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

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

.github/workflows/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/mint.yml

@@ -55,9 +55,10 @@ jobs:
         run: |
           ${GITHUB_WORKSPACE}/.github/workflows/run-mint.sh "erasure" "minio" "minio123" "${{ steps.vars.outputs.sha_short }}"
 
-      - name: resiliency
-        run: |
-          ${GITHUB_WORKSPACE}/.github/workflows/run-mint.sh "resiliency" "minio" "minio123" "${{ steps.vars.outputs.sha_short }}"
+      # FIXME: renable this back when we have a valid way to add deadlines for PUT()s (internode CreateFile)
+      # - name: resiliency
+      #   run: |
+      #     ${GITHUB_WORKSPACE}/.github/workflows/run-mint.sh "resiliency" "minio" "minio123" "${{ steps.vars.outputs.sha_short }}"
 
       - name: The job must cleanup
         if: ${{ always() }}

.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]

CONTRIBUTING.md

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

Dockerfile.hotfix

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

Dockerfile.release

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

Dockerfile.release.fips

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

Dockerfile.release.old_cpu

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

Makefile

@@ -34,11 +34,14 @@ verifiers: lint check-gen
 
 check-gen: ## check for updated autogenerated files
 	@go generate ./... >/dev/null
+	@go mod tidy -compat=1.21
 	@(! git diff --name-only | grep '_gen.go$$') || (echo "Non-committed changes in auto-generated code is detected, please commit them to proceed." && false)
+	@(! git diff --name-only | grep 'go.sum') || (echo "Non-committed changes in auto-generated go.sum is detected, please commit them to proceed." && false)
 
 lint: getdeps ## runs golangci-lint suite of linters
 	@echo "Running $@ check"
 	@$(GOLANGCI) run --build-tags kqueue --timeout=10m --config ./.golangci.yml
+	@command typos && typos ./ || echo "typos binary is not found.. skipping.."
 
 lint-fix: getdeps ## runs golangci-lint suite of linters with automatic fixes
 	@echo "Running $@ check"
@@ -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
 
@@ -106,7 +109,10 @@ test-replication-3site:
 test-delete-replication:
 	@(env bash $(PWD)/docs/bucket/replication/delete-replication.sh)
 
-test-replication: install-race test-replication-2site test-replication-3site test-delete-replication test-sio-error ## verify multi site replication
+test-delete-marker-proxying:
+	@(env bash $(PWD)/docs/bucket/replication/test_del_marker_proxying.sh)
+
+test-replication: install-race test-replication-2site test-replication-3site test-delete-replication test-sio-error test-delete-marker-proxying ## verify multi site replication
 	@echo "Running tests for replicating three sites"
 
 test-site-replication-ldap: install-race ## verify automatic site replication
@@ -162,9 +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)
@@ -193,7 +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

@@ -38,40 +38,47 @@ 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
+
+	# 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
+	# Wait for all drives to be healed
+	while [ $(/tmp/mc admin info --json myminio | jq '.info.servers[].drives[].healing | select(. != null) | select(. == true)' | wc -l) -gt 0 ]; do sleep 1; done
+
+	# Wait for Status: in MinIO output
+	while true; do
+		rv=$(check_online)
+		if [ "$rv" != "1" ]; then
+			# success
+			break
+		fi
+
+		# Check if we should retry
+		retry=$((retry + 1))
+		if [ $retry -le 20 ]; then
+			sleep 5
+			continue
+		fi
+
+		# Failure
+		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
@@ -83,14 +90,24 @@ 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
 }
 
 function purge() {
-	rm -rf "$1"
+	echo rm -rf "$1"
 }
 
 function __init__() {
@@ -117,18 +134,6 @@ function perform_test() {
 
 	set -x
 	start_minio_3_node $2
-
-	rv=$(check_online)
-	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
-	fi
 }
 
 function main() {

buildscripts/verify-healing.sh

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

cmd/admin-bucket-handlers.go

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

cmd/admin-handler-utils.go

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

cmd/admin-handlers-idp-ldap.go

@@ -20,6 +20,7 @@ package cmd
 import (
 	"encoding/json"
 	"errors"
+	"fmt"
 	"io"
 	"net/http"
 	"strings"
@@ -284,6 +285,18 @@ func (a adminAPIHandlers) AddServiceAccountLDAP(w http.ResponseWriter, r *http.R
 		opts.claims[ldapUser] = targetUser // DN
 		opts.claims[ldapActualUser] = lookupResult.ActualDN
 
+		// Check if this user or their groups have a policy applied.
+		ldapPolicies, err := globalIAMSys.PolicyDBGet(targetUser, targetGroups...)
+		if err != nil {
+			writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
+			return
+		}
+		if len(ldapPolicies) == 0 {
+			err = fmt.Errorf("No policy set for user `%s` or any of their groups: `%s`", opts.claims[ldapActualUser], strings.Join(targetGroups, "`,`"))
+			writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErrWithErr(ErrAdminNoSuchUser, err), r.URL)
+			return
+		}
+
 		// Add LDAP attributes that were looked up into the claims.
 		for attribKey, attribValue := range lookupResult.Attributes {
 			opts.claims[ldapAttribPrefix+attribKey] = attribValue
@@ -466,3 +479,180 @@ func (a adminAPIHandlers) ListAccessKeysLDAP(w http.ResponseWriter, r *http.Requ
 
 	writeSuccessResponseJSON(w, encryptedData)
 }
+
+// ListAccessKeysLDAPBulk - GET /minio/admin/v3/idp/ldap/list-access-keys-bulk
+func (a adminAPIHandlers) ListAccessKeysLDAPBulk(w http.ResponseWriter, r *http.Request) {
+	ctx := r.Context()
+
+	// Get current object layer instance.
+	objectAPI := newObjectLayerFn()
+	if objectAPI == nil || globalNotificationSys == nil {
+		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrServerNotInitialized), r.URL)
+		return
+	}
+
+	cred, owner, s3Err := validateAdminSignature(ctx, r, "")
+	if s3Err != ErrNone {
+		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(s3Err), r.URL)
+		return
+	}
+
+	dnList := r.Form["userDNs"]
+	isAll := r.Form.Get("all") == "true"
+	onlySelf := !isAll && len(dnList) == 0
+
+	if isAll && len(dnList) > 0 {
+		// This should be checked on client side, so return generic error
+		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrInvalidRequest), r.URL)
+		return
+	}
+
+	// Empty DN list and not self, list access keys for all users
+	if isAll {
+		if !globalIAMSys.IsAllowed(policy.Args{
+			AccountName:     cred.AccessKey,
+			Groups:          cred.Groups,
+			Action:          policy.ListUsersAdminAction,
+			ConditionValues: getConditionValues(r, "", cred),
+			IsOwner:         owner,
+			Claims:          cred.Claims,
+		}) {
+			writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrAccessDenied), r.URL)
+			return
+		}
+	} else if len(dnList) == 1 {
+		var dn string
+		foundResult, err := globalIAMSys.LDAPConfig.GetValidatedDNForUsername(dnList[0])
+		if err == nil {
+			dn = foundResult.NormDN
+		}
+		if dn == cred.ParentUser || dnList[0] == cred.ParentUser {
+			onlySelf = true
+		}
+	}
+
+	if !globalIAMSys.IsAllowed(policy.Args{
+		AccountName:     cred.AccessKey,
+		Groups:          cred.Groups,
+		Action:          policy.ListServiceAccountsAdminAction,
+		ConditionValues: getConditionValues(r, "", cred),
+		IsOwner:         owner,
+		Claims:          cred.Claims,
+		DenyOnly:        onlySelf,
+	}) {
+		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrAccessDenied), r.URL)
+		return
+	}
+
+	if onlySelf && len(dnList) == 0 {
+		selfDN := cred.AccessKey
+		if cred.ParentUser != "" {
+			selfDN = cred.ParentUser
+		}
+		dnList = append(dnList, selfDN)
+	}
+
+	var ldapUserList []string
+	if isAll {
+		ldapUsers, err := globalIAMSys.ListLDAPUsers(ctx)
+		if err != nil {
+			writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
+			return
+		}
+		for user := range ldapUsers {
+			ldapUserList = append(ldapUserList, user)
+		}
+	} else {
+		for _, userDN := range dnList {
+			// Validate the userDN
+			foundResult, err := globalIAMSys.LDAPConfig.GetValidatedDNForUsername(userDN)
+			if err != nil {
+				writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
+				return
+			}
+			if foundResult == nil {
+				continue
+			}
+			ldapUserList = append(ldapUserList, foundResult.NormDN)
+		}
+	}
+
+	listType := r.Form.Get("listType")
+	var listSTSKeys, listServiceAccounts bool
+	switch listType {
+	case madmin.AccessKeyListUsersOnly:
+		listSTSKeys = false
+		listServiceAccounts = false
+	case madmin.AccessKeyListSTSOnly:
+		listSTSKeys = true
+		listServiceAccounts = false
+	case madmin.AccessKeyListSvcaccOnly:
+		listSTSKeys = false
+		listServiceAccounts = true
+	case madmin.AccessKeyListAll:
+		listSTSKeys = true
+		listServiceAccounts = true
+	default:
+		err := errors.New("invalid list type")
+		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErrWithErr(ErrInvalidRequest, err), r.URL)
+		return
+	}
+
+	accessKeyMap := make(map[string]madmin.ListAccessKeysLDAPResp)
+	for _, internalDN := range ldapUserList {
+		externalDN := globalIAMSys.LDAPConfig.DecodeDN(internalDN)
+		accessKeys := madmin.ListAccessKeysLDAPResp{}
+		if listSTSKeys {
+			stsKeys, err := globalIAMSys.ListSTSAccounts(ctx, internalDN)
+			if err != nil {
+				writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
+				return
+			}
+			for _, sts := range stsKeys {
+				expiryTime := sts.Expiration
+				accessKeys.STSKeys = append(accessKeys.STSKeys, madmin.ServiceAccountInfo{
+					AccessKey:  sts.AccessKey,
+					Expiration: &expiryTime,
+				})
+			}
+			// if only STS keys, skip if user has no STS keys
+			if !listServiceAccounts && len(stsKeys) == 0 {
+				continue
+			}
+		}
+
+		if listServiceAccounts {
+			serviceAccounts, err := globalIAMSys.ListServiceAccounts(ctx, internalDN)
+			if err != nil {
+				writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
+				return
+			}
+			for _, svc := range serviceAccounts {
+				expiryTime := svc.Expiration
+				accessKeys.ServiceAccounts = append(accessKeys.ServiceAccounts, madmin.ServiceAccountInfo{
+					AccessKey:  svc.AccessKey,
+					Expiration: &expiryTime,
+				})
+			}
+			// if only service accounts, skip if user has no service accounts
+			if !listSTSKeys && len(serviceAccounts) == 0 {
+				continue
+			}
+		}
+		accessKeyMap[externalDN] = accessKeys
+	}
+
+	data, err := json.Marshal(accessKeyMap)
+	if err != nil {
+		writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
+		return
+	}
+
+	encryptedData, err := madmin.EncryptData(cred.SecretKey, data)
+	if err != nil {
+		writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
+		return
+	}
+
+	writeSuccessResponseJSON(w, encryptedData)
+}

cmd/admin-handlers-pools.go

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

cmd/admin-handlers-users-race_test.go

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

cmd/admin-handlers-users.go

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

cmd/admin-handlers-users_test.go

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

cmd/admin-handlers.go

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

cmd/admin-handlers_test.go

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

cmd/admin-heal-ops.go

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

cmd/admin-router.go

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

cmd/api-errors.go

@@ -28,7 +28,7 @@ import (
 	"strconv"
 	"strings"
 
-	"github.com/Azure/azure-storage-blob-go/azblob"
+	"github.com/Azure/azure-sdk-for-go/sdk/azcore"
 	"github.com/minio/minio/internal/ioutil"
 	"google.golang.org/api/googleapi"
 
