 )
 
 //go:generate stringer -type=healingMetric -trimprefix=healingMetric $GOFILE
@@ -144,36 +145,41 @@ func listAllBuckets(ctx context.Context, storageDisks []StorageAPI, healBuckets
 	return reduceReadQuorumErrs(ctx, g.Wait(), bucketMetadataOpIgnoredErrs, readQuorum)
 }
 
+var errLegacyXLMeta = errors.New("legacy XL meta")
+
+var errOutdatedXLMeta = errors.New("outdated XL meta")
+
+var errPartMissingOrCorrupt = errors.New("part missing or corrupt")
+
 // Only heal on disks where we are sure that healing is needed. We can expand
 // this list as and when we figure out more errors can be added to this list safely.
-func shouldHealObjectOnDisk(erErr, dataErr error, meta FileInfo, latestMeta FileInfo) bool {
-	switch {
-	case errors.Is(erErr, errFileNotFound) || errors.Is(erErr, errFileVersionNotFound):
-		return true
-	case errors.Is(erErr, errFileCorrupt):
-		return true
+func shouldHealObjectOnDisk(erErr error, partsErrs []int, meta FileInfo, latestMeta FileInfo) (bool, error) {
+	if errors.Is(erErr, errFileNotFound) || errors.Is(erErr, errFileVersionNotFound) || errors.Is(erErr, errFileCorrupt) {
+		return true, erErr
 	}
 	if erErr == nil {
 		if meta.XLV1 {
 			// Legacy means heal always
 			// always check first.
-			return true
+			return true, errLegacyXLMeta
+		}
+		if !latestMeta.Equals(meta) {
+			return true, errOutdatedXLMeta
 		}
 		if !meta.Deleted && !meta.IsRemote() {
 			// If xl.meta was read fine but there may be problem with the part.N files.
-			if IsErr(dataErr, []error{
-				errFileNotFound,
-				errFileVersionNotFound,
-				errFileCorrupt,
-			}...) {
-				return true
+			for _, partErr := range partsErrs {
+				if slices.Contains([]int{
+					checkPartFileNotFound,
+					checkPartFileCorrupt,
+				}, partErr) {
+					return true, errPartMissingOrCorrupt
+				}
 			}
 		}
-		if !latestMeta.Equals(meta) {
-			return true
-		}
+		return false, nil
 	}
-	return false
+	return false, erErr
 }
 
 const (
@@ -332,7 +338,7 @@ func (er *erasureObjects) healObject(ctx context.Context, bucket string, object
 	// used here for reconstruction. This is done to ensure that
 	// we do not skip drives that have inconsistent metadata to be
 	// skipped from purging when they are stale.
-	availableDisks, dataErrs, _ := disksWithAllParts(ctx, onlineDisks, partsMetadata,
+	availableDisks, dataErrsByDisk, dataErrsByPart := disksWithAllParts(ctx, onlineDisks, partsMetadata,
 		errs, latestMeta, bucket, object, scanMode)
 
 	var erasure Erasure
@@ -355,15 +361,20 @@ func (er *erasureObjects) healObject(ctx context.Context, bucket string, object
 	// to be healed.
 	outDatedDisks := make([]StorageAPI, len(storageDisks))
 	disksToHealCount := 0
-	for i, v := range availableDisks {
+	for i := range availableDisks {
+		yes, reason := shouldHealObjectOnDisk(errs[i], dataErrsByDisk[i], partsMetadata[i], latestMeta)
+		if yes {
+			outDatedDisks[i] = storageDisks[i]
+			disksToHealCount++
+		}
+
 		driveState := ""
 		switch {
-		case v != nil:
+		case reason == nil:
 			driveState = madmin.DriveStateOk
-		case errors.Is(errs[i], errDiskNotFound), errors.Is(dataErrs[i], errDiskNotFound):
+		case IsErr(reason, errDiskNotFound):
 			driveState = madmin.DriveStateOffline
-		case IsErr(errs[i], errFileNotFound, errFileVersionNotFound, errVolumeNotFound),
-			IsErr(dataErrs[i], errFileNotFound, errFileVersionNotFound, errVolumeNotFound):
+		case IsErr(reason, errFileNotFound, errFileVersionNotFound, errVolumeNotFound, errPartMissingOrCorrupt, errOutdatedXLMeta, errLegacyXLMeta):
 			driveState = madmin.DriveStateMissing
 		default:
 			// all remaining cases imply corrupt data/metadata
@@ -380,12 +391,6 @@ func (er *erasureObjects) healObject(ctx context.Context, bucket string, object
 			Endpoint: storageEndpoints[i].String(),
 			State:    driveState,
 		})
-
-		if shouldHealObjectOnDisk(errs[i], dataErrs[i], partsMetadata[i], latestMeta) {
-			outDatedDisks[i] = storageDisks[i]
-			disksToHealCount++
-			continue
-		}
 	}
 
 	if isAllNotFound(errs) {
@@ -412,7 +417,7 @@ func (er *erasureObjects) healObject(ctx context.Context, bucket string, object
 	if !latestMeta.XLV1 && !latestMeta.Deleted && disksToHealCount > latestMeta.Erasure.ParityBlocks {
 		// Allow for dangling deletes, on versions that have DataDir missing etc.
 		// this would end up restoring the correct readable versions.
-		m, err := er.deleteIfDangling(ctx, bucket, object, partsMetadata, errs, dataErrs, ObjectOptions{
+		m, err := er.deleteIfDangling(ctx, bucket, object, partsMetadata, errs, dataErrsByPart, ObjectOptions{
 			VersionID: versionID,
 		})
 		errs = make([]error, len(errs))
@@ -624,7 +629,7 @@ 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
 			}
 		}
@@ -908,35 +913,52 @@ func isObjectDirDangling(errs []error) (ok bool) {
 	return found < notFound && found > 0
 }
 
+func danglingMetaErrsCount(cerrs []error) (notFoundCount int, nonActionableCount int) {
+	for _, readErr := range cerrs {
+		if readErr == nil {
+			continue
+		}
+		switch {
+		case errors.Is(readErr, errFileNotFound) || errors.Is(readErr, errFileVersionNotFound):
+			notFoundCount++
+		default:
+			// All other errors are non-actionable
+			nonActionableCount++
+		}
+	}
+	return
+}
+
+func danglingPartErrsCount(results []int) (notFoundCount int, nonActionableCount int) {
+	for _, partResult := range results {
+		switch partResult {
+		case checkPartSuccess:
+			continue
+		case checkPartFileNotFound:
+			notFoundCount++
+		default:
+			// All other errors are non-actionable
+			nonActionableCount++
+		}
+	}
+	return
+}
+
 // Object is considered dangling/corrupted if and only
 // if total disks - a combination of corrupted and missing
 // files is lesser than number of data blocks.
-func isObjectDangling(metaArr []FileInfo, errs []error, dataErrs []error) (validMeta FileInfo, ok bool) {
+func isObjectDangling(metaArr []FileInfo, errs []error, dataErrsByPart map[int][]int) (validMeta FileInfo, ok bool) {
 	// We can consider an object data not reliable
 	// when xl.meta is not found in read quorum disks.
 	// or when xl.meta is not readable in read quorum disks.
-	danglingErrsCount := func(cerrs []error) (int, int) {
-		var (
-			notFoundCount      int
-			nonActionableCount int
-		)
-		for _, readErr := range cerrs {
-			if readErr == nil {
-				continue
-			}
-			switch {
-			case errors.Is(readErr, errFileNotFound) || errors.Is(readErr, errFileVersionNotFound):
-				notFoundCount++
-			default:
-				// All other errors are non-actionable
-				nonActionableCount++
-			}
-		}
-		return notFoundCount, nonActionableCount
-	}
+	notFoundMetaErrs, nonActionableMetaErrs := danglingMetaErrsCount(errs)
 
-	notFoundMetaErrs, nonActionableMetaErrs := danglingErrsCount(errs)
-	notFoundPartsErrs, nonActionablePartsErrs := danglingErrsCount(dataErrs)
+	notFoundPartsErrs, nonActionablePartsErrs := 0, 0
+	for _, dataErrs := range dataErrsByPart {
+		if nf, na := danglingPartErrsCount(dataErrs); nf > notFoundPartsErrs {
+			notFoundPartsErrs, nonActionablePartsErrs = nf, na
+		}
+	}
 
 	for _, m := range metaArr {
 		if m.IsValid() {
@@ -948,7 +970,7 @@ func isObjectDangling(metaArr []FileInfo, errs []error, dataErrs []error) (valid
 	if !validMeta.IsValid() {
 		// validMeta is invalid because all xl.meta is missing apparently
 		// we should figure out if dataDirs are also missing > dataBlocks.
-		dataBlocks := (len(dataErrs) + 1) / 2
+		dataBlocks := (len(metaArr) + 1) / 2
 		if notFoundPartsErrs > dataBlocks {
 			// Not using parity to ensure that we do not delete
 			// any valid content, if any is recoverable. But if

cmd/erasure-healing_test.go

@@ -49,7 +49,7 @@ func TestIsObjectDangling(t *testing.T) {
 		name             string
 		metaArr          []FileInfo
 		errs             []error
-		dataErrs         []error
+		dataErrs         map[int][]int
 		expectedMeta     FileInfo
 		expectedDangling bool
 	}{
@@ -165,11 +165,8 @@ func TestIsObjectDangling(t *testing.T) {
 				nil,
 				nil,
 			},
-			dataErrs: []error{
-				errFileCorrupt,
-				errFileNotFound,
-				nil,
-				errFileCorrupt,
+			dataErrs: map[int][]int{
+				0: {checkPartFileCorrupt, checkPartFileNotFound, checkPartSuccess, checkPartFileCorrupt},
 			},
 			expectedMeta:     fi,
 			expectedDangling: false,
@@ -188,11 +185,8 @@ func TestIsObjectDangling(t *testing.T) {
 				errFileNotFound,
 				nil,
 			},
-			dataErrs: []error{
-				errFileNotFound,
-				errFileCorrupt,
-				nil,
-				nil,
+			dataErrs: map[int][]int{
+				0: {checkPartFileNotFound, checkPartFileCorrupt, checkPartSuccess, checkPartSuccess},
 			},
 			expectedMeta:     fi,
 			expectedDangling: false,
@@ -247,15 +241,58 @@ func TestIsObjectDangling(t *testing.T) {
 				nil,
 				nil,
 			},
-			dataErrs: []error{
-				errFileNotFound,
-				errFileNotFound,
-				nil,
-				errFileNotFound,
+			dataErrs: map[int][]int{
+				0: {checkPartFileNotFound, checkPartFileNotFound, checkPartSuccess, checkPartFileNotFound},
 			},
 			expectedMeta:     fi,
 			expectedDangling: true,
 		},
+		{
+			name: "FileInfoDecided-case4-(missing data-dir for part 2)",
+			metaArr: []FileInfo{
+				{},
+				{},
+				{},
+				fi,
+			},
+			errs: []error{
+				errFileNotFound,
+				errFileNotFound,
+				nil,
+				nil,
+			},
+			dataErrs: map[int][]int{
+				0: {checkPartSuccess, checkPartSuccess, checkPartSuccess, checkPartSuccess},
+				1: {checkPartSuccess, checkPartFileNotFound, checkPartFileNotFound, checkPartFileNotFound},
+			},
+			expectedMeta:     fi,
+			expectedDangling: true,
+		},
+
+		{
+			name: "FileInfoDecided-case4-(enough data-dir existing for each part)",
+			metaArr: []FileInfo{
+				{},
+				{},
+				{},
+				fi,
+			},
+			errs: []error{
+				errFileNotFound,
+				errFileNotFound,
+				nil,
+				nil,
+			},
+			dataErrs: map[int][]int{
+				0: {checkPartFileNotFound, checkPartSuccess, checkPartSuccess, checkPartSuccess},
+				1: {checkPartSuccess, checkPartFileNotFound, checkPartSuccess, checkPartSuccess},
+				2: {checkPartSuccess, checkPartSuccess, checkPartFileNotFound, checkPartSuccess},
+				3: {checkPartSuccess, checkPartSuccess, checkPartSuccess, checkPartFileNotFound},
+			},
+			expectedMeta:     fi,
+			expectedDangling: false,
+		},
+
 		// Add new cases as seen
 	}
 	for _, testCase := range testCases {