@@ -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.",
@@ -1479,7 +1486,7 @@ var errorCodes = errorCodeMap{
 	},
 	ErrTooManyRequests: {
 		Code:           "TooManyRequests",
-		Description:    "Deadline exceeded while waiting in incoming queue, please reduce your request rate",
+		Description:    "Please reduce your request rate",
 		HTTPStatusCode: http.StatusTooManyRequests,
 	},
 	ErrUnsupportedMetadata: {
@@ -2542,11 +2549,11 @@ func toAPIError(ctx context.Context, err error) APIError {
 			if len(e.Errors) >= 1 {
 				apiErr.Code = e.Errors[0].Reason
 			}
-		case azblob.StorageError:
+		case *azcore.ResponseError:
 			apiErr = APIError{
-				Code:           string(e.ServiceCode()),
+				Code:           e.ErrorCode,
 				Description:    e.Error(),
-				HTTPStatusCode: e.Response().StatusCode,
+				HTTPStatusCode: e.StatusCode,
 			}
 			// Add more other SDK related errors here if any in future.
 		default:
@@ -2585,7 +2592,7 @@ func getAPIError(code APIErrorCode) APIError {
 	return errorCodes.ToAPIErr(ErrInternalError)
 }
 
-// getErrorResponse gets in standard error and resource value and
+// getAPIErrorResponse gets in standard error and resource value and
 // provides a encodable populated response values
 func getAPIErrorResponse(ctx context.Context, err APIError, resource, requestID, hostID string) APIErrorResponse {
 	reqInfo := logger.GetReqInfo(ctx)

cmd/api-response.go

@@ -789,8 +789,8 @@ func generateInitiateMultipartUploadResponse(bucket, key, uploadID string) Initi
 }
 
 // generates CompleteMultipartUploadResponse for given bucket, key, location and ETag.
-func generateCompleteMultpartUploadResponse(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 +946,11 @@ func writeSuccessResponseHeadersOnly(w http.ResponseWriter) {
 
 // writeErrorResponse writes error headers
 func writeErrorResponse(ctx context.Context, w http.ResponseWriter, err APIError, reqURL *url.URL) {
-	if err.HTTPStatusCode == http.StatusServiceUnavailable {
-		// Set retry-after header to indicate user-agents to retry request after 120secs.
+	switch err.HTTPStatusCode {
+	case http.StatusServiceUnavailable, http.StatusTooManyRequests:
+		// Set retry-after header to indicate user-agents to retry request after 60 seconds.
 		// https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Retry-After
-		w.Header().Set(xhttp.RetryAfter, "120")
+		w.Header().Set(xhttp.RetryAfter, "60")
 	}
 
 	switch err.Code {

cmd/api-router.go

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

cmd/background-newdisks-heal-ops.go

@@ -72,10 +72,12 @@ type healingTracker struct {
 
 	// Numbers when current bucket started healing,
 	// for resuming with correct numbers.
-	ResumeItemsHealed uint64 `json:"-"`
-	ResumeItemsFailed uint64 `json:"-"`
-	ResumeBytesDone   uint64 `json:"-"`
-	ResumeBytesFailed uint64 `json:"-"`
+	ResumeItemsHealed  uint64 `json:"-"`
+	ResumeItemsFailed  uint64 `json:"-"`
+	ResumeItemsSkipped uint64 `json:"-"`
+	ResumeBytesDone    uint64 `json:"-"`
+	ResumeBytesFailed  uint64 `json:"-"`
+	ResumeBytesSkipped uint64 `json:"-"`
 
 	// Filled on startup/restarts.
 	QueuedBuckets []string
@@ -88,6 +90,11 @@ type healingTracker struct {
 
 	ItemsSkipped uint64
 	BytesSkipped uint64
+
+	RetryAttempts uint64
+
+	Finished bool // finished healing, whether with errors or not
+
 	// Add future tracking capabilities
 	// Be sure that they are included in toHealingDisk
 }
@@ -141,14 +148,34 @@ func initHealingTracker(disk StorageAPI, healID string) *healingTracker {
 	return h
 }
 
-func (h healingTracker) getLastUpdate() time.Time {
+func (h *healingTracker) resetHealing() {
+	h.mu.Lock()
+	defer h.mu.Unlock()
+
+	h.ItemsHealed = 0
+	h.ItemsFailed = 0
+	h.BytesDone = 0
+	h.BytesFailed = 0
+	h.ResumeItemsHealed = 0
+	h.ResumeItemsFailed = 0
+	h.ResumeBytesDone = 0
+	h.ResumeBytesFailed = 0
+	h.ItemsSkipped = 0
+	h.BytesSkipped = 0
+
+	h.HealedBuckets = nil
+	h.Object = ""
+	h.Bucket = ""
+}
+
+func (h *healingTracker) getLastUpdate() time.Time {
 	h.mu.RLock()
 	defer h.mu.RUnlock()
 
 	return h.LastUpdate
 }
 
-func (h healingTracker) getBucket() string {
+func (h *healingTracker) getBucket() string {
 	h.mu.RLock()
 	defer h.mu.RUnlock()
 
@@ -162,7 +189,7 @@ func (h *healingTracker) setBucket(bucket string) {
 	h.Bucket = bucket
 }
 
-func (h healingTracker) getObject() string {
+func (h *healingTracker) getObject() string {
 	h.mu.RLock()
 	defer h.mu.RUnlock()
 
@@ -260,8 +287,10 @@ func (h *healingTracker) resume() {
 
 	h.ItemsHealed = h.ResumeItemsHealed
 	h.ItemsFailed = h.ResumeItemsFailed
+	h.ItemsSkipped = h.ResumeItemsSkipped
 	h.BytesDone = h.ResumeBytesDone
 	h.BytesFailed = h.ResumeBytesFailed
+	h.BytesSkipped = h.ResumeBytesSkipped
 }
 
 // bucketDone should be called when a bucket is done healing.
@@ -272,8 +301,10 @@ func (h *healingTracker) bucketDone(bucket string) {
 
 	h.ResumeItemsHealed = h.ItemsHealed
 	h.ResumeItemsFailed = h.ItemsFailed
+	h.ResumeItemsSkipped = h.ItemsSkipped
 	h.ResumeBytesDone = h.BytesDone
 	h.ResumeBytesFailed = h.BytesFailed
+	h.ResumeBytesSkipped = h.BytesSkipped
 	h.HealedBuckets = append(h.HealedBuckets, bucket)
 	for i, b := range h.QueuedBuckets {
 		if b == bucket {
@@ -322,6 +353,7 @@ func (h *healingTracker) toHealingDisk() madmin.HealingDisk {
 		PoolIndex:         h.PoolIndex,
 		SetIndex:          h.SetIndex,
 		DiskIndex:         h.DiskIndex,
+		Finished:          h.Finished,
 		Path:              h.Path,
 		Started:           h.Started.UTC(),
 		LastUpdate:        h.LastUpdate.UTC(),
@@ -337,6 +369,7 @@ func (h *healingTracker) toHealingDisk() madmin.HealingDisk {
 		Object:            h.Object,
 		QueuedBuckets:     h.QueuedBuckets,
 		HealedBuckets:     h.HealedBuckets,
+		RetryAttempts:     h.RetryAttempts,
 
 		ObjectsHealed: h.ItemsHealed, // Deprecated July 2021
 		ObjectsFailed: h.ItemsFailed, // Deprecated July 2021
@@ -351,24 +384,26 @@ func initAutoHeal(ctx context.Context, objAPI ObjectLayer) {
 	}
 
 	initBackgroundHealing(ctx, objAPI) // start quick background healing
-
-	if env.Get("_MINIO_AUTO_DRIVE_HEALING", config.EnableOn) == config.EnableOn || env.Get("_MINIO_AUTO_DISK_HEALING", config.EnableOn) == config.EnableOn {
+	if env.Get("_MINIO_AUTO_DRIVE_HEALING", config.EnableOn) == config.EnableOn {
 		globalBackgroundHealState.pushHealLocalDisks(getLocalDisksToHeal()...)
 		go monitorLocalDisksAndHeal(ctx, z)
 	}
+
+	go globalMRFState.startMRFPersistence()
+	go globalMRFState.healRoutine(z)
 }
 
 func getLocalDisksToHeal() (disksToHeal Endpoints) {
 	globalLocalDrivesMu.RLock()
-	localDrives := cloneDrives(globalLocalDrives)
+	localDrives := cloneDrives(globalLocalDrivesMap)
 	globalLocalDrivesMu.RUnlock()
 	for _, disk := range localDrives {
-		_, err := disk.GetDiskID()
+		_, err := disk.DiskInfo(context.Background(), DiskInfoOptions{})
 		if errors.Is(err, errUnformattedDisk) {
 			disksToHeal = append(disksToHeal, disk.Endpoint())
 			continue
 		}
-		if disk.Healing() != nil {
+		if h := disk.Healing(); h != nil && !h.Finished {
 			disksToHeal = append(disksToHeal, disk.Endpoint())
 		}
 	}
@@ -382,6 +417,8 @@ func getLocalDisksToHeal() (disksToHeal Endpoints) {
 
 var newDiskHealingTimeout = newDynamicTimeout(30*time.Second, 10*time.Second)
 
+var errRetryHealing = errors.New("some items failed to heal, we will retry healing this drive again")
+
 func healFreshDisk(ctx context.Context, z *erasureServerPools, endpoint Endpoint) error {
 	poolIdx, setIdx := endpoint.PoolIdx, endpoint.SetIdx
 	disk := getStorageViaEndpoint(endpoint)
@@ -389,6 +426,17 @@ func healFreshDisk(ctx context.Context, z *erasureServerPools, endpoint Endpoint
 		return fmt.Errorf("Unexpected error disk must be initialized by now after formatting: %s", endpoint)
 	}
 
+	_, err := disk.DiskInfo(ctx, DiskInfoOptions{})
+	if err != nil {
+		if errors.Is(err, errDriveIsRoot) {
+			// This is a root drive, ignore and move on
+			return nil
+		}
+		if !errors.Is(err, errUnformattedDisk) {
+			return err
+		}
+	}
+
 	// Prevent parallel erasure set healing
 	locker := z.NewNSLock(minioMetaBucket, fmt.Sprintf("new-drive-healing/%d/%d", poolIdx, setIdx))
 	lkctx, err := locker.GetLock(ctx, newDiskHealingTimeout)
@@ -451,12 +499,30 @@ func healFreshDisk(ctx context.Context, z *erasureServerPools, endpoint Endpoint
 		return err
 	}
 
-	healingLogEvent(ctx, "Healing of drive '%s' is finished (healed: %d, skipped: %d, failed: %d).", disk, tracker.ItemsHealed, tracker.ItemsSkipped, tracker.ItemsFailed)
+	// if objects have failed healing, we attempt a retry to heal the drive upto 3 times before giving up.
+	if tracker.ItemsFailed > 0 && tracker.RetryAttempts < 4 {
+		tracker.RetryAttempts++
 
-	if len(tracker.QueuedBuckets) > 0 {
-		return fmt.Errorf("not all buckets were healed: %v", tracker.QueuedBuckets)
+		healingLogEvent(ctx, "Healing of drive '%s' is incomplete, retrying %s time (healed: %d, skipped: %d, failed: %d).", disk,
+			humanize.Ordinal(int(tracker.RetryAttempts)), tracker.ItemsHealed, tracker.ItemsSkipped, tracker.ItemsFailed)
+
+		tracker.resetHealing()
+		bugLogIf(ctx, tracker.update(ctx))
+
+		return errRetryHealing
 	}
 
+	if tracker.ItemsFailed > 0 {
+		healingLogEvent(ctx, "Healing of drive '%s' is incomplete, retried %d times (healed: %d, skipped: %d, failed: %d).", disk,
+			tracker.RetryAttempts, tracker.ItemsHealed, tracker.ItemsSkipped, tracker.ItemsFailed)
+	} else {
+		if tracker.RetryAttempts > 0 {
+			healingLogEvent(ctx, "Healing of drive '%s' is complete, retried %d times (healed: %d, skipped: %d).", disk,
+				tracker.RetryAttempts-1, tracker.ItemsHealed, tracker.ItemsSkipped)
+		} else {
+			healingLogEvent(ctx, "Healing of drive '%s' is finished (healed: %d, skipped: %d).", disk, tracker.ItemsHealed, tracker.ItemsSkipped)
+		}
+	}
 	if serverDebugLog {
 		tracker.printTo(os.Stdout)
 		fmt.Printf("\n")
@@ -486,7 +552,8 @@ func healFreshDisk(ctx context.Context, z *erasureServerPools, endpoint Endpoint
 			continue
 		}
 		if t.HealID == tracker.HealID {
-			t.delete(ctx)
+			t.Finished = true
+			t.update(ctx)
 		}
 	}
 
@@ -528,7 +595,7 @@ func monitorLocalDisksAndHeal(ctx context.Context, z *erasureServerPools) {
 					if err := healFreshDisk(ctx, z, disk); err != nil {
 						globalBackgroundHealState.setDiskHealingStatus(disk, false)
 						timedout := OperationTimedOut{}
-						if !errors.Is(err, context.Canceled) && !errors.As(err, &timedout) {
+						if !errors.Is(err, context.Canceled) && !errors.As(err, &timedout) && !errors.Is(err, errRetryHealing) {
 							printEndpointError(disk, err, false)
 						}
 						return

cmd/background-newdisks-heal-ops_gen.go

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

cmd/batch-expire.go

@@ -27,6 +27,7 @@ import (
 	"net/http"
 	"runtime"
 	"strconv"
+	"strings"
 	"time"
 
 	"github.com/minio/minio-go/v7/pkg/tags"
@@ -36,6 +37,7 @@ import (
 	"github.com/minio/pkg/v3/env"
 	"github.com/minio/pkg/v3/wildcard"
 	"github.com/minio/pkg/v3/workers"
+	"github.com/minio/pkg/v3/xtime"
 	"gopkg.in/yaml.v3"
 )
 
@@ -116,7 +118,7 @@ func (p BatchJobExpirePurge) Validate() error {
 // BatchJobExpireFilter holds all the filters currently supported for batch replication
 type BatchJobExpireFilter struct {
 	line, col     int
-	OlderThan     time.Duration       `yaml:"olderThan,omitempty" json:"olderThan"`
+	OlderThan     xtime.Duration      `yaml:"olderThan,omitempty" json:"olderThan"`
 	CreatedBefore *time.Time          `yaml:"createdBefore,omitempty" json:"createdBefore"`
 	Tags          []BatchJobKV        `yaml:"tags,omitempty" json:"tags"`
 	Metadata      []BatchJobKV        `yaml:"metadata,omitempty" json:"metadata"`
@@ -162,7 +164,7 @@ func (ef BatchJobExpireFilter) Matches(obj ObjectInfo, now time.Time) bool {
 	if len(ef.Name) > 0 && !wildcard.Match(ef.Name, obj.Name) {
 		return false
 	}
-	if ef.OlderThan > 0 && now.Sub(obj.ModTime) <= ef.OlderThan {
+	if ef.OlderThan > 0 && now.Sub(obj.ModTime) <= ef.OlderThan.D() {
 		return false
 	}
 
@@ -280,7 +282,7 @@ type BatchJobExpire struct {
 	line, col       int
 	APIVersion      string                 `yaml:"apiVersion" json:"apiVersion"`
 	Bucket          string                 `yaml:"bucket" json:"bucket"`
-	Prefix          string                 `yaml:"prefix" json:"prefix"`
+	Prefix          BatchJobPrefix         `yaml:"prefix" json:"prefix"`
 	NotificationCfg BatchJobNotification   `yaml:"notify" json:"notify"`
 	Retry           BatchJobRetry          `yaml:"retry" json:"retry"`
 	Rules           []BatchJobExpireFilter `yaml:"rules" json:"rules"`
@@ -340,8 +342,24 @@ func (r *BatchJobExpire) Expire(ctx context.Context, api ObjectLayer, vc *versio
 		PrefixEnabledFn:  vc.PrefixEnabled,
 		VersionSuspended: vc.Suspended(),
 	}
-	_, errs := api.DeleteObjects(ctx, r.Bucket, objsToDel, opts)
-	return errs
+
+	allErrs := make([]error, 0, len(objsToDel))
+
+	for {
+		count := len(objsToDel)
+		if count == 0 {
+			break
+		}
+		if count > maxDeleteList {
+			count = maxDeleteList
+		}
+		_, errs := api.DeleteObjects(ctx, r.Bucket, objsToDel[:count], opts)
+		allErrs = append(allErrs, errs...)
+		// Next batch of deletion
+		objsToDel = objsToDel[count:]
+	}
+
+	return allErrs
 }
 
 const (
@@ -432,7 +450,7 @@ 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 +482,13 @@ 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 +532,7 @@ func (r *BatchJobExpire) Start(ctx context.Context, api ObjectLayer, job BatchJo
 		JobType:   string(job.Type()),
 		StartTime: job.Started,
 	}
-	if err := ri.load(ctx, api, job); err != nil {
+	if err := ri.loadOrInit(ctx, api, job); err != nil {
 		return err
 	}
 
@@ -534,40 +552,54 @@ func (r *BatchJobExpire) Start(ctx context.Context, api ObjectLayer, job BatchJo
 		return err
 	}
 
-	ctx, cancel := context.WithCancel(ctx)
-	defer cancel()
+	ctx, cancelCause := context.WithCancelCause(ctx)
+	defer cancelCause(nil)
 
 	results := make(chan itemOrErr[ObjectInfo], workerSize)
-	if err := api.Walk(ctx, r.Bucket, r.Prefix, results, WalkOptions{
-		Marker:       lastObject,
-		LatestOnly:   false, // we need to visit all versions of the object to implement purge: retainVersions
-		VersionsSort: WalkVersionsSortDesc,
-	}); err != nil {
-		// Do not need to retry if we can't list objects on source.
-		return err
-	}
+	go func() {
+		for _, prefix := range r.Prefix.F() {
+			prefixResultCh := make(chan itemOrErr[ObjectInfo], workerSize)
+			err := api.Walk(ctx, r.Bucket, strings.TrimSpace(prefix), prefixResultCh, WalkOptions{
+				Marker:       lastObject,
+				LatestOnly:   false, // we need to visit all versions of the object to implement purge: retainVersions
+				VersionsSort: WalkVersionsSortDesc,
+			})
+			if err != nil {
+				cancelCause(err)
+				xioutil.SafeClose(results)
+				return
+			}
+			for result := range prefixResultCh {
+				results <- result
+			}
+		}
+		xioutil.SafeClose(results)
+	}()
 
 	// Goroutine to periodically save batch-expire job's in-memory state
 	saverQuitCh := make(chan struct{})
 	go func() {
 		saveTicker := time.NewTicker(10 * time.Second)
 		defer saveTicker.Stop()
-		for {
+		quit := false
+		after := time.Minute
+		for !quit {
 			select {
 			case <-saveTicker.C:
-				// persist in-memory state to disk after every 10secs.
-				batchLogIf(ctx, ri.updateAfter(ctx, api, 10*time.Second, job))
-
 			case <-ctx.Done():
-				// persist in-memory state immediately before exiting due to context cancellation.
-				batchLogIf(ctx, ri.updateAfter(ctx, api, 0, job))
-				return
-
+				quit = true
 			case <-saverQuitCh:
-				// persist in-memory state immediately to disk.
-				batchLogIf(ctx, ri.updateAfter(ctx, api, 0, job))
-				return
+				quit = true
 			}
+
+			if quit {
+				// save immediately if we are quitting
+				after = 0
+			}
+
+			ctx, cancel := context.WithTimeout(GlobalContext, 30*time.Second) // independent context
+			batchLogIf(ctx, ri.updateAfter(ctx, api, after, job))
+			cancel()
 		}
 	}()
 
@@ -584,7 +616,7 @@ func (r *BatchJobExpire) Start(ctx context.Context, api ObjectLayer, job BatchJo
 		versionsCount int
 		toDel         []expireObjInfo
 	)
-	failed := true
+	failed := false
 	for result := range results {
 		if result.Err != nil {
 			failed = true
@@ -653,6 +685,10 @@ func (r *BatchJobExpire) Start(ctx context.Context, api ObjectLayer, job BatchJo
 			ObjectInfo: result.Item,
 		})
 	}
+	if context.Cause(ctx) != nil {
+		xioutil.SafeClose(expireCh)
+		return context.Cause(ctx)
+	}
 	// Send any remaining objects downstream
 	if len(toDel) > 0 {
 		select {

cmd/batch-expire_gen.go

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

cmd/batch-expire_test.go

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

cmd/batch-handlers.go

@@ -28,6 +28,7 @@ import (
 	"math/rand"
 	"net/http"
 	"net/url"
+	"path/filepath"
 	"runtime"
 	"strconv"
 	"strings"
@@ -57,6 +58,11 @@ import (
 
 var globalBatchConfig batch.Config
 
+const (
+	// Keep the completed/failed job stats 3 days before removing it
+	oldJobsExpiration = 3 * 24 * time.Hour
+)
+
 // BatchJobRequest this is an internal data structure not for external consumption.
 type BatchJobRequest struct {
 	ID        string               `yaml:"-" json:"name"`
@@ -262,7 +268,7 @@ func (r *BatchJobReplicateV1) StartFromSource(ctx context.Context, api ObjectLay
 		JobType:   string(job.Type()),
 		StartTime: job.Started,
 	}
-	if err := ri.load(ctx, api, job); err != nil {
+	if err := ri.loadOrInit(ctx, api, job); err != nil {
 		return err
 	}
 	if ri.Complete {
@@ -281,12 +287,12 @@ func (r *BatchJobReplicateV1) StartFromSource(ctx context.Context, api ObjectLay
 	isStorageClassOnly := len(r.Flags.Filter.Metadata) == 1 && strings.EqualFold(r.Flags.Filter.Metadata[0].Key, xhttp.AmzStorageClass)
 
 	skip := func(oi ObjectInfo) (ok bool) {
-		if r.Flags.Filter.OlderThan > 0 && time.Since(oi.ModTime) < r.Flags.Filter.OlderThan {
+		if r.Flags.Filter.OlderThan > 0 && time.Since(oi.ModTime) < r.Flags.Filter.OlderThan.D() {
 			// skip all objects that are newer than specified older duration
 			return true
 		}
 
-		if r.Flags.Filter.NewerThan > 0 && time.Since(oi.ModTime) >= r.Flags.Filter.NewerThan {
+		if r.Flags.Filter.NewerThan > 0 && time.Since(oi.ModTime) >= r.Flags.Filter.NewerThan.D() {
 			// skip all objects that are older than specified newer duration
 			return true
 		}
@@ -379,12 +385,23 @@ func (r *BatchJobReplicateV1) StartFromSource(ctx context.Context, api ObjectLay
 		s3Type := r.Target.Type == BatchJobReplicateResourceS3 || r.Source.Type == BatchJobReplicateResourceS3
 		minioSrc := r.Source.Type == BatchJobReplicateResourceMinIO
 		ctx, cancel := context.WithCancel(ctx)
-		objInfoCh := c.ListObjects(ctx, r.Source.Bucket, miniogo.ListObjectsOptions{
-			Prefix:       r.Source.Prefix,
-			WithVersions: minioSrc,
-			Recursive:    true,
-			WithMetadata: true,
-		})
+
+		objInfoCh := make(chan miniogo.ObjectInfo, 1)
+		go func() {
+			for _, prefix := range r.Source.Prefix.F() {
+				prefixObjInfoCh := c.ListObjects(ctx, r.Source.Bucket, miniogo.ListObjectsOptions{
+					Prefix:       strings.TrimSpace(prefix),
+					WithVersions: minioSrc,
+					Recursive:    true,
+					WithMetadata: true,
+				})
+				for obj := range prefixObjInfoCh {
+					objInfoCh <- obj
+				}
+			}
+			xioutil.SafeClose(objInfoCh)
+		}()
+
 		prevObj := ""
 		skipReplicate := false
 
@@ -534,7 +551,7 @@ func toObjectInfo(bucket, object string, objInfo miniogo.ObjectInfo) ObjectInfo
 	return oi
 }
 
-func (r BatchJobReplicateV1) writeAsArchive(ctx context.Context, objAPI ObjectLayer, remoteClnt *minio.Client, entries []ObjectInfo) error {
+func (r BatchJobReplicateV1) writeAsArchive(ctx context.Context, objAPI ObjectLayer, remoteClnt *minio.Client, entries []ObjectInfo, prefix string) error {
 	input := make(chan minio.SnowballObject, 1)
 	opts := minio.SnowballOptions{
 		Opts:     minio.PutObjectOptions{},
@@ -556,6 +573,10 @@ func (r BatchJobReplicateV1) writeAsArchive(ctx context.Context, objAPI ObjectLa
 				continue
 			}
 