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
@@ -456,6 +458,7 @@ 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 +469,13 @@ 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

@@ -483,6 +483,11 @@ func (er erasureObjects) newMultipartUpload(ctx context.Context, bucket string,
 	}
 	fi.DataDir = mustGetUUID()
 
+	if userDefined[ReplicationSsecChecksumHeader] != "" {
+		fi.Checksum, _ = base64.StdEncoding.DecodeString(userDefined[ReplicationSsecChecksumHeader])
+		delete(userDefined, ReplicationSsecChecksumHeader)
+	}
+
 	// Initialize erasure metadata.
 	for index := range partsMetadata {
 		partsMetadata[index] = fi
@@ -1294,9 +1299,18 @@ func (er erasureObjects) CompleteMultipartUpload(ctx context.Context, bucket str
 		defer lk.Unlock(lkctx)
 	}
 
+	// Accept encrypted checksum from incoming request.
+	if opts.UserDefined[ReplicationSsecChecksumHeader] != "" {
+		if v, err := base64.StdEncoding.DecodeString(opts.UserDefined[ReplicationSsecChecksumHeader]); err == nil {
+			fi.Checksum = v
+		}
+		delete(opts.UserDefined, ReplicationSsecChecksumHeader)
+	}
+
 	if checksumType.IsSet() {
 		checksumType |= hash.ChecksumMultipart | hash.ChecksumIncludesMultipart
-		cs := hash.NewChecksumFromData(checksumType, checksumCombined)
+		var cs *hash.Checksum
+		cs = hash.NewChecksumFromData(checksumType, checksumCombined)
 		fi.Checksum = cs.AppendTo(nil, checksumCombined)
 		if opts.EncryptFn != nil {
 			fi.Checksum = opts.EncryptFn("object-checksum", fi.Checksum)

cmd/erasure-object.go

@@ -20,6 +20,7 @@ package cmd
 import (
 	"bytes"
 	"context"
+	"encoding/base64"
 	"errors"
 	"fmt"
 	"io"
@@ -276,13 +277,13 @@ func (er erasureObjects) GetObjectNInfo(ctx context.Context, bucket, object stri
 		return gr.WithCleanupFuncs(nsUnlocker), nil
 	}
 
-	fn, off, length, err := NewGetObjectReader(rs, objInfo, opts)
+	fn, off, length, err := NewGetObjectReader(rs, objInfo, opts, h)
 	if err != nil {
 		return nil, err
 	}
 
 	if unlockOnDefer {
-		unlockOnDefer = fi.InlineData()
+		unlockOnDefer = fi.InlineData() || len(fi.Data) > 0
 	}
 
 	pr, pw := xioutil.WaitPipe()
@@ -483,8 +484,8 @@ func joinErrs(errs []error) []string {
 	return s
 }
 
-func (er erasureObjects) deleteIfDangling(ctx context.Context, bucket, object string, metaArr []FileInfo, errs []error, dataErrs []error, opts ObjectOptions) (FileInfo, error) {
-	m, ok := isObjectDangling(metaArr, errs, dataErrs)
+func (er erasureObjects) deleteIfDangling(ctx context.Context, bucket, object string, metaArr []FileInfo, errs []error, dataErrsByPart map[int][]int, opts ObjectOptions) (FileInfo, error) {
+	m, ok := isObjectDangling(metaArr, errs, dataErrsByPart)
 	if !ok {
 		// We only come here if we cannot figure out if the object
 		// can be deleted safely, in such a scenario return ReadQuorum error.
@@ -494,7 +495,7 @@ func (er erasureObjects) deleteIfDangling(ctx context.Context, bucket, object st
 	tags["set"] = er.setIndex
 	tags["pool"] = er.poolIndex
 	tags["merrs"] = joinErrs(errs)
-	tags["derrs"] = joinErrs(dataErrs)
+	tags["derrs"] = dataErrsByPart
 	if m.IsValid() {
 		tags["size"] = m.Size
 		tags["mtime"] = m.ModTime.Format(http.TimeFormat)
@@ -508,8 +509,20 @@ func (er erasureObjects) deleteIfDangling(ctx context.Context, bucket, object st
 
 	// count the number of offline disks
 	offline := 0
-	for i := 0; i < max(len(errs), len(dataErrs)); i++ {
-		if i < len(errs) && errors.Is(errs[i], errDiskNotFound) || i < len(dataErrs) && errors.Is(dataErrs[i], errDiskNotFound) {
+	for i := 0; i < len(errs); i++ {
+		var found bool
+		switch {
+		case errors.Is(errs[i], errDiskNotFound):
+			found = true
+		default:
+			for p := range dataErrsByPart {
+				if dataErrsByPart[p][i] == checkPartDiskNotFound {
+					found = true
+					break
+				}
+			}
+		}
+		if found {
 			offline++
 		}
 	}
@@ -563,35 +576,11 @@ func (er erasureObjects) deleteIfDangling(ctx context.Context, bucket, object st
 }
 
 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
+	return getFileInfo(ri.Buf, bucket, object, "", fileInfoOpts{
+		Data:             readData,
+		InclFreeVersions: inclFreeVers,
+		AllParts:         allParts,
+	})
 }
 
 func readAllRawFileInfo(ctx context.Context, disks []StorageAPI, bucket, object string, readData bool) ([]RawFileInfo, []error) {
@@ -743,8 +732,9 @@ 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)
@@ -918,6 +908,8 @@ 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))
@@ -925,7 +917,7 @@ func (er erasureObjects) getObjectFileInfo(ctx context.Context, bucket, object s
 			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
 		}
 	}
@@ -1355,6 +1347,12 @@ func (er erasureObjects) putObject(ctx context.Context, bucket string, object st
 	if opts.EncryptFn != nil {
 		fi.Checksum = opts.EncryptFn("object-checksum", fi.Checksum)
 	}
+	if userDefined[ReplicationSsecChecksumHeader] != "" {
+		if v, err := base64.StdEncoding.DecodeString(userDefined[ReplicationSsecChecksumHeader]); err == nil {
+			fi.Checksum = v
+		}
+	}
+	delete(userDefined, ReplicationSsecChecksumHeader)
 	uniqueID := mustGetUUID()
 	tempObj := uniqueID
 
@@ -1403,7 +1401,7 @@ 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))
@@ -1411,7 +1409,7 @@ 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))

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))
 		}
 	}
 
@@ -1327,11 +1327,7 @@ func (z *erasureServerPools) Status(ctx context.Context, idx int) (PoolStatus, e
 	poolInfo := z.poolMeta.Pools[idx].Clone()
 	if poolInfo.Decommission != nil {
 		poolInfo.Decommission.TotalSize = pi.Total
-		if poolInfo.Decommission.Failed || poolInfo.Decommission.Canceled {
-			poolInfo.Decommission.CurrentSize = pi.Free
-		} else {
-			poolInfo.Decommission.CurrentSize = poolInfo.Decommission.StartSize + poolInfo.Decommission.BytesDone
-		}
+		poolInfo.Decommission.CurrentSize = pi.Free
 	} else {
 		poolInfo.Decommission = &PoolDecommissionInfo{
 			TotalSize:   pi.Total,

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
@@ -794,7 +798,9 @@ 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
 

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
 		}
@@ -1526,14 +1526,12 @@ 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 != "" {
@@ -2008,10 +2006,12 @@ 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 +2025,13 @@ 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 +2085,6 @@ 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)
@@ -2199,9 +2201,8 @@ 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 +2213,15 @@ 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 +2276,16 @@ 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 +2341,18 @@ 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 +2445,7 @@ const (
 type HealthOptions struct {
 	Maintenance    bool
 	DeploymentType string
+	NoLogging      bool
 }
 
 // HealthResult returns the current state of the system, also
@@ -2449,6 +2470,24 @@ 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 +2605,18 @@ 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
 		}

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()
 	}()
@@ -235,8 +223,10 @@ 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
 			}
@@ -272,13 +262,7 @@ 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 +281,7 @@ 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 +295,7 @@ 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 +364,6 @@ 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,
@@ -662,14 +645,6 @@ 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 +1129,7 @@ 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

@@ -102,6 +102,8 @@ 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:

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/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
 		}
@@ -601,13 +604,6 @@ func setUploadForwardingMiddleware(h http.Handler) http.Handler {
 				h.ServeHTTP(w, r)
 				return
 			}
-			// forward request to peer handling this upload
-			if globalBucketTargetSys.isOffline(remote.EndpointURL) {
-				defer logger.AuditLog(r.Context(), w, r, mustGetClaimsFromToken(r))
-				writeErrorResponse(r.Context(), w, errorCodes.ToAPIErr(ErrReplicationRemoteConnectionError), r.URL)
-				return
-			}
-
 			r.URL.Scheme = remote.EndpointURL.Scheme
 			r.URL.Host = remote.EndpointURL.Host
 			// Make sure we remove any existing headers before

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

@@ -163,7 +163,10 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 	}
 
 	for _, bucket := range healBuckets {
-		_, err := objAPI.HealBucket(ctx, bucket, madmin.HealOpts{ScanMode: scanMode})
+		_, err := objAPI.HealBucket(ctx, bucket, madmin.HealOpts{
+			Recreate: true,
+			ScanMode: scanMode,
+		})
 		if err != nil {
 			// Log bucket healing error if any, we shall retry again.
 			healingLogIf(ctx, err)
@@ -262,6 +265,7 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 		// 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{
+			Recreate: true,
 			ScanMode: scanMode,
 		}); err != nil {
 			// Set this such that when we return this function
@@ -437,6 +441,8 @@ 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 +455,22 @@ 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 {
+					result = healEntrySuccess(uint64(version.Size))
+				} else {
 					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 +518,11 @@ 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 +530,10 @@ 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
 		}
 

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.
@@ -414,10 +414,9 @@ 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 +449,8 @@ 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 +39,15 @@ 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
+	requestsDeadline       time.Duration
+	requestsPool           chan struct{}
+	clusterDeadline        time.Duration
+	listQuorum             string
+	corsAllowOrigins       []string
+	replicationPriority    string
+	replicationMaxWorkers  int
+	replicationMaxLWorkers int
+	transitionWorkers      int
 
 	staleUploadsExpiry          time.Duration
 	staleUploadsCleanupInterval time.Duration
@@ -170,11 +171,12 @@ func (t *apiConfig) init(cfg api.Config, setDriveCounts []int, legacy bool) {
 	}
 	t.listQuorum = listQuorum
 	if globalReplicationPool != nil &&
-		(cfg.ReplicationPriority != t.replicationPriority || cfg.ReplicationMaxWorkers != t.replicationMaxWorkers) {
-		globalReplicationPool.ResizeWorkerPriority(cfg.ReplicationPriority, cfg.ReplicationMaxWorkers)
+		(cfg.ReplicationPriority != t.replicationPriority || cfg.ReplicationMaxWorkers != t.replicationMaxWorkers || cfg.ReplicationMaxLWorkers != t.replicationMaxLWorkers) {
+		globalReplicationPool.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 {
@@ -323,10 +325,9 @@ func maxClients(f http.HandlerFunc) http.HandlerFunc {
 		}
 
 		globalHTTPStats.addRequestsInQueue(1)
-		defer globalHTTPStats.addRequestsInQueue(-1)
-
 		pool, deadline := globalAPIConfig.getRequestsPool()
 		if pool == nil {
+			globalHTTPStats.addRequestsInQueue(-1)
 			f.ServeHTTP(w, r)
 			return
 		}
@@ -334,6 +335,7 @@ func maxClients(f http.HandlerFunc) http.HandlerFunc {
 		// 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
 		}
@@ -349,12 +351,14 @@ func maxClients(f http.HandlerFunc) http.HandlerFunc {
 		select {
 		case pool <- struct{}{}:
 			defer func() { <-pool }()
+			globalHTTPStats.addRequestsInQueue(-1)
 			if contextCanceled(ctx) {
 				w.WriteHeader(499)
 				return
 			}
 			f.ServeHTTP(w, r)
 		case <-deadlineTimer.C:
+			globalHTTPStats.addRequestsInQueue(-1)
 			if contextCanceled(ctx) {
 				w.WriteHeader(499)
 				return
@@ -364,6 +368,7 @@ func maxClients(f http.HandlerFunc) http.HandlerFunc {
 				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.
@@ -378,14 +383,16 @@ 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/handler-utils.go