+			if prefix != "" {
+				entry.Name = pathJoin(prefix, entry.Name)
+			}
+
 			snowballObj := minio.SnowballObject{
 				// Create path to store objects within the bucket.
 				Key:       entry.Name,
@@ -719,63 +740,85 @@ const (
 
 	batchReplJobAPIVersion        = "v1"
 	batchReplJobDefaultRetries    = 3
-	batchReplJobDefaultRetryDelay = 250 * time.Millisecond
+	batchReplJobDefaultRetryDelay = time.Second
 )
 
-func getJobReportPath(job BatchJobRequest) string {
-	var fileName string
-	switch {
-	case job.Replicate != nil:
-		fileName = batchReplName
-	case job.KeyRotate != nil:
-		fileName = batchKeyRotationName
-	case job.Expire != nil:
-		fileName = batchExpireName
-	}
-	return pathJoin(batchJobReportsPrefix, job.ID, fileName)
-}
-
 func getJobPath(job BatchJobRequest) string {
 	return pathJoin(batchJobPrefix, job.ID)
 }
 
+func (ri *batchJobInfo) getJobReportPath() (string, error) {
+	var fileName string
+	switch madmin.BatchJobType(ri.JobType) {
+	case madmin.BatchJobReplicate:
+		fileName = batchReplName
+	case madmin.BatchJobKeyRotate:
+		fileName = batchKeyRotationName
+	case madmin.BatchJobExpire:
+		fileName = batchExpireName
+	default:
+		return "", fmt.Errorf("unknown job type: %v", ri.JobType)
+	}
+	return pathJoin(batchJobReportsPrefix, ri.JobID, fileName), nil
+}
+
+func (ri *batchJobInfo) loadOrInit(ctx context.Context, api ObjectLayer, job BatchJobRequest) error {
+	err := ri.load(ctx, api, job)
+	if errors.Is(err, errNoSuchJob) {
+		switch {
+		case job.Replicate != nil:
+			ri.Version = batchReplVersionV1
+			ri.RetryAttempts = batchReplJobDefaultRetries
+			if job.Replicate.Flags.Retry.Attempts > 0 {
+				ri.RetryAttempts = job.Replicate.Flags.Retry.Attempts
+			}
+		case job.KeyRotate != nil:
+			ri.Version = batchKeyRotateVersionV1
+			ri.RetryAttempts = batchKeyRotateJobDefaultRetries
+			if job.KeyRotate.Flags.Retry.Attempts > 0 {
+				ri.RetryAttempts = job.KeyRotate.Flags.Retry.Attempts
+			}
+		case job.Expire != nil:
+			ri.Version = batchExpireVersionV1
+			ri.RetryAttempts = batchExpireJobDefaultRetries
+			if job.Expire.Retry.Attempts > 0 {
+				ri.RetryAttempts = job.Expire.Retry.Attempts
+			}
+		}
+		return nil
+	}
+	return err
+}
+
 func (ri *batchJobInfo) load(ctx context.Context, api ObjectLayer, job BatchJobRequest) error {
+	path, err := job.getJobReportPath()
+	if err != nil {
+		batchLogIf(ctx, err)
+		return err
+	}
+	return ri.loadByPath(ctx, api, path)
+}
+
+func (ri *batchJobInfo) loadByPath(ctx context.Context, api ObjectLayer, path string) error {
 	var format, version uint16
-	switch {
-	case job.Replicate != nil:
+	switch filepath.Base(path) {
+	case batchReplName:
 		version = batchReplVersionV1
 		format = batchReplFormat
-	case job.KeyRotate != nil:
+	case batchKeyRotationName:
 		version = batchKeyRotateVersionV1
 		format = batchKeyRotationFormat
-	case job.Expire != nil:
+	case batchExpireName:
 		version = batchExpireVersionV1
 		format = batchExpireFormat
 	default:
 		return errors.New("no supported batch job request specified")
 	}
-	data, err := readConfig(ctx, api, getJobReportPath(job))
+
+	data, err := readConfig(ctx, api, path)
 	if err != nil {
 		if errors.Is(err, errConfigNotFound) || isErrObjectNotFound(err) {
-			ri.Version = int(version)
-			switch {
-			case job.Replicate != nil:
-				ri.RetryAttempts = batchReplJobDefaultRetries
-				if job.Replicate.Flags.Retry.Attempts > 0 {
-					ri.RetryAttempts = job.Replicate.Flags.Retry.Attempts
-				}
-			case job.KeyRotate != nil:
-				ri.RetryAttempts = batchKeyRotateJobDefaultRetries
-				if job.KeyRotate.Flags.Retry.Attempts > 0 {
-					ri.RetryAttempts = job.KeyRotate.Flags.Retry.Attempts
-				}
-			case job.Expire != nil:
-				ri.RetryAttempts = batchExpireJobDefaultRetries
-				if job.Expire.Retry.Attempts > 0 {
-					ri.RetryAttempts = job.Expire.Retry.Attempts
-				}
-			}
-			return nil
+			return errNoSuchJob
 		}
 		return err
 	}
@@ -919,7 +962,12 @@ func (ri *batchJobInfo) updateAfter(ctx context.Context, api ObjectLayer, durati
 		if err != nil {
 			return err
 		}
-		return saveConfig(ctx, api, getJobReportPath(job), buf)
+		path, err := ri.getJobReportPath()
+		if err != nil {
+			batchLogIf(ctx, err)
+			return err
+		}
+		return saveConfig(ctx, api, path, buf)
 	}
 	ri.mu.Unlock()
 	return nil
@@ -944,8 +992,10 @@ func (ri *batchJobInfo) trackCurrentBucketObject(bucket string, info ObjectInfo,
 	ri.mu.Lock()
 	defer ri.mu.Unlock()
 
-	ri.Bucket = bucket
-	ri.Object = info.Name
+	if success {
+		ri.Bucket = bucket
+		ri.Object = info.Name
+	}
 	ri.countItem(info.Size, info.DeleteMarker, success, attempt)
 }
 
@@ -971,7 +1021,7 @@ func (r *BatchJobReplicateV1) Start(ctx context.Context, api ObjectLayer, job Ba
 		JobType:   string(job.Type()),
 		StartTime: job.Started,
 	}
-	if err := ri.load(ctx, api, job); err != nil {
+	if err := ri.loadOrInit(ctx, api, job); err != nil {
 		return err
 	}
 	if ri.Complete {
@@ -981,18 +1031,18 @@ func (r *BatchJobReplicateV1) Start(ctx context.Context, api ObjectLayer, job Ba
 	lastObject := ri.Object
 
 	delay := job.Replicate.Flags.Retry.Delay
-	if delay == 0 {
+	if delay < time.Second {
 		delay = batchReplJobDefaultRetryDelay
 	}
 	rnd := rand.New(rand.NewSource(time.Now().UnixNano()))
 
 	selectObj := func(info FileInfo) (ok bool) {
-		if r.Flags.Filter.OlderThan > 0 && time.Since(info.ModTime) < r.Flags.Filter.OlderThan {
+		if r.Flags.Filter.OlderThan > 0 && time.Since(info.ModTime) < r.Flags.Filter.OlderThan.D() {
 			// skip all objects that are newer than specified older duration
 			return false
 		}
 
-		if r.Flags.Filter.NewerThan > 0 && time.Since(info.ModTime) >= r.Flags.Filter.NewerThan {
+		if r.Flags.Filter.NewerThan > 0 && time.Since(info.ModTime) >= r.Flags.Filter.NewerThan.D() {
 			// skip all objects that are older than specified newer duration
 			return false
 		}
@@ -1071,99 +1121,106 @@ func (r *BatchJobReplicateV1) Start(ctx context.Context, api ObjectLayer, job Ba
 
 	c.SetAppInfo("minio-"+batchJobPrefix, r.APIVersion+" "+job.ID)
 
-	var (
-		walkCh = make(chan itemOrErr[ObjectInfo], 100)
-		slowCh = make(chan itemOrErr[ObjectInfo], 100)
-	)
-
-	if !*r.Source.Snowball.Disable && r.Source.Type.isMinio() && r.Target.Type.isMinio() {
-		go func() {
-			// Snowball currently needs the high level minio-go Client, not the Core one
-			cl, err := miniogo.New(u.Host, &miniogo.Options{
-				Creds:        credentials.NewStaticV4(cred.AccessKey, cred.SecretKey, cred.SessionToken),
-				Secure:       u.Scheme == "https",
-				Transport:    getRemoteInstanceTransport(),
-				BucketLookup: lookupStyle(r.Target.Path),
-			})
-			if err != nil {
-				batchLogIf(ctx, err)
-				return
-			}
-
-			// Already validated before arriving here
-			smallerThan, _ := humanize.ParseBytes(*r.Source.Snowball.SmallerThan)
-
-			batch := make([]ObjectInfo, 0, *r.Source.Snowball.Batch)
-			writeFn := func(batch []ObjectInfo) {
-				if len(batch) > 0 {
-					if err := r.writeAsArchive(ctx, api, cl, batch); err != nil {
-						batchLogIf(ctx, err)
-						for _, b := range batch {
-							slowCh <- itemOrErr[ObjectInfo]{Item: b}
-						}
-					} else {
-						ri.trackCurrentBucketBatch(r.Source.Bucket, batch)
-						globalBatchJobsMetrics.save(job.ID, ri)
-						// persist in-memory state to disk after every 10secs.
-						batchLogIf(ctx, ri.updateAfter(ctx, api, 10*time.Second, job))
-					}
-				}
-			}
-			for obj := range walkCh {
-				if obj.Item.DeleteMarker || !obj.Item.VersionPurgeStatus.Empty() || obj.Item.Size >= int64(smallerThan) {
-					slowCh <- obj
-					continue
-				}
-
-				batch = append(batch, obj.Item)
-
-				if len(batch) < *r.Source.Snowball.Batch {
-					continue
-				}
-				writeFn(batch)
-				batch = batch[:0]
-			}
-			writeFn(batch)
-			xioutil.SafeClose(slowCh)
-		}()
-	} else {
-		slowCh = walkCh
-	}
-
-	workerSize, err := strconv.Atoi(env.Get("_MINIO_BATCH_REPLICATION_WORKERS", strconv.Itoa(runtime.GOMAXPROCS(0)/2)))
-	if err != nil {
-		return err
-	}
-
-	wk, err := workers.New(workerSize)
-	if err != nil {
-		// invalid worker size.
-		return err
-	}
-
-	walkQuorum := env.Get("_MINIO_BATCH_REPLICATION_WALK_QUORUM", "strict")
-	if walkQuorum == "" {
-		walkQuorum = "strict"
-	}
-
 	retryAttempts := ri.RetryAttempts
 	retry := false
 	for attempts := 1; attempts <= retryAttempts; attempts++ {
 		attempts := attempts
+		var (
+			walkCh = make(chan itemOrErr[ObjectInfo], 100)
+			slowCh = make(chan itemOrErr[ObjectInfo], 100)
+		)
 
-		ctx, cancel := context.WithCancel(ctx)
+		if r.Source.Snowball.Disable != nil && !*r.Source.Snowball.Disable && r.Source.Type.isMinio() && r.Target.Type.isMinio() {
+			go func() {
+				// Snowball currently needs the high level minio-go Client, not the Core one
+				cl, err := miniogo.New(u.Host, &miniogo.Options{
+					Creds:        credentials.NewStaticV4(cred.AccessKey, cred.SecretKey, cred.SessionToken),
+					Secure:       u.Scheme == "https",
+					Transport:    getRemoteInstanceTransport(),
+					BucketLookup: lookupStyle(r.Target.Path),
+				})
+				if err != nil {
+					batchLogOnceIf(ctx, err, job.ID+"miniogo.New")
+					return
+				}
+
+				// Already validated before arriving here
+				smallerThan, _ := humanize.ParseBytes(*r.Source.Snowball.SmallerThan)
+
+				batch := make([]ObjectInfo, 0, *r.Source.Snowball.Batch)
+				writeFn := func(batch []ObjectInfo) {
+					if len(batch) > 0 {
+						if err := r.writeAsArchive(ctx, api, cl, batch, r.Target.Prefix); err != nil {
+							batchLogOnceIf(ctx, err, job.ID+"writeAsArchive")
+							for _, b := range batch {
+								slowCh <- itemOrErr[ObjectInfo]{Item: b}
+							}
+						} else {
+							ri.trackCurrentBucketBatch(r.Source.Bucket, batch)
+							globalBatchJobsMetrics.save(job.ID, ri)
+							// persist in-memory state to disk after every 10secs.
+							batchLogOnceIf(ctx, ri.updateAfter(ctx, api, 10*time.Second, job), job.ID+"updateAfter")
+						}
+					}
+				}
+				for obj := range walkCh {
+					if obj.Item.DeleteMarker || !obj.Item.VersionPurgeStatus.Empty() || obj.Item.Size >= int64(smallerThan) {
+						slowCh <- obj
+						continue
+					}
+
+					batch = append(batch, obj.Item)
+
+					if len(batch) < *r.Source.Snowball.Batch {
+						continue
+					}
+					writeFn(batch)
+					batch = batch[:0]
+				}
+				writeFn(batch)
+				xioutil.SafeClose(slowCh)
+			}()
+		} else {
+			slowCh = walkCh
+		}
+
+		workerSize, err := strconv.Atoi(env.Get("_MINIO_BATCH_REPLICATION_WORKERS", strconv.Itoa(runtime.GOMAXPROCS(0)/2)))
+		if err != nil {
+			return err
+		}
+
+		wk, err := workers.New(workerSize)
+		if err != nil {
+			// invalid worker size.
+			return err
+		}
+
+		walkQuorum := env.Get("_MINIO_BATCH_REPLICATION_WALK_QUORUM", "strict")
+		if walkQuorum == "" {
+			walkQuorum = "strict"
+		}
+		ctx, cancelCause := context.WithCancelCause(ctx)
 		// one of source/target is s3, skip delete marker and all versions under the same object name.
 		s3Type := r.Target.Type == BatchJobReplicateResourceS3 || r.Source.Type == BatchJobReplicateResourceS3
 
-		if err := api.Walk(ctx, r.Source.Bucket, r.Source.Prefix, walkCh, WalkOptions{
-			Marker:   lastObject,
-			Filter:   selectObj,
-			AskDisks: walkQuorum,
-		}); err != nil {
-			cancel()
-			// Do not need to retry if we can't list objects on source.
-			return err
-		}
+		go func() {
+			for _, prefix := range r.Source.Prefix.F() {
+				prefixWalkCh := make(chan itemOrErr[ObjectInfo], 100)
+				if err := api.Walk(ctx, r.Source.Bucket, strings.TrimSpace(prefix), prefixWalkCh, WalkOptions{
+					Marker:   lastObject,
+					Filter:   selectObj,
+					AskDisks: walkQuorum,
+				}); err != nil {
+					cancelCause(err)
+					xioutil.SafeClose(walkCh)
+					return
+				}
+				for obj := range prefixWalkCh {
+					walkCh <- obj
+				}
+			}
+			xioutil.SafeClose(walkCh)
+		}()
 
 		prevObj := ""
 
@@ -1171,7 +1228,7 @@ func (r *BatchJobReplicateV1) Start(ctx context.Context, api ObjectLayer, job Ba
 		for res := range slowCh {
 			if res.Err != nil {
 				ri.Failed = true
-				batchLogIf(ctx, res.Err)
+				batchLogOnceIf(ctx, res.Err, job.ID+"res.Err")
 				continue
 			}
 			result := res.Item
@@ -1198,7 +1255,7 @@ func (r *BatchJobReplicateV1) Start(ctx context.Context, api ObjectLayer, job Ba
 						return
 					}
 					stopFn(result, err)
-					batchLogIf(ctx, err)
+					batchLogOnceIf(ctx, err, job.ID+"ReplicateToTarget")
 					success = false
 				} else {
 					stopFn(result, nil)
@@ -1206,7 +1263,7 @@ func (r *BatchJobReplicateV1) Start(ctx context.Context, api ObjectLayer, job Ba
 				ri.trackCurrentBucketObject(r.Source.Bucket, result, success, attempts)
 				globalBatchJobsMetrics.save(job.ID, ri)
 				// persist in-memory state to disk after every 10secs.
-				batchLogIf(ctx, ri.updateAfter(ctx, api, 10*time.Second, job))
+				batchLogOnceIf(ctx, ri.updateAfter(ctx, api, 10*time.Second, job), job.ID+"updateAfter2")
 
 				if wait := globalBatchConfig.ReplicationWait(); wait > 0 {
 					time.Sleep(wait)
@@ -1214,20 +1271,23 @@ func (r *BatchJobReplicateV1) Start(ctx context.Context, api ObjectLayer, job Ba
 			}()
 		}
 		wk.Wait()
-
+		// Do not need to retry if we can't list objects on source.
+		if context.Cause(ctx) != nil {
+			return context.Cause(ctx)
+		}
 		ri.RetryAttempts = attempts
 		ri.Complete = ri.ObjectsFailed == 0
 		ri.Failed = ri.ObjectsFailed > 0
 
 		globalBatchJobsMetrics.save(job.ID, ri)
 		// persist in-memory state to disk.
-		batchLogIf(ctx, ri.updateAfter(ctx, api, 0, job))
+		batchLogOnceIf(ctx, ri.updateAfter(ctx, api, 0, job), job.ID+"updateAfter3")
 
 		if err := r.Notify(ctx, ri); err != nil {
-			batchLogIf(ctx, fmt.Errorf("unable to notify %v", err))
+			batchLogOnceIf(ctx, fmt.Errorf("unable to notify %v", err), job.ID+"notify")
 		}
 
-		cancel()
+		cancelCause(nil)
 		if ri.Failed {
 			ri.ObjectsFailed = 0
 			ri.Bucket = ""
@@ -1436,10 +1496,24 @@ func (j BatchJobRequest) Validate(ctx context.Context, o ObjectLayer) error {
 }
 
 func (j BatchJobRequest) delete(ctx context.Context, api ObjectLayer) {
-	deleteConfig(ctx, api, getJobReportPath(j))
 	deleteConfig(ctx, api, getJobPath(j))
 }
 
+func (j BatchJobRequest) getJobReportPath() (string, error) {
+	var fileName string
+	switch {
+	case j.Replicate != nil:
+		fileName = batchReplName
+	case j.KeyRotate != nil:
+		fileName = batchKeyRotationName
+	case j.Expire != nil:
+		fileName = batchExpireName
+	default:
+		return "", errors.New("unknown job type")
+	}
+	return pathJoin(batchJobReportsPrefix, j.ID, fileName), nil
+}
+
 func (j *BatchJobRequest) save(ctx context.Context, api ObjectLayer) error {
 	if j.Replicate == nil && j.KeyRotate == nil && j.Expire == nil {
 		return errInvalidArgument
@@ -1476,7 +1550,7 @@ 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 +1594,9 @@ func (a adminAPIHandlers) ListBatchJobs(w http.ResponseWriter, r *http.Request)
 			writeErrorResponseJSON(ctx, w, toAPIError(ctx, result.Err), r.URL)
 			return
 		}
+		if strings.HasPrefix(result.Item.Name, batchJobReportsPrefix+slashSeparator) {
+			continue
+		}
 		req := &BatchJobRequest{}
 		if err := req.load(ctx, objectAPI, result.Item.Name); err != nil {
 			if !errors.Is(err, errNoSuchJob) {
@@ -1542,6 +1619,55 @@ func (a adminAPIHandlers) ListBatchJobs(w http.ResponseWriter, r *http.Request)
 	batchLogIf(ctx, json.NewEncoder(w).Encode(&listResult))
 }
 
+// BatchJobStatus - returns the status of a batch job saved in the disk
+func (a adminAPIHandlers) BatchJobStatus(w http.ResponseWriter, r *http.Request) {
+	ctx := r.Context()
+
+	objectAPI, _ := validateAdminReq(ctx, w, r, policy.ListBatchJobsAction)
+	if objectAPI == nil {
+		return
+	}
+
+	jobID := r.Form.Get("jobId")
+	if jobID == "" {
+		writeErrorResponseJSON(ctx, w, toAPIError(ctx, errInvalidArgument), r.URL)
+		return
+	}
+
+	req := BatchJobRequest{ID: jobID}
+	if i := strings.Index(jobID, "-"); i > 0 {
+		switch madmin.BatchJobType(jobID[:i]) {
+		case madmin.BatchJobReplicate:
+			req.Replicate = &BatchJobReplicateV1{}
+		case madmin.BatchJobKeyRotate:
+			req.KeyRotate = &BatchJobKeyRotateV1{}
+		case madmin.BatchJobExpire:
+			req.Expire = &BatchJobExpire{}
+		default:
+			writeErrorResponseJSON(ctx, w, toAPIError(ctx, errors.New("job ID format unrecognized")), r.URL)
+			return
+		}
+	}
+
+	ri := &batchJobInfo{}
+	if err := ri.load(ctx, objectAPI, req); err != nil {
+		if !errors.Is(err, errNoSuchJob) {
+			batchLogIf(ctx, err)
+		}
+		writeErrorResponseJSON(ctx, w, toAPIError(ctx, err), r.URL)
+		return
+	}
+
+	buf, err := json.Marshal(madmin.BatchJobStatus{LastMetric: ri.metric()})
+	if err != nil {
+		batchLogIf(ctx, err)
+		writeErrorResponseJSON(ctx, w, toAPIError(ctx, err), r.URL)
+		return
+	}
+
+	w.Write(buf)
+}
+
 var errNoSuchJob = errors.New("no such job")
 
 // DescribeBatchJob returns the currently active batch job definition
@@ -1579,7 +1705,7 @@ func (a adminAPIHandlers) DescribeBatchJob(w http.ResponseWriter, r *http.Reques
 	w.Write(buf)
 }
 
-// StarBatchJob queue a new job for execution
+// StartBatchJob queue a new job for execution
 func (a adminAPIHandlers) StartBatchJob(w http.ResponseWriter, r *http.Request) {
 	ctx := r.Context()
 
@@ -1633,7 +1759,7 @@ func (a adminAPIHandlers) StartBatchJob(w http.ResponseWriter, r *http.Request)
 		return
 	}
 
-	job.ID = fmt.Sprintf("%s:%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 +1847,60 @@ func newBatchJobPool(ctx context.Context, o ObjectLayer, workers int) *BatchJobP
 		jobCancelers: make(map[string]context.CancelFunc),
 	}
 	jpool.ResizeWorkers(workers)
-	jpool.resume()
+
+	randomWait := func() time.Duration {
+		// randomWait depends on the number of nodes to avoid triggering resume and cleanups at the same time.
+		return time.Duration(rand.Float64() * float64(time.Duration(globalEndpoints.NEndpoints())*time.Hour))
+	}
+
+	go func() {
+		jpool.resume(randomWait)
+		jpool.cleanupReports(randomWait)
+	}()
+
 	return jpool
 }
 