@@ -33,8 +33,6 @@ import (
 	"github.com/minio/minio/internal/logger"
 	"github.com/minio/minio/internal/mcontext"
 	xnet "github.com/minio/pkg/v3/net"
-	"golang.org/x/exp/maps"
-	"golang.org/x/exp/slices"
 )
 
 const (
@@ -90,6 +88,7 @@ var supportedHeaders = []string{
 	"X-Minio-Replication-Server-Side-Encryption-Iv",
 	"X-Minio-Replication-Encrypted-Multipart",
 	"X-Minio-Replication-Actual-Object-Size",
+	ReplicationSsecChecksumHeader,
 	// Add more supported headers here.
 }
 
@@ -110,6 +109,7 @@ var replicationToInternalHeaders = map[string]string{
 	"X-Minio-Replication-Server-Side-Encryption-Iv":             "X-Minio-Internal-Server-Side-Encryption-Iv",
 	"X-Minio-Replication-Encrypted-Multipart":                   "X-Minio-Internal-Encrypted-Multipart",
 	"X-Minio-Replication-Actual-Object-Size":                    "X-Minio-Internal-Actual-Object-Size",
+	ReplicationSsecChecksumHeader:                               ReplicationSsecChecksumHeader,
 	// Add more supported headers here.
 }
 
@@ -206,8 +206,8 @@ func extractMetadataFromMime(ctx context.Context, v textproto.MIMEHeader, m map[
 	for _, supportedHeader := range supportedHeaders {
 		value, ok := nv[http.CanonicalHeaderKey(supportedHeader)]
 		if ok {
-			if slices.Contains(maps.Keys(replicationToInternalHeaders), supportedHeader) {
-				m[replicationToInternalHeaders[supportedHeader]] = strings.Join(value, ",")
+			if v, ok := replicationToInternalHeaders[supportedHeader]; ok {
+				m[v] = strings.Join(value, ",")
 			} else {
 				m[supportedHeader] = strings.Join(value, ",")
 			}

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/iam-etcd-store.go

@@ -236,9 +236,8 @@ func (ies *IAMEtcdStore) addUser(ctx context.Context, user string, userType IAMU
 				// for the expiring credentials.
 				deleteKeyEtcd(ctx, ies.client, getUserIdentityPath(user, userType))
 				deleteKeyEtcd(ctx, ies.client, getMappedPolicyPath(user, userType, false))
-				return nil
 			}
-			return err
+			return nil
 		}
 		u.Credentials.Claims = jwtClaims.Map()
 	}

cmd/iam-object-store.go

@@ -25,7 +25,6 @@ import (
 	"path"
 	"strings"
 	"sync"
-	"sync/atomic"
 	"time"
 	"unicode/utf8"
 
@@ -34,6 +33,7 @@ import (
 	"github.com/minio/minio/internal/config"
 	xioutil "github.com/minio/minio/internal/ioutil"
 	"github.com/minio/minio/internal/kms"
+	"github.com/minio/minio/internal/logger"
 	"github.com/puzpuzpuz/xsync/v3"
 )
 
@@ -44,8 +44,6 @@ type IAMObjectStore struct {
 
 	*iamCache
 
-	cachedIAMListing atomic.Value
-
 	usersSysType UsersSysType
 
 	objAPI ObjectLayer
@@ -256,9 +254,8 @@ func (iamOS *IAMObjectStore) loadUser(ctx context.Context, user string, userType
 				// for the expiring credentials.
 				iamOS.deleteIAMConfig(ctx, getUserIdentityPath(user, userType))
 				iamOS.deleteIAMConfig(ctx, getMappedPolicyPath(user, userType, false))
-				return nil
 			}
-			return err
+			return nil
 
 		}
 		u.Credentials.Claims = jwtClaims.Map()
@@ -421,8 +418,8 @@ func splitPath(s string, lastIndex bool) (string, string) {
 	return s[:i+1], s[i+1:]
 }
 
-func (iamOS *IAMObjectStore) listAllIAMConfigItems(ctx context.Context) (map[string][]string, error) {
-	res := make(map[string][]string)
+func (iamOS *IAMObjectStore) listAllIAMConfigItems(ctx context.Context) (res map[string][]string, err error) {
+	res = make(map[string][]string)
 	ctx, cancel := context.WithCancel(ctx)
 	defer cancel()
 	for item := range listIAMConfigItems(ctx, iamOS.objAPI, iamConfigPrefix+SlashSeparator) {
@@ -438,79 +435,48 @@ func (iamOS *IAMObjectStore) listAllIAMConfigItems(ctx context.Context) (map[str
 
 		res[listKey] = append(res[listKey], trimmedItem)
 	}
-	// Store the listing for later re-use.
-	iamOS.cachedIAMListing.Store(res)
+
 	return res, nil
 }
 
-// PurgeExpiredSTS - purge expired STS credentials from object store.
-func (iamOS *IAMObjectStore) PurgeExpiredSTS(ctx context.Context) error {
-	if iamOS.objAPI == nil {
-		return errServerNotInitialized
-	}
-
-	bootstrapTraceMsg("purging expired STS credentials")
-
-	iamListing, ok := iamOS.cachedIAMListing.Load().(map[string][]string)
-	if !ok {
-		// There has been no store yet. This should never happen!
-		iamLogIf(GlobalContext, errors.New("WARNING: no cached IAM listing found"))
-		return nil
-	}
-
-	// Scan STS users on disk and purge expired ones. We do not need to hold a
-	// lock with store.lock() here.
-	stsAccountsFromStore := map[string]UserIdentity{}
-	stsAccPoliciesFromStore := xsync.NewMapOf[string, MappedPolicy]()
-	for _, item := range iamListing[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) {
-			iamLogIf(GlobalContext,
-				fmt.Errorf("unable to load user during STS purge: %w (%s)", err, item))
-		}
-
-	}
-	// 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 iamListing[policyDBSTSUsersListKey] {
-		stsName := strings.TrimSuffix(item, ".json")
-		err := iamOS.loadMappedPolicy(ctx, stsName, stsUser, false, stsAccPoliciesFromStore)
-		if err != nil && !errors.Is(err, errNoSuchPolicy) {
-			iamLogIf(GlobalContext,
-				fmt.Errorf("unable to load policies during STS purge: %w (%s)", err, item))
-		}
-
-	}
-
-	// Store the newly populated map in the iam cache. This takes care of
-	// removing stale entries from the existing map.
-	iamOS.Lock()
-	defer iamOS.Unlock()
-	iamOS.iamCache.iamSTSAccountsMap = stsAccountsFromStore
-	iamOS.iamCache.iamSTSPolicyMap = stsAccPoliciesFromStore
-
-	return nil
-}
+const (
+	maxIAMLoadOpTime = 5 * time.Second
+)
 
 // 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)
@@ -518,58 +484,88 @@ func (iamOS *IAMObjectStore) loadAllFromObjStore(ctx context.Context, cache *iam
 			return fmt.Errorf("unable to load the policy doc `%s`: %w", policyName, err)
 		}
 	}
-	setDefaultCannedPolicies(cache.iamPolicyDocsMap)
+	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 {
-		bootstrapTraceMsg("loading regular IAM users")
+		bootstrapTraceMsgFirstTime("loading regular IAM users")
+		regUsersLoadStartTime := UTCNow()
 		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)
+				return fmt.Errorf("unable to load the user: %w", err)
 			}
 		}
+		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)
+		}
 
-		bootstrapTraceMsg("loading regular IAM groups")
+		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)
 			}
 		}
+		if took := time.Since(groupsLoadStartTime); took > maxIAMLoadOpTime {
+			logger.Info("Groups load took %.2fs (for %d items)", took.Seconds(), len(groupsList))
+		}
 	}
 
-	bootstrapTraceMsg("loading user policy mapping")
+	bootstrapTraceMsgFirstTime("loading user policy mapping")
+	userPolicyMappingLoadStartTime := UTCNow()
 	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)
+			return fmt.Errorf("unable to load the policy mapping for the user: %w", err)
 		}
 	}
+	if took := time.Since(userPolicyMappingLoadStartTime); took > maxIAMLoadOpTime {
+		logger.Info("User policy mappings load took %.2fs (for %d items)", took.Seconds(), len(userPolicyMappingsList))
+	}
 
-	bootstrapTraceMsg("loading group policy mapping")
+	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
@@ -584,16 +580,59 @@ 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
 	}
 
 	cache.buildUserGroupMemberships()
+
+	purgeStart := time.Now()
+
+	// Purge expired STS credentials.
+
+	// Scan STS users on disk and purge expired ones.
+	stsAccountsFromStore := map[string]UserIdentity{}
+	stsAccPoliciesFromStore := xsync.NewMapOf[string, MappedPolicy]()
+	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)
+		}
+
+	}
+	// 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)
+		}
+
+	}
+
+	took := time.Since(purgeStart).Seconds()
+	if took > maxDurationSecondsForLog {
+		// Log if we took a lot of time to load.
+		logger.Info("IAM expired STS purge took %.2fs", took)
+	}
+
+	// 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
+
 	return nil
 }
 

cmd/iam-store.go

@@ -33,8 +33,10 @@ import (
 	"github.com/minio/minio/internal/auth"
 	"github.com/minio/minio/internal/config/identity/openid"
 	"github.com/minio/minio/internal/jwt"
+	"github.com/minio/pkg/v3/console"
 	"github.com/minio/pkg/v3/policy"
 	"github.com/puzpuzpuz/xsync/v3"
+	"golang.org/x/sync/singleflight"
 )
 
 const (
@@ -429,8 +431,41 @@ func (c *iamCache) policyDBGet(store *IAMStoreSys, name string, isGroup bool) ([
 		}
 	}
 
-	// returned policy could be empty
-	policies := mp.toSlice()
+	// returned policy could be empty, we use set to de-duplicate.
+	policies := set.CreateStringSet(mp.toSlice()...)
+
+	for _, group := range u.Credentials.Groups {
+		if store.getUsersSysType() == MinIOUsersSysType {
+			g, ok := c.iamGroupsMap[group]
+			if !ok {
+				if err := store.loadGroup(context.Background(), group, c.iamGroupsMap); err != nil {
+					return nil, time.Time{}, err
+				}
+				g, ok = c.iamGroupsMap[group]
+				if !ok {
+					return nil, time.Time{}, errNoSuchGroup
+				}
+			}
+
+			// Group is disabled, so we return no policy - this
+			// ensures the request is denied.
+			if g.Status == statusDisabled {
+				return nil, time.Time{}, nil
+			}
+		}
+
+		policy, ok := c.iamGroupPolicyMap.Load(group)
+		if !ok {
+			if err := store.loadMappedPolicyWithRetry(context.TODO(), group, regUser, true, c.iamGroupPolicyMap, 3); err != nil && !errors.Is(err, errNoSuchPolicy) {
+				return nil, time.Time{}, err
+			}
+			policy, _ = c.iamGroupPolicyMap.Load(group)
+		}
+
+		for _, p := range policy.toSlice() {
+			policies.Add(p)
+		}
+	}
 
 	for _, group := range c.iamUserGroupMemberships[name].ToSlice() {
 		if store.getUsersSysType() == MinIOUsersSysType {
@@ -460,10 +495,12 @@ func (c *iamCache) policyDBGet(store *IAMStoreSys, name string, isGroup bool) ([
 			policy, _ = c.iamGroupPolicyMap.Load(group)
 		}
 