-func (j *BatchJobPool) resume() {
+func (j *BatchJobPool) cleanupReports(randomWait func() time.Duration) {
+	t := time.NewTimer(randomWait())
+	defer t.Stop()
+
+	for {
+		select {
+		case <-GlobalContext.Done():
+			return
+		case <-t.C:
+			results := make(chan itemOrErr[ObjectInfo], 100)
+			ctx, cancel := context.WithCancel(j.ctx)
+			defer cancel()
+			if err := j.objLayer.Walk(ctx, minioMetaBucket, batchJobReportsPrefix, results, WalkOptions{}); err != nil {
+				batchLogIf(j.ctx, err)
+				t.Reset(randomWait())
+				continue
+			}
+			for result := range results {
+				if result.Err != nil {
+					batchLogIf(j.ctx, result.Err)
+					continue
+				}
+				ri := &batchJobInfo{}
+				if err := ri.loadByPath(ctx, j.objLayer, result.Item.Name); err != nil {
+					batchLogIf(ctx, err)
+					continue
+				}
+				if (ri.Complete || ri.Failed) && time.Since(ri.LastUpdate) > oldJobsExpiration {
+					deleteConfig(ctx, j.objLayer, result.Item.Name)
+				}
+			}
+
+			t.Reset(randomWait())
+		}
+	}
+}
+
+func (j *BatchJobPool) resume(randomWait func() time.Duration) {
+	time.Sleep(randomWait())
+
 	results := make(chan itemOrErr[ObjectInfo], 100)
 	ctx, cancel := context.WithCancel(j.ctx)
 	defer cancel()
@@ -1738,6 +1913,9 @@ func (j *BatchJobPool) resume() {
 			batchLogIf(j.ctx, result.Err)
 			continue
 		}
+		if strings.HasPrefix(result.Item.Name, batchJobReportsPrefix+slashSeparator) {
+			continue
+		}
 		// ignore batch-replicate.bin and batch-rotate.bin entries
 		if strings.HasSuffix(result.Item.Name, slashSeparator) {
 			continue
@@ -1988,7 +2166,7 @@ func (m *batchJobMetrics) purgeJobMetrics() {
 			var toDeleteJobMetrics []string
 			m.RLock()
 			for id, metrics := range m.metrics {
-				if time.Since(metrics.LastUpdate) > 24*time.Hour && (metrics.Complete || metrics.Failed) {
+				if time.Since(metrics.LastUpdate) > oldJobsExpiration && (metrics.Complete || metrics.Failed) {
 					toDeleteJobMetrics = append(toDeleteJobMetrics, id)
 				}
 			}
@@ -2077,3 +2255,30 @@ func lookupStyle(s string) miniogo.BucketLookupType {
 	}
 	return lookup
 }
+
+// BatchJobPrefix - to support prefix field yaml unmarshalling with string or slice of strings
+type BatchJobPrefix []string
+
+var _ yaml.Unmarshaler = &BatchJobPrefix{}
+
+// UnmarshalYAML - to support prefix field yaml unmarshalling with string or slice of strings
+func (b *BatchJobPrefix) UnmarshalYAML(value *yaml.Node) error {
+	// try slice first
+	tmpSlice := []string{}
+	if err := value.Decode(&tmpSlice); err == nil {
+		*b = tmpSlice
+		return nil
+	}
+	// try string
+	tmpStr := ""
+	if err := value.Decode(&tmpStr); err == nil {
+		*b = []string{tmpStr}
+		return nil
+	}
+	return fmt.Errorf("unable to decode %s", value.Value)
+}
+
+// F - return prefix(es) as slice
+func (b *BatchJobPrefix) F() []string {
+	return *b
+}

cmd/batch-handlers_gen.go

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

cmd/batch-handlers_gen_test.go

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

cmd/batch-handlers_test.go

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

cmd/batch-replicate.go

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

cmd/batch-replicate_gen.go

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

cmd/batch-replicate_test.go

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

cmd/batch-rotate.go

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

cmd/bitrot-streaming.go

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

cmd/bootstrap-peer-server.go

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

cmd/bootstrap-peer-server_gen.go

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

cmd/bucket-handlers.go

@@ -51,7 +51,6 @@ import (
 	sse "github.com/minio/minio/internal/bucket/encryption"
 	objectlock "github.com/minio/minio/internal/bucket/object/lock"
 	"github.com/minio/minio/internal/bucket/replication"
-	"github.com/minio/minio/internal/config/cache"
 	"github.com/minio/minio/internal/config/dns"
 	"github.com/minio/minio/internal/crypto"
 	"github.com/minio/minio/internal/event"
@@ -92,7 +91,7 @@ const (
 // -- If IP of the entry doesn't match, this means entry is
 //
 //	for another instance. Log an error to console.
-func initFederatorBackend(buckets []BucketInfo, objLayer ObjectLayer) {
+func initFederatorBackend(buckets []string, objLayer ObjectLayer) {
 	if len(buckets) == 0 {
 		return
 	}
@@ -113,10 +112,10 @@ func initFederatorBackend(buckets []BucketInfo, objLayer ObjectLayer) {
 	domainMissing := err == dns.ErrDomainMissing
 	if dnsBuckets != nil {
 		for _, bucket := range buckets {
-			bucketsSet.Add(bucket.Name)
-			r, ok := dnsBuckets[bucket.Name]
+			bucketsSet.Add(bucket)
+			r, ok := dnsBuckets[bucket]
 			if !ok {
-				bucketsToBeUpdated.Add(bucket.Name)
+				bucketsToBeUpdated.Add(bucket)
 				continue
 			}
 			if !globalDomainIPs.Intersection(set.CreateStringSet(getHostsSlice(r)...)).IsEmpty() {
@@ -135,7 +134,7 @@ func initFederatorBackend(buckets []BucketInfo, objLayer ObjectLayer) {
 				// but if we do see a difference with local domain IPs with
 				// hostSlice from etcd then we should update with newer
 				// domainIPs, we proceed to do that here.
-				bucketsToBeUpdated.Add(bucket.Name)
+				bucketsToBeUpdated.Add(bucket)
 				continue
 			}
 
@@ -144,7 +143,7 @@ func initFederatorBackend(buckets []BucketInfo, objLayer ObjectLayer) {
 			// bucket names are globally unique in federation at a given
 			// path prefix, name collision is not allowed. We simply log
 			// an error and continue.
-			bucketsInConflict.Add(bucket.Name)
+			bucketsInConflict.Add(bucket)
 		}
 	}
 
@@ -1342,22 +1341,6 @@ func (api objectAPIHandlers) PostPolicyBucketHandler(w http.ResponseWriter, r *h
 					continue
 				}
 
-				asize, err := objInfo.GetActualSize()
-				if err != nil {
-					asize = objInfo.Size
-				}
-
-				globalCacheConfig.Set(&cache.ObjectInfo{
-					Key:          objInfo.Name,
-					Bucket:       objInfo.Bucket,
-					ETag:         getDecryptedETag(formValues, objInfo, false),
-					ModTime:      objInfo.ModTime,
-					Expires:      objInfo.Expires.UTC().Format(http.TimeFormat),
-					CacheControl: objInfo.CacheControl,
-					Metadata:     cleanReservedKeys(objInfo.UserDefined),
-					Size:         asize,
-				})
-
 				fanOutResp = append(fanOutResp, minio.PutObjectFanOutResponse{
 					Key:          objInfo.Name,
 					ETag:         getDecryptedETag(formValues, objInfo, false),
@@ -1420,7 +1403,7 @@ func (api objectAPIHandlers) PostPolicyBucketHandler(w http.ResponseWriter, r *h
 	if formValues.Get(postPolicyBucketTagging) != "" {
 		tags, err := tags.ParseObjectXML(strings.NewReader(formValues.Get(postPolicyBucketTagging)))
 		if err != nil {
-			writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL)
+			writeErrorResponse(ctx, w, errorCodes.ToAPIErr(ErrMalformedPOSTRequest), r.URL)
 			return
 		}
 		tagsStr := tags.String()
@@ -1452,22 +1435,6 @@ func (api objectAPIHandlers) PostPolicyBucketHandler(w http.ResponseWriter, r *h
 		w.Header().Set(xhttp.Location, obj)
 	}
 
-	asize, err := objInfo.GetActualSize()
-	if err != nil {
-		asize = objInfo.Size
-	}
-
-	defer globalCacheConfig.Set(&cache.ObjectInfo{
-		Key:          objInfo.Name,
-		Bucket:       objInfo.Bucket,
-		ETag:         etag,
-		ModTime:      objInfo.ModTime,
-		Expires:      objInfo.ExpiresStr(),
-		CacheControl: objInfo.CacheControl,
-		Metadata:     cleanReservedKeys(objInfo.UserDefined),
-		Size:         asize,
-	})
-
 	// Notify object created event.
 	defer sendEvent(eventArgs{
 		EventName:    event.ObjectCreatedPost,
@@ -1715,7 +1682,7 @@ func (api objectAPIHandlers) DeleteBucketHandler(w http.ResponseWriter, r *http.
 	}
 
 	globalNotificationSys.DeleteBucketMetadata(ctx, bucket)
-	globalReplicationPool.deleteResyncMetadata(ctx, bucket)
+	globalReplicationPool.Get().deleteResyncMetadata(ctx, bucket)
 
 	// Call site replication hook.
 	replLogIf(ctx, globalSiteReplicationSys.DeleteBucketHook(ctx, bucket, forceDelete))

cmd/bucket-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-audit.go

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

cmd/bucket-lifecycle-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

@@ -71,7 +71,8 @@ func NewLifecycleSys() *LifecycleSys {
 	return &LifecycleSys{}
 }
 
-func ilmTrace(startTime time.Time, duration time.Duration, oi ObjectInfo, event string) madmin.TraceInfo {
+func ilmTrace(startTime time.Time, duration time.Duration, oi ObjectInfo, event string, metadata map[string]string) madmin.TraceInfo {
+	sz, _ := oi.GetActualSize()
 	return madmin.TraceInfo{
 		TraceType: madmin.TraceILM,
 		Time:      startTime,
@@ -79,18 +80,19 @@ func ilmTrace(startTime time.Time, duration time.Duration, oi ObjectInfo, event
 		FuncName:  event,
 		Duration:  duration,
 		Path:      pathJoin(oi.Bucket, oi.Name),
+		Bytes:     sz,
 		Error:     "",
 		Message:   getSource(4),
-		Custom:    map[string]string{"version-id": oi.VersionID},
+		Custom:    metadata,
 	}
 }
 
-func (sys *LifecycleSys) trace(oi ObjectInfo) func(event string) {
+func (sys *LifecycleSys) trace(oi ObjectInfo) func(event string, metadata map[string]string) {
 	startTime := time.Now()
-	return func(event string) {
+	return func(event string, metadata map[string]string) {
 		duration := time.Since(startTime)
 		if globalTrace.NumSubscribers(madmin.TraceILM) > 0 {
-			globalTrace.Publish(ilmTrace(startTime, duration, oi, event))
+			globalTrace.Publish(ilmTrace(startTime, duration, oi, event, metadata))
 		}
 	}
 }
@@ -277,6 +279,10 @@ func (es *expiryState) getWorkerCh(h uint64) chan<- expiryOp {
 }
 
 func (es *expiryState) ResizeWorkers(n int) {
+	if n == 0 {
+		n = 100
+	}
+
 	// Lock to avoid multiple resizes to happen at the same time.
 	es.mu.Lock()
 	defer es.mu.Unlock()
@@ -538,6 +544,10 @@ func (t *transitionState) UpdateWorkers(n int) {
 }
 
 func (t *transitionState) updateWorkers(n int) {
+	if n == 0 {
+		n = 100
+	}
+
 	for t.numWorkers < n {
 		go t.worker(t.objAPI)
 		t.numWorkers++
@@ -573,6 +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)
 		}
 	}
@@ -693,6 +707,11 @@ type auditTierOp struct {
 	Error            string `json:"error,omitempty"`
 }
 
+func (op auditTierOp) String() string {
+	// flattening the auditTierOp{} for audit
+	return fmt.Sprintf("tier:%s,respNS:%d,tx:%d,err:%s", op.Tier, op.TimeToResponseNS, op.OutputBytes, op.Error)
+}
+
 func auditTierActions(ctx context.Context, tier string, bytes int64) func(err error) {
 	startTime := time.Now()
 	return func(err error) {
@@ -716,7 +735,7 @@ func auditTierActions(ctx context.Context, tier string, bytes int64) func(err er
 			globalTierMetrics.logFailure(tier)
 		}
 
-		logger.GetReqInfo(ctx).AppendTags("tierStats", op)
+		logger.GetReqInfo(ctx).AppendTags("tierStats", op.String())
 	}
 }
 
@@ -727,7 +746,7 @@ 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)
 	}
@@ -984,7 +1003,7 @@ func ongoingRestoreObj() restoreObjStatus {
 	}
 }
 
-// completeRestoreObj constructs restoreObjStatus for a completed restore-object with given expiry.
+// completedRestoreObj constructs restoreObjStatus for a completed restore-object with given expiry.
 func completedRestoreObj(expiry time.Time) restoreObjStatus {
 	return restoreObjStatus{
 		ongoing: false,

cmd/bucket-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

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

cmd/bucket-metadata.go

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

cmd/bucket-metadata_gen.go

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

cmd/bucket-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-quota.go

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

cmd/bucket-replication-handlers.go

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

cmd/bucket-replication-metrics.go

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

cmd/bucket-replication-stats.go

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

cmd/bucket-replication-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"
@@ -49,6 +50,7 @@ import (
 	xhttp "github.com/minio/minio/internal/http"
 	xioutil "github.com/minio/minio/internal/ioutil"
 	"github.com/minio/minio/internal/logger"
+	"github.com/minio/minio/internal/once"
 	"github.com/tinylib/msgp/msgp"
 	"github.com/zeebo/xxh3"
 	"golang.org/x/exp/maps"
@@ -74,8 +76,9 @@ 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 +92,10 @@ func getReplicationConfig(ctx context.Context, bucketName string) (rc *replicati
 
 // validateReplicationDestination returns error if replication destination bucket missing or not configured
 // It also returns true if replication destination is same as this server.
-func validateReplicationDestination(ctx context.Context, bucket string, rCfg *replication.Config, checkRemote bool) (bool, APIError) {
+func validateReplicationDestination(ctx context.Context, bucket string, rCfg *replication.Config, opts *validateReplicationDestinationOptions) (bool, APIError) {
+	if opts == nil {
+		opts = &validateReplicationDestinationOptions{}
+	}
 	var arns []string
 	if rCfg.RoleArn != "" {
 		arns = append(arns, rCfg.RoleArn)
@@ -111,7 +117,7 @@ func validateReplicationDestination(ctx context.Context, bucket string, rCfg *re
 		if clnt == nil {
 			return sameTarget, toAPIError(ctx, BucketRemoteTargetNotFound{Bucket: bucket})
 		}
-		if checkRemote { // validate remote bucket
+		if opts.CheckRemoteBucket { // validate remote bucket
 			found, err := clnt.BucketExists(ctx, arn.Bucket)
 			if err != nil {
 				return sameTarget, errorCodes.ToAPIErrWithErr(ErrRemoteDestinationNotFoundError, err)
@@ -131,23 +137,29 @@ func validateReplicationDestination(ctx context.Context, bucket string, rCfg *re
 				}
 			}
 		}
+		// if checked bucket, then check the ready is unnecessary
+		if !opts.CheckRemoteBucket && opts.CheckReady {
+			endpoint := clnt.EndpointURL().String()
+			if errInt, ok := opts.checkReadyErr.Load(endpoint); !ok {
+				err = checkRemoteEndpoint(ctx, clnt.EndpointURL())
+				opts.checkReadyErr.Store(endpoint, err)
+			} else {
+				if errInt == nil {
+					err = nil
+				} else {
+					err = errInt.(error)
+				}
+			}
+			switch err.(type) {
+			case BucketRemoteIdenticalToSource:
+				return true, errorCodes.ToAPIErrWithErr(ErrBucketRemoteIdenticalToSource, fmt.Errorf("remote target endpoint %s is self referential", clnt.EndpointURL().String()))
+			default:
+			}
+		}
 		// validate replication ARN against target endpoint
-		c := globalBucketTargetSys.GetRemoteTargetClient(bucket, arnStr)
-		if c != nil {
-			if err := checkRemoteEndpoint(ctx, c.EndpointURL()); err != nil {
-				switch err.(type) {
-				case BucketRemoteIdenticalToSource:
-					return true, errorCodes.ToAPIErrWithErr(ErrBucketRemoteIdenticalToSource, fmt.Errorf("remote target endpoint %s is self referential", c.EndpointURL().String()))
-				default:
-				}
-			}
-			if c.EndpointURL().String() == clnt.EndpointURL().String() {
-				selfTarget, _ := isLocalHost(clnt.EndpointURL().Hostname(), clnt.EndpointURL().Port(), globalMinioPort)
-				if !sameTarget {
-					sameTarget = selfTarget
-				}
-				continue
-			}
+		selfTarget, _ := isLocalHost(clnt.EndpointURL().Hostname(), clnt.EndpointURL().Port(), globalMinioPort)
+		if !sameTarget {
+			sameTarget = selfTarget
 		}
 	}
 
@@ -467,7 +479,7 @@ func replicateDelete(ctx context.Context, dobj DeletedObjectReplicationInfo, obj
 	lk := objectAPI.NewNSLock(bucket, "/[replicate]/"+dobj.ObjectName)
 	lkctx, err := lk.GetLock(ctx, globalOperationTimeout)
 	if err != nil {
-		globalReplicationPool.queueMRFSave(dobj.ToMRFEntry())
+		globalReplicationPool.Get().queueMRFSave(dobj.ToMRFEntry())
 		sendEvent(eventArgs{
 			BucketName: bucket,
 			Object: ObjectInfo{
@@ -537,7 +549,7 @@ func replicateDelete(ctx context.Context, dobj DeletedObjectReplicationInfo, obj
 	// to decrement pending count later.
 	for _, rinfo := range rinfos.Targets {
 		if rinfo.ReplicationStatus != rinfo.PrevReplicationStatus {
-			globalReplicationStats.Update(dobj.Bucket, rinfo, replicationStatus,
+			globalReplicationStats.Load().Update(dobj.Bucket, rinfo, replicationStatus,
 				prevStatus)
 		}
 	}
@@ -545,7 +557,7 @@ func replicateDelete(ctx context.Context, dobj DeletedObjectReplicationInfo, obj
 	eventName := event.ObjectReplicationComplete
 	if replicationStatus == replication.Failed {
 		eventName = event.ObjectReplicationFailed
-		globalReplicationPool.queueMRFSave(dobj.ToMRFEntry())
+		globalReplicationPool.Get().queueMRFSave(dobj.ToMRFEntry())
 	}
 	drs := getReplicationState(rinfos, dobj.ReplicationState, dobj.VersionID)
 	if replicationStatus != prevStatus {
@@ -763,11 +775,13 @@ 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 +796,21 @@ 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
 	}
@@ -1026,7 +1055,7 @@ func replicateObject(ctx context.Context, ri ReplicateObjectInfo, objectAPI Obje
 			UserAgent:  "Internal: [Replication]",
 			Host:       globalLocalNodeName,
 		})
-		globalReplicationPool.queueMRFSave(ri.ToMRFEntry())
+		globalReplicationPool.Get().queueMRFSave(ri.ToMRFEntry())
 		return
 	}
 	ctx = lkctx.Context()
@@ -1111,7 +1140,7 @@ func replicateObject(ctx context.Context, ri ReplicateObjectInfo, objectAPI Obje
 		for _, rinfo := range rinfos.Targets {
 			if rinfo.ReplicationStatus != rinfo.PrevReplicationStatus {
 				rinfo.OpType = opType // update optype to reflect correct operation.
-				globalReplicationStats.Update(bucket, rinfo, rinfo.ReplicationStatus, rinfo.PrevReplicationStatus)
+				globalReplicationStats.Load().Update(bucket, rinfo, rinfo.ReplicationStatus, rinfo.PrevReplicationStatus)
 			}
 		}
 	}
@@ -1131,7 +1160,7 @@ func replicateObject(ctx context.Context, ri ReplicateObjectInfo, objectAPI Obje
 		ri.EventType = ReplicateMRF
 		ri.ReplicationStatusInternal = rinfos.ReplicationStatusInternal()
 		ri.RetryCount++
-		globalReplicationPool.queueMRFSave(ri.ToMRFEntry())
+		globalReplicationPool.Get().queueMRFSave(ri.ToMRFEntry())
 	}
 }
 
@@ -1201,17 +1230,23 @@ 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 +1274,7 @@ 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 +1300,7 @@ func (ri ReplicateObjectInfo) replicateObject(ctx context.Context, objectAPI Obj
 		HeaderSize: headerSize,
 	}
 	newCtx := ctx
-	if globalBucketMonitor.IsThrottled(bucket, tgt.ARN) {
+	if globalBucketMonitor.IsThrottled(bucket, tgt.ARN) && objInfo.Size < minLargeObjSize {
 		var cancel context.CancelFunc
 		newCtx, cancel = context.WithTimeout(ctx, throttleDeadline)
 		defer cancel()
@@ -1377,7 +1412,8 @@ 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 +1473,13 @@ 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 +1507,7 @@ func (ri ReplicateObjectInfo) replicateAll(ctx context.Context, objectAPI Object
 			return
 		}
 	}
+applyAction:
 	rinfo.ReplicationStatus = replication.Completed
 	rinfo.Size = size
 	rinfo.ReplicationAction = rAction
@@ -1506,7 +1550,7 @@ 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 +1575,7 @@ 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 +1641,13 @@ func replicateObjectWithMultipart(ctx context.Context, c *minio.Core, bucket, ob
 	}()
 
 	var (
-		hr    *hash.Reader
-		pInfo minio.ObjectPart
+		hr     *hash.Reader
+		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)
@@ -1614,35 +1658,51 @@ 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) {
-			objectSize += partInfo.Size
-			pInfo, err = c.PutObjectPart(ctx, bucket, object, uploadID, partInfo.Number, hr, partInfo.Size, popts)
+		var size int64
+		if isSSEC {
+			size = partInfo.Size
 		} else {
-			objectSize += partInfo.ActualSize
-			pInfo, err = c.PutObjectPart(ctx, bucket, object, uploadID, partInfo.Number, hr, partInfo.ActualSize, popts)
+			size = partInfo.ActualSize
 		}
+		objectSize += size
+		pInfo, err := c.PutObjectPart(ctx, bucket, object, uploadID, partInfo.Number, hr, size, popts)
 		if err != nil {
 			return err
 		}
-		if !crypto.SSEC.IsEncrypted(objInfo.UserDefined) && pInfo.Size != partInfo.ActualSize {
-			return fmt.Errorf("Part size mismatch: got %d, want %d", pInfo.Size, partInfo.ActualSize)
+		if pInfo.Size != size {
+			return fmt.Errorf("ssec(%t): Part size mismatch: got %d, want %d", isSSEC, pInfo.Size, size)
 		}
 		uploadedParts = append(uploadedParts, minio.CompletePart{
-			PartNumber: pInfo.PartNumber,
-			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{
+		xhttp.MinIOReplicationActualObjectSize: 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,
@@ -1728,23 +1788,27 @@ const (
 )
 
 var (
-	globalReplicationPool  *ReplicationPool
-	globalReplicationStats *ReplicationStats
+	globalReplicationPool  = once.NewSingleton[ReplicationPool]()
+	globalReplicationStats atomic.Pointer[ReplicationStats]
 )
 
 // ReplicationPool describes replication pool
 type ReplicationPool struct {
 	// atomic ops:
 	activeWorkers    int32
+	activeLrgWorkers int32
 	activeMRFWorkers int32
 
-	objLayer   ObjectLayer
-	ctx        context.Context
-	priority   string
-	maxWorkers int
-	mu         sync.RWMutex
-	mrfMU      sync.Mutex
-	resyncer   *replicationResyncer
+	objLayer    ObjectLayer
+	ctx         context.Context
+	priority    string
+	maxWorkers  int
+	maxLWorkers int
+	stats       *ReplicationStats
+
+	mu       sync.RWMutex
+	mrfMU    sync.Mutex
+	resyncer *replicationResyncer
 
 	// workers:
 	workers    []chan ReplicationWorkerOperation
@@ -1787,7 +1851,7 @@ const (
 )
 
 // NewReplicationPool creates a pool of replication workers of specified size
-func NewReplicationPool(ctx context.Context, o ObjectLayer, opts replicationPoolOpts) *ReplicationPool {
+func NewReplicationPool(ctx context.Context, o ObjectLayer, opts replicationPoolOpts, stats *ReplicationStats) *ReplicationPool {
 	var workers, failedWorkers int
 	priority := "auto"
 	maxWorkers := WorkerMaxLimit
@@ -1815,9 +1879,13 @@ func NewReplicationPool(ctx context.Context, o ObjectLayer, opts replicationPool
 	if maxWorkers > 0 && failedWorkers > maxWorkers {
 		failedWorkers = maxWorkers
 	}
+	maxLWorkers := LargeWorkerCount
+	if opts.MaxLWorkers > 0 {
+		maxLWorkers = opts.MaxLWorkers
+	}
 	pool := &ReplicationPool{
 		workers:         make([]chan ReplicationWorkerOperation, 0, workers),
-		lrgworkers:      make([]chan ReplicationWorkerOperation, 0, LargeWorkerCount),
+		lrgworkers:      make([]chan ReplicationWorkerOperation, 0, maxLWorkers),
 		mrfReplicaCh:    make(chan ReplicationWorkerOperation, 100000),
 		mrfWorkerKillCh: make(chan struct{}, failedWorkers),
 		resyncer:        newresyncer(),
@@ -1825,11 +1893,13 @@ func NewReplicationPool(ctx context.Context, o ObjectLayer, opts replicationPool
 		mrfStopCh:       make(chan struct{}, 1),
 		ctx:             ctx,
 		objLayer:        o,
+		stats:           stats,
 		priority:        priority,
 		maxWorkers:      maxWorkers,
+		maxLWorkers:     maxLWorkers,
 	}
 