-		policies = append(policies, policy.toSlice()...)
+		for _, p := range policy.toSlice() {
+			policies.Add(p)
+		}
 	}
 
-	return policies, mp.UpdatedAt, nil
+	return policies.ToSlice(), mp.UpdatedAt, nil
 }
 
 func (c *iamCache) updateUserWithClaims(key string, u UserIdentity) error {
@@ -532,38 +569,28 @@ func setDefaultCannedPolicies(policies map[string]PolicyDoc) {
 	}
 }
 
-// PurgeExpiredSTS - purges expired STS credentials.
-func (store *IAMStoreSys) PurgeExpiredSTS(ctx context.Context) error {
-	iamOS, ok := store.IAMStorageAPI.(*IAMObjectStore)
-	if !ok {
-		// No purging is done for non-object storage.
-		return nil
-	}
-	return iamOS.PurgeExpiredSTS(ctx)
-}
-
 // LoadIAMCache reads all IAM items and populates a new iamCache object and
 // replaces the in-memory cache object.
 func (store *IAMStoreSys) LoadIAMCache(ctx context.Context, firstTime bool) error {
-	bootstrapTraceMsg := func(s string) {
+	bootstrapTraceMsgFirstTime := func(s string) {
 		if firstTime {
 			bootstrapTraceMsg(s)
 		}
 	}
-	bootstrapTraceMsg("loading IAM data")
+	bootstrapTraceMsgFirstTime("loading IAM data")
 
 	newCache := newIamCache()
 
 	loadedAt := time.Now()
 
 	if iamOS, ok := store.IAMStorageAPI.(*IAMObjectStore); ok {
-		err := iamOS.loadAllFromObjStore(ctx, newCache)
+		err := iamOS.loadAllFromObjStore(ctx, newCache, firstTime)
 		if err != nil {
 			return err
 		}
 	} else {
-
-		bootstrapTraceMsg("loading policy documents")
+		// Only non-object IAM store (i.e. only etcd backend).
+		bootstrapTraceMsgFirstTime("loading policy documents")
 		if err := store.loadPolicyDocs(ctx, newCache.iamPolicyDocsMap); err != nil {
 			return err
 		}
@@ -572,29 +599,29 @@ func (store *IAMStoreSys) LoadIAMCache(ctx context.Context, firstTime bool) erro
 		setDefaultCannedPolicies(newCache.iamPolicyDocsMap)
 
 		if store.getUsersSysType() == MinIOUsersSysType {
-			bootstrapTraceMsg("loading regular users")
+			bootstrapTraceMsgFirstTime("loading regular users")
 			if err := store.loadUsers(ctx, regUser, newCache.iamUsersMap); err != nil {
 				return err
 			}
-			bootstrapTraceMsg("loading regular groups")
+			bootstrapTraceMsgFirstTime("loading regular groups")
 			if err := store.loadGroups(ctx, newCache.iamGroupsMap); err != nil {
 				return err
 			}
 		}
 
-		bootstrapTraceMsg("loading user policy mapping")
+		bootstrapTraceMsgFirstTime("loading user policy mapping")
 		// load polices mapped to users
 		if err := store.loadMappedPolicies(ctx, regUser, false, newCache.iamUserPolicyMap); err != nil {
 			return err
 		}
 
-		bootstrapTraceMsg("loading group policy mapping")
+		bootstrapTraceMsgFirstTime("loading group policy mapping")
 		// load policies mapped to groups
 		if err := store.loadMappedPolicies(ctx, regUser, true, newCache.iamGroupPolicyMap); err != nil {
 			return err
 		}
 
-		bootstrapTraceMsg("loading service accounts")
+		bootstrapTraceMsgFirstTime("loading service accounts")
 		// load service accounts
 		if err := store.loadUsers(ctx, svcUser, newCache.iamUsersMap); err != nil {
 			return err
@@ -643,6 +670,8 @@ func (store *IAMStoreSys) LoadIAMCache(ctx context.Context, firstTime bool) erro
 // layer.
 type IAMStoreSys struct {
 	IAMStorageAPI
+
+	group *singleflight.Group
 }
 
 // HasWatcher - returns if the storage system has a watcher.
@@ -943,12 +972,7 @@ func (store *IAMStoreSys) GetGroupDescription(group string) (gd madmin.GroupDesc
 	}, nil
 }
 
-// ListGroups - lists groups. Since this is not going to be a frequent
-// operation, we fetch this info from storage, and refresh the cache as well.
-func (store *IAMStoreSys) ListGroups(ctx context.Context) (res []string, err error) {
-	cache := store.lock()
-	defer store.unlock()
-
+func (store *IAMStoreSys) updateGroups(ctx context.Context, cache *iamCache) (res []string, err error) {
 	if store.getUsersSysType() == MinIOUsersSysType {
 		m := map[string]GroupInfo{}
 		err = store.loadGroups(ctx, m)
@@ -976,7 +1000,16 @@ func (store *IAMStoreSys) ListGroups(ctx context.Context) (res []string, err err
 		})
 	}
 
-	return
+	return res, nil
+}
+
+// ListGroups - lists groups. Since this is not going to be a frequent
+// operation, we fetch this info from storage, and refresh the cache as well.
+func (store *IAMStoreSys) ListGroups(ctx context.Context) (res []string, err error) {
+	cache := store.lock()
+	defer store.unlock()
+
+	return store.updateGroups(ctx, cache)
 }
 
 // listGroups - lists groups - fetch groups from cache
@@ -1451,16 +1484,51 @@ func filterPolicies(cache *iamCache, policyName string, bucketName string) (stri
 	return strings.Join(policies, ","), policy.MergePolicies(toMerge...)
 }
 
-// FilterPolicies - accepts a comma separated list of policy names as a string
-// and bucket and returns only policies that currently exist in MinIO. If
-// bucketName is non-empty, additionally filters policies matching the bucket.
-// The first returned value is the list of currently existing policies, and the
-// second is their combined policy definition.
-func (store *IAMStoreSys) FilterPolicies(policyName string, bucketName string) (string, policy.Policy) {
-	cache := store.rlock()
-	defer store.runlock()
+// MergePolicies - accepts a comma separated list of policy names as a string
+// and returns only policies that currently exist in MinIO. It includes hot loading
+// of policies if not in the memory
+func (store *IAMStoreSys) MergePolicies(policyName string) (string, policy.Policy) {
+	var policies []string
+	var missingPolicies []string
+	var toMerge []policy.Policy
 
-	return filterPolicies(cache, policyName, bucketName)
+	cache := store.rlock()
+	for _, policy := range newMappedPolicy(policyName).toSlice() {
+		if policy == "" {
+			continue
+		}
+		p, found := cache.iamPolicyDocsMap[policy]
+		if !found {
+			missingPolicies = append(missingPolicies, policy)
+			continue
+		}
+		policies = append(policies, policy)
+		toMerge = append(toMerge, p.Policy)
+	}
+	store.runlock()
+
+	if len(missingPolicies) > 0 {
+		m := make(map[string]PolicyDoc)
+		for _, policy := range missingPolicies {
+			ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+			_ = store.loadPolicyDoc(ctx, policy, m)
+			cancel()
+		}
+
+		cache := store.lock()
+		for policy, p := range m {
+			cache.iamPolicyDocsMap[policy] = p
+		}
+		store.unlock()
+
+		for policy, p := range m {
+			policies = append(policies, policy)
+			toMerge = append(toMerge, p.Policy)
+		}
+
+	}
+
+	return strings.Join(policies, ","), policy.MergePolicies(toMerge...)
 }
 
 // GetBucketUsers - returns users (not STS or service accounts) that have access
@@ -1810,6 +1878,7 @@ func (store *IAMStoreSys) DeleteUser(ctx context.Context, accessKey string, user
 		err = nil
 	}
 	delete(cache.iamUsersMap, accessKey)
+	store.group.Forget(accessKey)
 
 	cache.updatedAt = time.Now()
 
@@ -1885,6 +1954,7 @@ func (store *IAMStoreSys) DeleteUsers(ctx context.Context, users []string) error
 			err := store.deleteUserIdentity(ctx, user, userType)
 			iamLogIf(GlobalContext, err)
 			delete(cache.iamUsersMap, user)
+			store.group.Forget(user)
 
 			deleted = true
 		}
@@ -1911,6 +1981,11 @@ func (store *IAMStoreSys) GetAllParentUsers() map[string]ParentUserInfo {
 	cache := store.rlock()
 	defer store.runlock()
 
+	return store.getParentUsers(cache)
+}
+
+// assumes store is locked by caller.
+func (store *IAMStoreSys) getParentUsers(cache *iamCache) map[string]ParentUserInfo {
 	res := map[string]ParentUserInfo{}
 	for _, ui := range cache.iamUsersMap {
 		cred := ui.Credentials
@@ -1981,50 +2056,104 @@ func (store *IAMStoreSys) GetAllParentUsers() map[string]ParentUserInfo {
 	return res
 }
 
-// Assumes store is locked by caller. If users is empty, returns all user mappings.
-func (store *IAMStoreSys) listUserPolicyMappings(cache *iamCache, users []string,
-	userPredicate func(string) bool,
+// GetAllSTSUserMappings - Loads all STS user policy mappings from storage and
+// returns them. Also gets any STS users that do not have policy mappings but have
+// Service Accounts or STS keys (This is useful if the user is part of a group)
+func (store *IAMStoreSys) GetAllSTSUserMappings(userPredicate func(string) bool) (map[string]string, error) {
+	cache := store.rlock()
+	defer store.runlock()
+
+	stsMap := make(map[string]string)
+	m := xsync.NewMapOf[string, MappedPolicy]()
+	if err := store.loadMappedPolicies(context.Background(), stsUser, false, m); err != nil {
+		return nil, err
+	}
+
+	m.Range(func(user string, mappedPolicy MappedPolicy) bool {
+		if userPredicate != nil && !userPredicate(user) {
+			return true
+		}
+		stsMap[user] = mappedPolicy.Policies
+		return true
+	})
+
+	for user := range store.getParentUsers(cache) {
+		if _, ok := stsMap[user]; !ok {
+			if userPredicate != nil && !userPredicate(user) {
+				continue
+			}
+			stsMap[user] = ""
+		}
+	}
+	return stsMap, nil
+}
+
+// Assumes store is locked by caller. If userMap is empty, returns all user mappings.
+func (store *IAMStoreSys) listUserPolicyMappings(cache *iamCache, userMap map[string]set.StringSet,
+	userPredicate func(string) bool, decodeFunc func(string) string,
 ) []madmin.UserPolicyEntities {
+	stsMap := xsync.NewMapOf[string, MappedPolicy]()
+	resMap := make(map[string]madmin.UserPolicyEntities, len(userMap))
+
+	for user, groupSet := range userMap {
+		// Attempt to load parent user mapping for STS accounts
+		store.loadMappedPolicy(context.TODO(), user, stsUser, false, stsMap)
+		decodeUser := user
+		if decodeFunc != nil {
+			decodeUser = decodeFunc(user)
+		}
+		blankEntities := madmin.UserPolicyEntities{User: decodeUser}
+		if !groupSet.IsEmpty() {
+			blankEntities.MemberOfMappings = store.listGroupPolicyMappings(cache, groupSet, nil, decodeFunc)
+		}
+		resMap[user] = blankEntities
+	}
+
 	var r []madmin.UserPolicyEntities
-	usersSet := set.CreateStringSet(users...)
 	cache.iamUserPolicyMap.Range(func(user string, mappedPolicy MappedPolicy) bool {
 		if userPredicate != nil && !userPredicate(user) {
 			return true
 		}
 
-		if !usersSet.IsEmpty() && !usersSet.Contains(user) {
-			return true
+		entitiesWithMemberOf, ok := resMap[user]
+		if !ok {
+			if len(userMap) > 0 {
+				return true
+			}
+			decodeUser := user
+			if decodeFunc != nil {
+				decodeUser = decodeFunc(user)
+			}
+			entitiesWithMemberOf = madmin.UserPolicyEntities{User: decodeUser}
 		}
 
 		ps := mappedPolicy.toSlice()
 		sort.Strings(ps)
-		r = append(r, madmin.UserPolicyEntities{
-			User:     user,
-			Policies: ps,
-		})
+		entitiesWithMemberOf.Policies = ps
+		resMap[user] = entitiesWithMemberOf
 		return true
 	})
 
-	stsMap := xsync.NewMapOf[string, MappedPolicy]()
-	for _, user := range users {
-		// Attempt to load parent user mapping for STS accounts
-		store.loadMappedPolicy(context.TODO(), user, stsUser, false, stsMap)
-	}
-
 	stsMap.Range(func(user string, mappedPolicy MappedPolicy) bool {
 		if userPredicate != nil && !userPredicate(user) {
 			return true
 		}
 
+		entitiesWithMemberOf := resMap[user]
+
 		ps := mappedPolicy.toSlice()
 		sort.Strings(ps)
-		r = append(r, madmin.UserPolicyEntities{
-			User:     user,
-			Policies: ps,
-		})
+		entitiesWithMemberOf.Policies = ps
+		resMap[user] = entitiesWithMemberOf
 		return true
 	})
 
+	for _, v := range resMap {
+		if v.Policies != nil || v.MemberOfMappings != nil {
+			r = append(r, v)
+		}
+	}
+
 	sort.Slice(r, func(i, j int) bool {
 		return r[i].User < r[j].User
 	})
@@ -2033,11 +2162,11 @@ func (store *IAMStoreSys) listUserPolicyMappings(cache *iamCache, users []string
 }
 
 // Assumes store is locked by caller. If groups is empty, returns all group mappings.
-func (store *IAMStoreSys) listGroupPolicyMappings(cache *iamCache, groups []string,
-	groupPredicate func(string) bool,
+func (store *IAMStoreSys) listGroupPolicyMappings(cache *iamCache, groupsSet set.StringSet,
+	groupPredicate func(string) bool, decodeFunc func(string) string,
 ) []madmin.GroupPolicyEntities {
 	var r []madmin.GroupPolicyEntities
-	groupsSet := set.CreateStringSet(groups...)
+
 	cache.iamGroupPolicyMap.Range(func(group string, mappedPolicy MappedPolicy) bool {
 		if groupPredicate != nil && !groupPredicate(group) {
 			return true
@@ -2047,10 +2176,15 @@ func (store *IAMStoreSys) listGroupPolicyMappings(cache *iamCache, groups []stri
 			return true
 		}
 
+		decodeGroup := group
+		if decodeFunc != nil {
+			decodeGroup = decodeFunc(group)
+		}
+
 		ps := mappedPolicy.toSlice()
 		sort.Strings(ps)
 		r = append(r, madmin.GroupPolicyEntities{
-			Group:    group,
+			Group:    decodeGroup,
 			Policies: ps,
 		})
 		return true
@@ -2064,17 +2198,20 @@ func (store *IAMStoreSys) listGroupPolicyMappings(cache *iamCache, groups []stri
 }
 
 // Assumes store is locked by caller. If policies is empty, returns all policy mappings.
-func (store *IAMStoreSys) listPolicyMappings(cache *iamCache, policies []string,
-	userPredicate, groupPredicate func(string) bool,
+func (store *IAMStoreSys) listPolicyMappings(cache *iamCache, queryPolSet set.StringSet,
+	userPredicate, groupPredicate func(string) bool, decodeFunc func(string) string,
 ) []madmin.PolicyEntities {
-	queryPolSet := set.CreateStringSet(policies...)
-
 	policyToUsersMap := make(map[string]set.StringSet)
 	cache.iamUserPolicyMap.Range(func(user string, mappedPolicy MappedPolicy) bool {
 		if userPredicate != nil && !userPredicate(user) {
 			return true
 		}
 
+		decodeUser := user
+		if decodeFunc != nil {
+			decodeUser = decodeFunc(user)
+		}
+
 		commonPolicySet := mappedPolicy.policySet()
 		if !queryPolSet.IsEmpty() {
 			commonPolicySet = commonPolicySet.Intersection(queryPolSet)
@@ -2082,9 +2219,9 @@ func (store *IAMStoreSys) listPolicyMappings(cache *iamCache, policies []string,
 		for _, policy := range commonPolicySet.ToSlice() {
 			s, ok := policyToUsersMap[policy]
 			if !ok {
-				policyToUsersMap[policy] = set.CreateStringSet(user)
+				policyToUsersMap[policy] = set.CreateStringSet(decodeUser)
 			} else {
-				s.Add(user)
+				s.Add(decodeUser)
 				policyToUsersMap[policy] = s
 			}
 		}
@@ -2098,6 +2235,11 @@ func (store *IAMStoreSys) listPolicyMappings(cache *iamCache, policies []string,
 				continue
 			}
 
+			decodeUser := user
+			if decodeFunc != nil {
+				decodeUser = decodeFunc(user)
+			}
+
 			var mappedPolicy MappedPolicy
 			store.loadIAMConfig(context.Background(), &mappedPolicy, getMappedPolicyPath(user, stsUser, false))
 
@@ -2108,9 +2250,9 @@ func (store *IAMStoreSys) listPolicyMappings(cache *iamCache, policies []string,
 			for _, policy := range commonPolicySet.ToSlice() {
 				s, ok := policyToUsersMap[policy]
 				if !ok {
-					policyToUsersMap[policy] = set.CreateStringSet(user)
+					policyToUsersMap[policy] = set.CreateStringSet(decodeUser)
 				} else {
-					s.Add(user)
+					s.Add(decodeUser)
 					policyToUsersMap[policy] = s
 				}
 			}
@@ -2125,6 +2267,11 @@ func (store *IAMStoreSys) listPolicyMappings(cache *iamCache, policies []string,
 					return true
 				}
 
+				decodeUser := user
+				if decodeFunc != nil {
+					decodeUser = decodeFunc(user)
+				}
+
 				commonPolicySet := mappedPolicy.policySet()
 				if !queryPolSet.IsEmpty() {
 					commonPolicySet = commonPolicySet.Intersection(queryPolSet)
@@ -2132,9 +2279,9 @@ func (store *IAMStoreSys) listPolicyMappings(cache *iamCache, policies []string,
 				for _, policy := range commonPolicySet.ToSlice() {
 					s, ok := policyToUsersMap[policy]
 					if !ok {
-						policyToUsersMap[policy] = set.CreateStringSet(user)
+						policyToUsersMap[policy] = set.CreateStringSet(decodeUser)
 					} else {
-						s.Add(user)
+						s.Add(decodeUser)
 						policyToUsersMap[policy] = s
 					}
 				}
@@ -2149,6 +2296,11 @@ func (store *IAMStoreSys) listPolicyMappings(cache *iamCache, policies []string,
 			return true
 		}
 
+		decodeGroup := group
+		if decodeFunc != nil {
+			decodeGroup = decodeFunc(group)
+		}
+
 		commonPolicySet := mappedPolicy.policySet()
 		if !queryPolSet.IsEmpty() {
 			commonPolicySet = commonPolicySet.Intersection(queryPolSet)
@@ -2156,9 +2308,9 @@ func (store *IAMStoreSys) listPolicyMappings(cache *iamCache, policies []string,
 		for _, policy := range commonPolicySet.ToSlice() {
 			s, ok := policyToGroupsMap[policy]
 			if !ok {
-				policyToGroupsMap[policy] = set.CreateStringSet(group)
+				policyToGroupsMap[policy] = set.CreateStringSet(decodeGroup)
 			} else {
-				s.Add(group)
+				s.Add(decodeGroup)
 				policyToGroupsMap[policy] = s
 			}
 		}
@@ -2198,24 +2350,24 @@ func (store *IAMStoreSys) listPolicyMappings(cache *iamCache, policies []string,
 }
 
 // ListPolicyMappings - return users/groups mapped to policies.
-func (store *IAMStoreSys) ListPolicyMappings(q madmin.PolicyEntitiesQuery,
-	userPredicate, groupPredicate func(string) bool,
+func (store *IAMStoreSys) ListPolicyMappings(q cleanEntitiesQuery,
+	userPredicate, groupPredicate func(string) bool, decodeFunc func(string) string,
 ) madmin.PolicyEntitiesResult {
 	cache := store.rlock()
 	defer store.runlock()
 
 	var result madmin.PolicyEntitiesResult
 
-	isAllPoliciesQuery := len(q.Users) == 0 && len(q.Groups) == 0 && len(q.Policy) == 0
+	isAllPoliciesQuery := len(q.Users) == 0 && len(q.Groups) == 0 && len(q.Policies) == 0
 
 	if len(q.Users) > 0 {
-		result.UserMappings = store.listUserPolicyMappings(cache, q.Users, userPredicate)
+		result.UserMappings = store.listUserPolicyMappings(cache, q.Users, userPredicate, decodeFunc)
 	}
 	if len(q.Groups) > 0 {
-		result.GroupMappings = store.listGroupPolicyMappings(cache, q.Groups, groupPredicate)
+		result.GroupMappings = store.listGroupPolicyMappings(cache, q.Groups, groupPredicate, decodeFunc)
 	}
-	if len(q.Policy) > 0 || isAllPoliciesQuery {
-		result.PolicyMappings = store.listPolicyMappings(cache, q.Policy, userPredicate, groupPredicate)
+	if len(q.Policies) > 0 || isAllPoliciesQuery {
+		result.PolicyMappings = store.listPolicyMappings(cache, q.Policies, userPredicate, groupPredicate, decodeFunc)
 	}
 	return result
 }
@@ -2572,69 +2724,100 @@ func (store *IAMStoreSys) UpdateUserIdentity(ctx context.Context, cred auth.Cred
 }
 
 // LoadUser - attempts to load user info from storage and updates cache.
-func (store *IAMStoreSys) LoadUser(ctx context.Context, accessKey string) {
-	cache := store.lock()
-	defer store.unlock()
+func (store *IAMStoreSys) LoadUser(ctx context.Context, accessKey string) error {
+	// We use singleflight to de-duplicate requests when server
+	// is coming up and loading accessKey and its associated assets
+	val, err, shared := store.group.Do(accessKey, func() (val interface{}, err error) {
+		cache := store.lock()
+		defer func() {
+			cache.updatedAt = time.Now()
+			store.unlock()
+		}()
 
-	cache.updatedAt = time.Now()
+		_, found := cache.iamUsersMap[accessKey]
 
-	_, found := cache.iamUsersMap[accessKey]
-
-	// Check for regular user access key
-	if !found {
-		store.loadUser(ctx, accessKey, regUser, cache.iamUsersMap)
-		if _, found = cache.iamUsersMap[accessKey]; found {
-			// load mapped policies
-			store.loadMappedPolicyWithRetry(ctx, accessKey, regUser, false, cache.iamUserPolicyMap, 3)
-		}
-	}
-
-	// Check for service account
-	if !found {
-		store.loadUser(ctx, accessKey, svcUser, cache.iamUsersMap)
-		var svc UserIdentity
-		svc, found = cache.iamUsersMap[accessKey]
-		if found {
-			// Load parent user and mapped policies.
-			if store.getUsersSysType() == MinIOUsersSysType {
-				store.loadUser(ctx, svc.Credentials.ParentUser, regUser, cache.iamUsersMap)
-				store.loadMappedPolicyWithRetry(ctx, svc.Credentials.ParentUser, regUser, false, cache.iamUserPolicyMap, 3)
-			} else {
-				// In case of LDAP the parent user's policy mapping needs to be
-				// loaded into sts map
-				store.loadMappedPolicyWithRetry(ctx, svc.Credentials.ParentUser, stsUser, false, cache.iamSTSPolicyMap, 3)
+		// Check for regular user access key
+		if !found {
+			err = store.loadUser(ctx, accessKey, regUser, cache.iamUsersMap)
+			if _, found = cache.iamUsersMap[accessKey]; found {
+				// load mapped policies
+				err = store.loadMappedPolicyWithRetry(ctx, accessKey, regUser, false, cache.iamUserPolicyMap, 3)
 			}
 		}
-	}
 
-	// Check for STS account
-	stsAccountFound := false
-	var stsUserCred UserIdentity
-	if !found {
-		store.loadUser(ctx, accessKey, stsUser, cache.iamSTSAccountsMap)
-		if stsUserCred, found = cache.iamSTSAccountsMap[accessKey]; found {
-			// Load mapped policy
-			store.loadMappedPolicyWithRetry(ctx, stsUserCred.Credentials.ParentUser, stsUser, false, cache.iamSTSPolicyMap, 3)
-			stsAccountFound = true
-		}
-	}
-
-	// Load any associated policy definitions
-	if !stsAccountFound {
-		pols, _ := cache.iamUserPolicyMap.Load(accessKey)
-		for _, policy := range pols.toSlice() {
-			if _, found = cache.iamPolicyDocsMap[policy]; !found {
-				store.loadPolicyDocWithRetry(ctx, policy, cache.iamPolicyDocsMap, 3)
+		// Check for service account
+		if !found {
+			err = store.loadUser(ctx, accessKey, svcUser, cache.iamUsersMap)
+			var svc UserIdentity
+			svc, found = cache.iamUsersMap[accessKey]
+			if found {
+				// Load parent user and mapped policies.
+				if store.getUsersSysType() == MinIOUsersSysType {
+					err = store.loadUser(ctx, svc.Credentials.ParentUser, regUser, cache.iamUsersMap)
+					if err == nil {
+						err = store.loadMappedPolicyWithRetry(ctx, svc.Credentials.ParentUser, regUser, false, cache.iamUserPolicyMap, 3)
+					}
+				} else {
+					// In case of LDAP the parent user's policy mapping needs to be
+					// loaded into sts map
+					err = store.loadMappedPolicyWithRetry(ctx, svc.Credentials.ParentUser, stsUser, false, cache.iamSTSPolicyMap, 3)
+				}
 			}
 		}
-	} else {
-		pols, _ := cache.iamSTSPolicyMap.Load(stsUserCred.Credentials.AccessKey)
-		for _, policy := range pols.toSlice() {
-			if _, found = cache.iamPolicyDocsMap[policy]; !found {
-				store.loadPolicyDocWithRetry(ctx, policy, cache.iamPolicyDocsMap, 3)
+
+		// Check for STS account
+		stsAccountFound := false
+		var stsUserCred UserIdentity
+		if !found {
+			err = store.loadUser(ctx, accessKey, stsUser, cache.iamSTSAccountsMap)
+			if stsUserCred, found = cache.iamSTSAccountsMap[accessKey]; found {
+				// Load mapped policy
+				err = store.loadMappedPolicyWithRetry(ctx, stsUserCred.Credentials.ParentUser, stsUser, false, cache.iamSTSPolicyMap, 3)
+				stsAccountFound = true
 			}
 		}
+
+		// Load any associated policy definitions
+		if !stsAccountFound {
+			pols, _ := cache.iamUserPolicyMap.Load(accessKey)
+			for _, policy := range pols.toSlice() {
+				if _, found = cache.iamPolicyDocsMap[policy]; !found {
+					err = store.loadPolicyDocWithRetry(ctx, policy, cache.iamPolicyDocsMap, 3)
+				}
+			}
+		} else {
+			pols, _ := cache.iamSTSPolicyMap.Load(stsUserCred.Credentials.AccessKey)
+			for _, policy := range pols.toSlice() {
+				if _, found = cache.iamPolicyDocsMap[policy]; !found {
+					err = store.loadPolicyDocWithRetry(ctx, policy, cache.iamPolicyDocsMap, 3)
+				}
+			}
+		}
+
+		load := len(cache.iamGroupsMap) == 0
+		if store.getUsersSysType() == LDAPUsersSysType && cache.iamGroupPolicyMap.Size() == 0 {
+			load = true
+		}
+		if load {
+			if _, err = store.updateGroups(ctx, cache); err != nil {
+				return "done", err
+			}
+		}
+
+		cache.buildUserGroupMemberships()
+
+		return "done", err
+	})
+
+	if serverDebugLog {
+		console.Debugln("loadUser: loading shared", val, err, shared)
 	}
+
+	if IsErr(err, errNoSuchUser, errNoSuchPolicy, errNoSuchGroup) {
+		return nil
+	}
+
+	return err
 }
 
 func extractJWTClaims(u UserIdentity) (*jwt.MapClaims, error) {

cmd/iam.go

@@ -52,6 +52,7 @@ import (
 	"github.com/minio/pkg/v3/ldap"
 	"github.com/minio/pkg/v3/policy"
 	etcd "go.etcd.io/etcd/client/v3"
+	"golang.org/x/sync/singleflight"
 )
 
 // UsersSysType - defines the type of users and groups system that is
@@ -177,9 +178,9 @@ func (sys *IAMSys) initStore(objAPI ObjectLayer, etcdClient *etcd.Client) {
 	}
 
 	if etcdClient == nil {
-		sys.store = &IAMStoreSys{newIAMObjectStore(objAPI, sys.usersSysType)}
+		sys.store = &IAMStoreSys{newIAMObjectStore(objAPI, sys.usersSysType), &singleflight.Group{}}
 	} else {
-		sys.store = &IAMStoreSys{newIAMEtcdStore(etcdClient, sys.usersSysType)}
+		sys.store = &IAMStoreSys{newIAMEtcdStore(etcdClient, sys.usersSysType), &singleflight.Group{}}
 	}
 }
 
@@ -216,6 +217,9 @@ func (sys *IAMSys) Load(ctx context.Context, firstTime bool) error {
 
 	if firstTime {
 		bootstrapTraceMsg(fmt.Sprintf("globalIAMSys.Load(): (duration: %s)", loadDuration))
+		if globalIsDistErasure {
+			logger.Info("IAM load(startup) finished. (duration: %s)", loadDuration)
+		}
 	}
 
 	select {
@@ -314,6 +318,24 @@ func (sys *IAMSys) Init(ctx context.Context, objAPI ObjectLayer, etcdClient *etc
 		break
 	}
 
+	cache := sys.store.lock()
+	setDefaultCannedPolicies(cache.iamPolicyDocsMap)
+	sys.store.unlock()
+
+	// Load RoleARNs
+	sys.rolesMap = make(map[arn.ARN]string)
+
+	// From OpenID
+	if riMap := sys.OpenIDConfig.GetRoleInfo(); riMap != nil {
+		sys.validateAndAddRolePolicyMappings(ctx, riMap)
+	}
+
+	// From AuthN plugin if enabled.
+	if authn := newGlobalAuthNPluginFn(); authn != nil {
+		riMap := authn.GetRoleInfo()
+		sys.validateAndAddRolePolicyMappings(ctx, riMap)
+	}
+
 	// Load IAM data from storage.
 	for {
 		if err := sys.Load(retryCtx, true); err != nil {
@@ -333,25 +355,13 @@ func (sys *IAMSys) Init(ctx context.Context, objAPI ObjectLayer, etcdClient *etc
 
 	go sys.periodicRoutines(ctx, refreshInterval)
 
-	// Load RoleARNs
-	sys.rolesMap = make(map[arn.ARN]string)
-
-	// From OpenID
-	if riMap := sys.OpenIDConfig.GetRoleInfo(); riMap != nil {
-		sys.validateAndAddRolePolicyMappings(ctx, riMap)
-	}
-
-	// From AuthN plugin if enabled.
-	if authn := newGlobalAuthNPluginFn(); authn != nil {
-		riMap := authn.GetRoleInfo()
-		sys.validateAndAddRolePolicyMappings(ctx, riMap)
-	}
-
 	sys.printIAMRoles()
 
 	bootstrapTraceMsg("finishing IAM loading")
 }
 
+const maxDurationSecondsForLog = 5
+
 func (sys *IAMSys) periodicRoutines(ctx context.Context, baseInterval time.Duration) {
 	// Watch for IAM config changes for iamStorageWatcher.
 	watcher, isWatcher := sys.store.IAMStorageAPI.(iamStorageWatcher)
@@ -384,7 +394,6 @@ func (sys *IAMSys) periodicRoutines(ctx context.Context, baseInterval time.Durat
 		return baseInterval/2 + randAmt
 	}
 
-	var maxDurationSecondsForLog float64 = 5
 	timer := time.NewTimer(waitInterval())
 	defer timer.Stop()
 
@@ -394,24 +403,12 @@ 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)
-				}
-			}
-
-			// Purge expired STS credentials.
-			purgeStart := time.Now()
-			if err := sys.store.PurgeExpiredSTS(ctx); err != nil {
-				iamLogIf(ctx, fmt.Errorf("Failure in periodic STS purge for IAM (took %.2fs): %v", time.Since(purgeStart).Seconds(), err))
-			} else {
-				took := time.Since(purgeStart).Seconds()
-				if took > maxDurationSecondsForLog {
-					// Log if we took a lot of time to load.
-					logger.Info("IAM expired STS purge took %.2fs", took)
+					logger.Info("IAM refresh took (duration: %s)", took)
 				}
 			}
 
@@ -446,7 +443,7 @@ 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 {
@@ -682,7 +679,7 @@ func (sys *IAMSys) CurrentPolicies(policyName string) string {
 		return ""
 	}
 
-	policies, _ := sys.store.FilterPolicies(policyName, "")
+	policies, _ := sys.store.MergePolicies(policyName)
 	return policies
 }
 
@@ -796,11 +793,15 @@ 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
@@ -809,6 +810,57 @@ 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() {
@@ -821,7 +873,8 @@ 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():
@@ -895,6 +948,7 @@ 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 {
@@ -904,7 +958,7 @@ 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():
@@ -1388,11 +1442,13 @@ 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() {
@@ -1435,25 +1491,28 @@ 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)
@@ -1461,8 +1520,9 @@ 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...)
@@ -1520,11 +1580,11 @@ 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
 		}
@@ -1539,11 +1599,11 @@ 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
 			}
@@ -1578,31 +1638,16 @@ func (sys *IAMSys) NormalizeLDAPAccessKeypairs(ctx context.Context, accessKeyMap
 
 func (sys *IAMSys) getStoredLDAPPolicyMappingKeys(ctx context.Context, isGroup bool) set.StringSet {
 	entityKeysInStorage := set.NewStringSet()
-	if iamOS, ok := sys.store.IAMStorageAPI.(*IAMObjectStore); ok {
-		// Load existing mapping keys from the cached listing for
-		// `IAMObjectStore`.
-		iamFilesListing := iamOS.cachedIAMListing.Load().(map[string][]string)
-		listKey := policyDBSTSUsersListKey
-		if isGroup {
-			listKey = policyDBGroupsListKey
-		}
-		for _, item := range iamFilesListing[listKey] {
-			stsUserName := strings.TrimSuffix(item, ".json")
-			entityKeysInStorage.Add(stsUserName)
-		}
-	} else {
-		// For non-iam object store, we copy the mapping keys from the cache.
-		cache := sys.store.rlock()
-		defer sys.store.runlock()
-		cachedPolicyMap := cache.iamSTSPolicyMap
-		if isGroup {
-			cachedPolicyMap = cache.iamGroupPolicyMap
-		}
-		cachedPolicyMap.Range(func(k string, v MappedPolicy) bool {
-			entityKeysInStorage.Add(k)
-			return true
-		})
+	cache := sys.store.rlock()
+	defer sys.store.runlock()
+	cachedPolicyMap := cache.iamSTSPolicyMap
+	if isGroup {
+		cachedPolicyMap = cache.iamGroupPolicyMap
 	}
+	cachedPolicyMap.Range(func(k string, v MappedPolicy) bool {
+		entityKeysInStorage.Add(k)
+		return true
+	})
 
 	return entityKeysInStorage
 }
@@ -1648,7 +1693,7 @@ 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
 		}
@@ -1728,31 +1773,46 @@ func (sys *IAMSys) NormalizeLDAPMappingImport(ctx context.Context, isGroup bool,
 	return nil
 }
 
-// GetUser - get user credentials
-func (sys *IAMSys) GetUser(ctx context.Context, accessKey string) (u UserIdentity, ok bool) {
+// CheckKey validates the incoming accessKey
+func (sys *IAMSys) CheckKey(ctx context.Context, accessKey string) (u UserIdentity, ok bool, err error) {
 	if !sys.Initialized() {
-		return u, false
+		return u, false, nil
 	}
 
 	if accessKey == globalActiveCred.AccessKey {
-		return newUserIdentity(globalActiveCred), true
+		return newUserIdentity(globalActiveCred), true, nil
 	}
 
 	loadUserCalled := false
 	select {
 	case <-sys.configLoaded:
 	default:
-		sys.store.LoadUser(ctx, accessKey)
+		err = sys.store.LoadUser(ctx, accessKey)
 		loadUserCalled = true
 	}
 
 	u, ok = sys.store.GetUser(accessKey)
 	if !ok && !loadUserCalled {
-		sys.store.LoadUser(ctx, accessKey)
+		err = sys.store.LoadUser(ctx, accessKey)
+		loadUserCalled = true
+
 		u, ok = sys.store.GetUser(accessKey)
 	}
 
-	return u, ok && u.Credentials.IsValid()
+	if !ok && loadUserCalled && err != nil {
+		iamLogOnceIf(ctx, err, accessKey)
+
+		// return 503 to application
+		return u, false, errIAMNotInitialized
+	}
+
+	return u, ok && u.Credentials.IsValid(), nil
+}
+
+// GetUser - get user credentials
+func (sys *IAMSys) GetUser(ctx context.Context, accessKey string) (u UserIdentity, ok bool) {
+	u, ok, _ = sys.CheckKey(ctx, accessKey)
+	return u, ok
 }
 
 // Notify all other MinIO peers to load group.
@@ -1976,7 +2036,7 @@ 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
@@ -1986,20 +2046,22 @@ func (sys *IAMSys) PolicyDBUpdateLDAP(ctx context.Context, isAttach bool,
 		}
 		isGroup = false
 	} else {
-		if isAttach {
-			var underBaseDN bool
-			if dnResult, underBaseDN, err = sys.LDAPConfig.GetValidatedGroupDN(nil, r.Group); err != nil {
-				iamLogIf(ctx, err)
-				return
-			} else if dnResult == nil || !underBaseDN {
+		var underBaseDN bool
+		if dnResult, underBaseDN, err = sys.LDAPConfig.GetValidatedGroupDN(nil, r.Group); err != nil {
+			iamLogIf(ctx, err)
+			return
+		}
+		if dnResult == nil || !underBaseDN {
+			if !isAttach {
+				dn = sys.LDAPConfig.QuickNormalizeDN(r.Group)
+			} else {
 				err = errNoSuchGroup
 				return
 			}
+		} else {
 			// We use the group DN returned by the LDAP server (this may not
 			// equal the input group name, but we assume it is canonical).
 			dn = dnResult.NormDN
-		} else {
-			dn = r.Group
 		}
 		isGroup = true
 	}
@@ -2126,7 +2188,7 @@ 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
 		}
@@ -2218,22 +2280,16 @@ 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.
@@ -2354,7 +2410,7 @@ 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

@@ -24,6 +24,7 @@ import (
 
 	"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 +57,7 @@ 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 +84,7 @@ 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 +115,7 @@ 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 +160,20 @@ 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 +185,9 @@ 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 +197,7 @@ 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{
+	allKeyNames, _, err := GlobalKMS.ListKeyNames(ctx, &kms.ListRequest{
 		Prefix: r.Form.Get("pattern"),
 	})
 	if err != nil {
@@ -188,8 +205,24 @@ func (a kmsAPIHandlers) KMSListKeysHandler(w http.ResponseWriter, r *http.Reques
 		return
 	}
 
-	values := make([]kes.KeyInfo, 0, len(names))
-	for _, name := range names {
+	// 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
+	}
+
+	// Now we have all the key names, for each of them, check whether the policy grants permission for
+	// the user to list it.
+	keyNames := []string{}
+	for _, name := range allKeyNames {
+		if checkKMSActionAllowed(r, owner, cred, policy.KMSListKeysAction, name) {
+			keyNames = append(keyNames, name)
+		}
+	}
+
+	values := make([]kes.KeyInfo, 0, len(keyNames))
+	for _, name := range keyNames {
 		values = append(values, kes.KeyInfo{
 			Name: name,
 		})
@@ -224,6 +257,17 @@ 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 +318,16 @@ 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 +1,851 @@
+// 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 {
+		gotKeyNames := keyNamesFromListKeysResp(t, rec.Body.Bytes())
+		if len(test.wantKeyNames) != len(gotKeyNames) {
+			t.Fatalf("want keys len: %d, got len: %d", len(test.wantKeyNames), len(gotKeyNames))
+		}
+		for i, wantKeyName := range test.wantKeyNames {
+			if gotKeyNames[i] != wantKeyName {
+				t.Fatalf("want key name %s, in position %d, got %s", wantKeyName, i, gotKeyNames[i])
+			}
+		}
+	}
+
+	// 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)
+		}
+	}
+}
+
+func keyNamesFromListKeysResp(t *testing.T, b []byte) []string {
+	var keyInfos []madmin.KMSKeyInfo
+	err := json.Unmarshal(b, &keyInfos)
+	if err != nil {
+		t.Fatalf("cannot unmarshal '%s', err: %v", b, err)
+	}
+
+	var gotKeyNames []string
+	for _, keyInfo := range keyInfos {
+		gotKeyNames = append(gotKeyNames, keyInfo.Name)
+	}
+	return gotKeyNames
+}

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)

cmd/lock-rest-client.go

@@ -107,5 +107,5 @@ func newLockAPI(endpoint Endpoint) dsync.NetLocker {
 
 // Returns a lock rest client.
 func newlockRESTClient(ep Endpoint) *lockRESTClient {
-	return &lockRESTClient{globalGrid.Load().Connection(ep.GridHost())}
+	return &lockRESTClient{globalLockGrid.Load().Connection(ep.GridHost())}
 }

cmd/lock-rest-client_test.go

@@ -39,7 +39,7 @@ func TestLockRESTlient(t *testing.T) {
 
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
-	err = initGlobalGrid(ctx, []PoolEndpoints{{Endpoints: Endpoints{endpoint, endpointLocal}}})
+	err = initGlobalLockGrid(ctx, []PoolEndpoints{{Endpoints: Endpoints{endpoint, endpointLocal}}})
 	if err != nil {
 		t.Fatal(err)
 	}

cmd/lock-rest-server-common_test.go

@@ -44,7 +44,7 @@ func createLockTestServer(ctx context.Context, t *testing.T) (string, *lockRESTS
 		},
 	}
 	creds := globalActiveCred
-	token, err := authenticateNode(creds.AccessKey, creds.SecretKey, "")
+	token, err := authenticateNode(creds.AccessKey, creds.SecretKey)
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -63,15 +63,15 @@ func TestLockRpcServerRemoveEntry(t *testing.T) {
 		Owner:           "owner",
 		Writer:          true,
 		UID:             "0123-4567",
-		Timestamp:       UTCNow(),
-		TimeLastRefresh: UTCNow(),
+		Timestamp:       UTCNow().UnixNano(),
+		TimeLastRefresh: UTCNow().UnixNano(),
 	}
 	lockRequesterInfo2 := lockRequesterInfo{
 		Owner:           "owner",
 		Writer:          true,
 		UID:             "89ab-cdef",
-		Timestamp:       UTCNow(),
-		TimeLastRefresh: UTCNow(),
+		Timestamp:       UTCNow().UnixNano(),
+		TimeLastRefresh: UTCNow().UnixNano(),
 	}
 
 	locker.ll.lockMap["name"] = []lockRequesterInfo{

cmd/lock-rest-server.go

@@ -111,17 +111,17 @@ func newLockHandler(h grid.HandlerID) *grid.SingleHandler[*dsync.LockArgs, *dsyn
 }
 
 // registerLockRESTHandlers - register lock rest router.
-func registerLockRESTHandlers() {
+func registerLockRESTHandlers(gm *grid.Manager) {
 	lockServer := &lockRESTServer{
 		ll: newLocker(),
 	}
 
-	logger.FatalIf(lockRPCForceUnlock.Register(globalGrid.Load(), lockServer.ForceUnlockHandler), "unable to register handler")
-	logger.FatalIf(lockRPCRefresh.Register(globalGrid.Load(), lockServer.RefreshHandler), "unable to register handler")
-	logger.FatalIf(lockRPCLock.Register(globalGrid.Load(), lockServer.LockHandler), "unable to register handler")
-	logger.FatalIf(lockRPCUnlock.Register(globalGrid.Load(), lockServer.UnlockHandler), "unable to register handler")
-	logger.FatalIf(lockRPCRLock.Register(globalGrid.Load(), lockServer.RLockHandler), "unable to register handler")
-	logger.FatalIf(lockRPCRUnlock.Register(globalGrid.Load(), lockServer.RUnlockHandler), "unable to register handler")
+	logger.FatalIf(lockRPCForceUnlock.Register(gm, lockServer.ForceUnlockHandler), "unable to register handler")
+	logger.FatalIf(lockRPCRefresh.Register(gm, lockServer.RefreshHandler), "unable to register handler")
+	logger.FatalIf(lockRPCLock.Register(gm, lockServer.LockHandler), "unable to register handler")
+	logger.FatalIf(lockRPCUnlock.Register(gm, lockServer.UnlockHandler), "unable to register handler")
+	logger.FatalIf(lockRPCRLock.Register(gm, lockServer.RLockHandler), "unable to register handler")
+	logger.FatalIf(lockRPCRUnlock.Register(gm, lockServer.RUnlockHandler), "unable to register handler")
 
 	globalLockServer = lockServer.ll
 

cmd/logging.go

@@ -20,6 +20,10 @@ func replLogOnceIf(ctx context.Context, err error, id string, errKind ...interfa
 	logger.LogOnceIf(ctx, "replication", err, id, errKind...)
 }
 
+func iamLogOnceIf(ctx context.Context, err error, id string, errKind ...interface{}) {
+	logger.LogOnceIf(ctx, "iam", err, id, errKind...)
+}
+
 func iamLogIf(ctx context.Context, err error, errKind ...interface{}) {
 	if !errors.Is(err, grid.ErrDisconnected) {
 		logger.LogIf(ctx, "iam", err, errKind...)
@@ -180,10 +184,6 @@ func etcdLogOnceIf(ctx context.Context, err error, id string, errKind ...interfa
 	logger.LogOnceIf(ctx, "etcd", err, id, errKind...)
 }
 
-func subnetLogIf(ctx context.Context, err error, errKind ...interface{}) {
-	logger.LogIf(ctx, "subnet", err, errKind...)
-}
-
 func metricsLogIf(ctx context.Context, err error, errKind ...interface{}) {
 	logger.LogIf(ctx, "metrics", err, errKind...)
 }

cmd/main.go

@@ -189,12 +189,14 @@ func printMinIOVersion(c *cli.Context) {
 	io.Copy(c.App.Writer, versionBanner(c))
 }
 
+var debugNoExit = env.Get("_MINIO_DEBUG_NO_EXIT", "") != ""
+
 // Main main for minio server.
 func Main(args []string) {
 	// Set the minio app name.
 	appName := filepath.Base(args[0])
 
-	if env.Get("_MINIO_DEBUG_NO_EXIT", "") != "" {
+	if debugNoExit {
 		freeze := func(_ int) {
 			// Infinite blocking op
 			<-make(chan struct{})

cmd/metacache-entries.go

@@ -20,7 +20,6 @@ package cmd
 import (
 	"bytes"
 	"context"
-	"errors"
 	"os"
 	"path"
 	"sort"
@@ -260,7 +259,7 @@ func (e *metaCacheEntry) fileInfo(bucket string) (FileInfo, error) {
 		}
 		return e.cached.ToFileInfo(bucket, e.name, "", false, false)
 	}
-	return getFileInfo(e.metadata, bucket, e.name, "", false, false)
+	return getFileInfo(e.metadata, bucket, e.name, "", fileInfoOpts{})
 }
 
 // xlmeta returns the decoded metadata.
@@ -301,7 +300,7 @@ func (e *metaCacheEntry) fileInfoVersions(bucket string) (FileInfoVersions, erro
 		}, nil
 	}
 	// Too small gains to reuse cache here.
-	return getFileInfoVersions(e.metadata, bucket, e.name, false)
+	return getFileInfoVersions(e.metadata, bucket, e.name, false, true)
 }
 
 // metaCacheEntries is a slice of metacache entries.
@@ -385,9 +384,6 @@ func (m metaCacheEntries) resolve(r *metadataResolutionParams) (selected *metaCa
 		// shallow decode.
 		xl, err := entry.xlmeta()
 		if err != nil {
-			if !errors.Is(err, errFileNotFound) {
-				internalLogIf(GlobalContext, err)
-			}
 			continue
 		}
 		objsValid++
@@ -536,6 +532,9 @@ func (m *metaCacheEntriesSorted) fileInfoVersions(bucket, prefix, delimiter, aft
 			}
 
 			for _, version := range fiVersions {
+				if !version.VersionPurgeStatus().Empty() {
+					continue
+				}
 				versioned := vcfg != nil && vcfg.Versioned(entry.name)
 				versions = append(versions, version.ToObjectInfo(bucket, entry.name, versioned))
 			}
@@ -597,7 +596,7 @@ func (m *metaCacheEntriesSorted) fileInfos(bucket, prefix, delimiter string) (ob
 			}
 
 			fi, err := entry.fileInfo(bucket)
-			if err == nil {
+			if err == nil && fi.VersionPurgeStatus().Empty() {
 				versioned := vcfg != nil && vcfg.Versioned(entry.name)
 				objects = append(objects, fi.ToObjectInfo(bucket, entry.name, versioned))
 			}
@@ -752,7 +751,9 @@ func mergeEntryChannels(ctx context.Context, in []chan metaCacheEntry, out chan<
 				if !dirMatches {
 					// We have an object `name` or 'name/' and a directory `name/`.
 					if other.isDir() {
-						console.Debugln("mergeEntryChannels: discarding directory", other.name, "for object", best.name)
+						if serverDebugLog {
+							console.Debugln("mergeEntryChannels: discarding directory", other.name, "for object", best.name)
+						}
 						// Discard the directory.
 						if err := selectFrom(otherIdx); err != nil {
 							return err
@@ -760,7 +761,9 @@ func mergeEntryChannels(ctx context.Context, in []chan metaCacheEntry, out chan<
 						continue
 					}
 					// Replace directory with object.
-					console.Debugln("mergeEntryChannels: discarding directory", best.name, "for object", other.name)
+					if serverDebugLog {
+						console.Debugln("mergeEntryChannels: discarding directory", best.name, "for object", other.name)
+					}
 					toMerge = toMerge[:0]
 					best = other
 					bestIdx = otherIdx

cmd/metacache-server-pool.go

@@ -153,19 +153,7 @@ func (z *erasureServerPools) listPath(ctx context.Context, o *listPathOptions) (
 			} else {
 				// Continue listing
 				o.ID = c.id
-				go func(meta metacache) {
-					// Continuously update while we wait.
-					t := time.NewTicker(metacacheMaxClientWait / 10)
-					defer t.Stop()
-					select {
-					case <-ctx.Done():
-						// Request is done, stop updating.
-						return
-					case <-t.C:
-						meta.lastHandout = time.Now()
-						meta, _ = rpc.UpdateMetacacheListing(ctx, meta)
-					}
-				}(*c)
+				go c.keepAlive(ctx, rpc)
 			}
 		}
 	}
@@ -219,6 +207,9 @@ func (z *erasureServerPools) listPath(ctx context.Context, o *listPathOptions) (
 		o.ID = ""
 	}
 
+	if contextCanceled(ctx) {
+		return entries, ctx.Err()
+	}
 	// Do listing in-place.
 	// Create output for our results.
 	// Create filter for results.
@@ -449,5 +440,10 @@ func (z *erasureServerPools) listAndSave(ctx context.Context, o *listPathOptions
 		xioutil.SafeClose(saveCh)
 	}()
 
-	return filteredResults()
+	entries, err = filteredResults()
+	if err == nil {
+		// Check if listing recorded an error.
+		err = meta.getErr()
+	}
+	return entries, err
 }

cmd/metacache-set.go

@@ -391,7 +391,7 @@ func (r *metacacheReader) filter(o listPathOptions) (entries metaCacheEntriesSor
 			if !o.InclDeleted && entry.isObject() && entry.isLatestDeletemarker() && !entry.isObjectDir() {
 				return true
 			}
-			if entry.isAllFreeVersions() {
+			if !o.InclDeleted && entry.isAllFreeVersions() {
 				return true
 			}
 			entries.o = append(entries.o, entry)
@@ -805,6 +805,17 @@ func (m *metaCacheRPC) setErr(err string) {
 	*m.meta = meta
 }
 
+// getErr will return an error if the listing failed.
+// The error is not type safe.
+func (m *metaCacheRPC) getErr() error {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	if m.meta.status == scanStateError {
+		return errors.New(m.meta.error)
+	}
+	return nil
+}
+
 func (er *erasureObjects) saveMetaCacheStream(ctx context.Context, mc *metaCacheRPC, entries <-chan metaCacheEntry) (err error) {
 	o := mc.o
 	o.debugf(color.Green("saveMetaCacheStream:")+" with options: %#v", o)

cmd/metacache.go

@@ -24,6 +24,8 @@ import (
 	"path"
 	"strings"
 	"time"
+
+	"github.com/minio/pkg/v3/console"
 )
 
 type scanStatus uint8
@@ -97,6 +99,37 @@ func (m *metacache) worthKeeping() bool {
 	return true
 }
 
+// keepAlive will continuously update lastHandout until ctx is canceled.
+func (m metacache) keepAlive(ctx context.Context, rpc *peerRESTClient) {
+	// we intentionally operate on a copy of m, so we can update without locks.
+	t := time.NewTicker(metacacheMaxClientWait / 10)
+	defer t.Stop()
+	for {
+		select {
+		case <-ctx.Done():
+			// Request is done, stop updating.
+			return
+		case <-t.C:
+			m.lastHandout = time.Now()
+
+			if m2, err := rpc.UpdateMetacacheListing(ctx, m); err == nil {
+				if m2.status != scanStateStarted {
+					if serverDebugLog {
+						console.Debugln("returning", m.id, "due to scan state", m2.status, time.Now().Format(time.RFC3339))
+					}
+					return
+				}
+				m = m2
+				if serverDebugLog {
+					console.Debugln("refreshed", m.id, time.Now().Format(time.RFC3339))
+				}
+			} else if serverDebugLog {
+				console.Debugln("error refreshing", m.id, time.Now().Format(time.RFC3339))
+			}
+		}
+	}
+}
+
 // baseDirFromPrefix will return the base directory given an object path.
 // For example an object with name prefix/folder/object.ext will return `prefix/folder/`.
 func baseDirFromPrefix(prefix string) string {
@@ -116,13 +149,17 @@ func baseDirFromPrefix(prefix string) string {
 // update cache with new status.
 // The updates are conditional so multiple callers can update with different states.
 func (m *metacache) update(update metacache) {
-	m.lastUpdate = UTCNow()
+	now := UTCNow()
+	m.lastUpdate = now
 
-	if m.lastHandout.After(m.lastHandout) {
-		m.lastHandout = UTCNow()
+	if update.lastHandout.After(m.lastHandout) {
+		m.lastHandout = update.lastUpdate
+		if m.lastHandout.After(now) {
+			m.lastHandout = now
+		}
 	}
 	if m.status == scanStateStarted && update.status == scanStateSuccess {
-		m.ended = UTCNow()
+		m.ended = now
 	}
 
 	if m.status == scanStateStarted && update.status != scanStateStarted {
@@ -138,7 +175,7 @@ func (m *metacache) update(update metacache) {
 	if m.error == "" && update.error != "" {
 		m.error = update.error
 		m.status = scanStateError
-		m.ended = UTCNow()
+		m.ended = now
 	}
 	m.fileNotFound = m.fileNotFound || update.fileNotFound
 }

cmd/metrics-resource.go

@@ -262,7 +262,7 @@ func collectDriveMetrics(m madmin.RealtimeMetrics) {
 	latestDriveStatsMu.Unlock()
 
 	globalLocalDrivesMu.RLock()
-	localDrives := cloneDrives(globalLocalDrives)
+	localDrives := cloneDrives(globalLocalDrivesMap)
 	globalLocalDrivesMu.RUnlock()
 
 	for _, d := range localDrives {

cmd/metrics-router.go

@@ -38,7 +38,8 @@ const (
 
 // Standard env prometheus auth type
 const (
-	EnvPrometheusAuthType = "MINIO_PROMETHEUS_AUTH_TYPE"
+	EnvPrometheusAuthType    = "MINIO_PROMETHEUS_AUTH_TYPE"
+	EnvPrometheusOpenMetrics = "MINIO_PROMETHEUS_OPEN_METRICS"
 )
 
 type prometheusAuthType string
@@ -58,14 +59,15 @@ func registerMetricsRouter(router *mux.Router) {
 	if authType == prometheusPublic {
 		auth = NoAuthMiddleware
 	}
+
 	metricsRouter.Handle(prometheusMetricsPathLegacy, auth(metricsHandler()))
 	metricsRouter.Handle(prometheusMetricsV2ClusterPath, auth(metricsServerHandler()))
 	metricsRouter.Handle(prometheusMetricsV2BucketPath, auth(metricsBucketHandler()))
 	metricsRouter.Handle(prometheusMetricsV2NodePath, auth(metricsNodeHandler()))
 	metricsRouter.Handle(prometheusMetricsV2ResourcePath, auth(metricsResourceHandler()))
 
-	// Metrics v3!
-	metricsV3Server := newMetricsV3Server(authType)
+	// Metrics v3
+	metricsV3Server := newMetricsV3Server(auth)
 
 	// Register metrics v3 handler. It also accepts an optional query
 	// parameter `?list` - see handler for details.