-	pool.AddLargeWorkers()
+	pool.ResizeLrgWorkers(maxLWorkers, 0)
 	pool.ResizeWorkers(workers, 0)
 	pool.ResizeFailedWorkers(failedWorkers)
 	go pool.resyncer.PersistToDisk(ctx, o)
@@ -1851,11 +1921,11 @@ func (p *ReplicationPool) AddMRFWorker() {
 			}
 			switch v := oi.(type) {
 			case ReplicateObjectInfo:
-				globalReplicationStats.incQ(v.Bucket, v.Size, v.DeleteMarker, v.OpType)
+				p.stats.incQ(v.Bucket, v.Size, v.DeleteMarker, v.OpType)
 				atomic.AddInt32(&p.activeMRFWorkers, 1)
 				replicateObject(p.ctx, v, p.objLayer)
 				atomic.AddInt32(&p.activeMRFWorkers, -1)
-				globalReplicationStats.decQ(v.Bucket, v.Size, v.DeleteMarker, v.OpType)
+				p.stats.decQ(v.Bucket, v.Size, v.DeleteMarker, v.OpType)
 
 			default:
 				bugLogIf(p.ctx, fmt.Errorf("unknown mrf replication type: %T", oi), "unknown-mrf-replicate-type")
@@ -1883,9 +1953,9 @@ func (p *ReplicationPool) AddWorker(input <-chan ReplicationWorkerOperation, opT
 				if opTracker != nil {
 					atomic.AddInt32(opTracker, 1)
 				}
-				globalReplicationStats.incQ(v.Bucket, v.Size, v.DeleteMarker, v.OpType)
+				p.stats.incQ(v.Bucket, v.Size, v.DeleteMarker, v.OpType)
 				replicateObject(p.ctx, v, p.objLayer)
-				globalReplicationStats.decQ(v.Bucket, v.Size, v.DeleteMarker, v.OpType)
+				p.stats.decQ(v.Bucket, v.Size, v.DeleteMarker, v.OpType)
 				if opTracker != nil {
 					atomic.AddInt32(opTracker, -1)
 				}
@@ -1893,10 +1963,10 @@ func (p *ReplicationPool) AddWorker(input <-chan ReplicationWorkerOperation, opT
 				if opTracker != nil {
 					atomic.AddInt32(opTracker, 1)
 				}
-				globalReplicationStats.incQ(v.Bucket, 0, true, v.OpType)
+				p.stats.incQ(v.Bucket, 0, true, v.OpType)
 
 				replicateDelete(p.ctx, v, p.objLayer)
-				globalReplicationStats.decQ(v.Bucket, 0, true, v.OpType)
+				p.stats.decQ(v.Bucket, 0, true, v.OpType)
 
 				if opTracker != nil {
 					atomic.AddInt32(opTracker, -1)
@@ -1908,23 +1978,8 @@ func (p *ReplicationPool) AddWorker(input <-chan ReplicationWorkerOperation, opT
 	}
 }
 
-// AddLargeWorkers adds a static number of workers to handle large uploads
-func (p *ReplicationPool) AddLargeWorkers() {
-	for i := 0; i < LargeWorkerCount; i++ {
-		p.lrgworkers = append(p.lrgworkers, make(chan ReplicationWorkerOperation, 100000))
-		i := i
-		go p.AddLargeWorker(p.lrgworkers[i])
-	}
-	go func() {
-		<-p.ctx.Done()
-		for i := 0; i < LargeWorkerCount; i++ {
-			xioutil.SafeClose(p.lrgworkers[i])
-		}
-	}()
-}
-
 // AddLargeWorker adds a replication worker to the static pool for large uploads.
-func (p *ReplicationPool) AddLargeWorker(input <-chan ReplicationWorkerOperation) {
+func (p *ReplicationPool) AddLargeWorker(input <-chan ReplicationWorkerOperation, opTracker *int32) {
 	for {
 		select {
 		case <-p.ctx.Done():
@@ -1935,11 +1990,23 @@ func (p *ReplicationPool) AddLargeWorker(input <-chan ReplicationWorkerOperation
 			}
 			switch v := oi.(type) {
 			case ReplicateObjectInfo:
-				globalReplicationStats.incQ(v.Bucket, v.Size, v.DeleteMarker, v.OpType)
+				if opTracker != nil {
+					atomic.AddInt32(opTracker, 1)
+				}
+				p.stats.incQ(v.Bucket, v.Size, v.DeleteMarker, v.OpType)
 				replicateObject(p.ctx, v, p.objLayer)
-				globalReplicationStats.decQ(v.Bucket, v.Size, v.DeleteMarker, v.OpType)
+				p.stats.decQ(v.Bucket, v.Size, v.DeleteMarker, v.OpType)
+				if opTracker != nil {
+					atomic.AddInt32(opTracker, -1)
+				}
 			case DeletedObjectReplicationInfo:
+				if opTracker != nil {
+					atomic.AddInt32(opTracker, 1)
+				}
 				replicateDelete(p.ctx, v, p.objLayer)
+				if opTracker != nil {
+					atomic.AddInt32(opTracker, -1)
+				}
 			default:
 				bugLogIf(p.ctx, fmt.Errorf("unknown replication type: %T", oi), "unknown-replicate-type")
 			}
@@ -1947,6 +2014,30 @@ func (p *ReplicationPool) AddLargeWorker(input <-chan ReplicationWorkerOperation
 	}
 }
 
+// ResizeLrgWorkers sets replication workers pool for large transfers(>=128MiB) to new size.
+// checkOld can be set to an expected value.
+// If the worker count changed
+func (p *ReplicationPool) ResizeLrgWorkers(n, checkOld int) {
+	p.mu.Lock()
+	defer p.mu.Unlock()
+
+	if (checkOld > 0 && len(p.lrgworkers) != checkOld) || n == len(p.lrgworkers) || n < 1 {
+		// Either already satisfied or worker count changed while we waited for the lock.
+		return
+	}
+	for len(p.lrgworkers) < n {
+		input := make(chan ReplicationWorkerOperation, 100000)
+		p.lrgworkers = append(p.lrgworkers, input)
+
+		go p.AddLargeWorker(input, &p.activeLrgWorkers)
+	}
+	for len(p.lrgworkers) > n {
+		worker := p.lrgworkers[len(p.lrgworkers)-1]
+		p.lrgworkers = p.lrgworkers[:len(p.lrgworkers)-1]
+		xioutil.SafeClose(worker)
+	}
+}
+
 // ActiveWorkers returns the number of active workers handling replication traffic.
 func (p *ReplicationPool) ActiveWorkers() int {
 	return int(atomic.LoadInt32(&p.activeWorkers))
@@ -1957,6 +2048,11 @@ func (p *ReplicationPool) ActiveMRFWorkers() int {
 	return int(atomic.LoadInt32(&p.activeMRFWorkers))
 }
 
+// ActiveLrgWorkers returns the number of active workers handling traffic > 128MiB object size.
+func (p *ReplicationPool) ActiveLrgWorkers() int {
+	return int(atomic.LoadInt32(&p.activeLrgWorkers))
+}
+
 // ResizeWorkers sets replication workers pool to new size.
 // checkOld can be set to an expected value.
 // If the worker count changed
@@ -1982,7 +2078,7 @@ func (p *ReplicationPool) ResizeWorkers(n, checkOld int) {
 }
 
 // ResizeWorkerPriority sets replication failed workers pool size
-func (p *ReplicationPool) ResizeWorkerPriority(pri string, maxWorkers int) {
+func (p *ReplicationPool) ResizeWorkerPriority(pri string, maxWorkers, maxLWorkers int) {
 	var workers, mrfWorkers int
 	p.mu.Lock()
 	switch pri {
@@ -2009,11 +2105,15 @@ func (p *ReplicationPool) ResizeWorkerPriority(pri string, maxWorkers int) {
 	if maxWorkers > 0 && mrfWorkers > maxWorkers {
 		mrfWorkers = maxWorkers
 	}
+	if maxLWorkers <= 0 {
+		maxLWorkers = LargeWorkerCount
+	}
 	p.priority = pri
 	p.maxWorkers = maxWorkers
 	p.mu.Unlock()
 	p.ResizeWorkers(workers, 0)
 	p.ResizeFailedWorkers(mrfWorkers)
+	p.ResizeLrgWorkers(maxLWorkers, 0)
 }
 
 // ResizeFailedWorkers sets replication failed workers pool size
@@ -2057,9 +2157,18 @@ func (p *ReplicationPool) queueReplicaTask(ri ReplicateObjectInfo) {
 		h := xxh3.HashString(ri.Bucket + ri.Name)
 		select {
 		case <-p.ctx.Done():
-		case p.lrgworkers[h%LargeWorkerCount] <- ri:
+		case p.lrgworkers[h%uint64(len(p.lrgworkers))] <- ri:
 		default:
-			globalReplicationPool.queueMRFSave(ri.ToMRFEntry())
+			p.queueMRFSave(ri.ToMRFEntry())
+			p.mu.RLock()
+			maxLWorkers := p.maxLWorkers
+			existing := len(p.lrgworkers)
+			p.mu.RUnlock()
+			maxLWorkers = min(maxLWorkers, LargeWorkerCount)
+			if p.ActiveLrgWorkers() < maxLWorkers {
+				workers := min(existing+1, maxLWorkers)
+				p.ResizeLrgWorkers(workers, existing)
+			}
 		}
 		return
 	}
@@ -2081,7 +2190,7 @@ func (p *ReplicationPool) queueReplicaTask(ri ReplicateObjectInfo) {
 	case healCh <- ri:
 	case ch <- ri:
 	default:
-		globalReplicationPool.queueMRFSave(ri.ToMRFEntry())
+		globalReplicationPool.Get().queueMRFSave(ri.ToMRFEntry())
 		p.mu.RLock()
 		prio := p.priority
 		maxWorkers := p.maxWorkers
@@ -2117,7 +2226,7 @@ func queueReplicateDeletesWrapper(doi DeletedObjectReplicationInfo, existingObje
 			doi.ResetID = v.ResetID
 			doi.TargetArn = k
 
-			globalReplicationPool.queueReplicaDeleteTask(doi)
+			globalReplicationPool.Get().queueReplicaDeleteTask(doi)
 		}
 	}
 }
@@ -2138,7 +2247,7 @@ func (p *ReplicationPool) queueReplicaDeleteTask(doi DeletedObjectReplicationInf
 	case <-p.ctx.Done():
 	case ch <- doi:
 	default:
-		globalReplicationPool.queueMRFSave(doi.ToMRFEntry())
+		p.queueMRFSave(doi.ToMRFEntry())
 		p.mu.RLock()
 		prio := p.priority
 		maxWorkers := p.maxWorkers
@@ -2162,14 +2271,16 @@ func (p *ReplicationPool) queueReplicaDeleteTask(doi DeletedObjectReplicationInf
 }
 
 type replicationPoolOpts struct {
-	Priority   string
-	MaxWorkers int
+	Priority    string
+	MaxWorkers  int
+	MaxLWorkers int
 }
 
 func initBackgroundReplication(ctx context.Context, objectAPI ObjectLayer) {
-	globalReplicationPool = NewReplicationPool(ctx, objectAPI, globalAPIConfig.getReplicationOpts())
-	globalReplicationStats = NewReplicationStats(ctx, objectAPI)
-	go globalReplicationStats.trackEWMA()
+	stats := NewReplicationStats(ctx, objectAPI)
+	globalReplicationPool.Set(NewReplicationPool(ctx, objectAPI, globalAPIConfig.getReplicationOpts(), stats))
+	globalReplicationStats.Store(stats)
+	go stats.trackEWMA()
 }
 
 type proxyResult struct {
@@ -2179,12 +2290,12 @@ 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
 	}
@@ -2369,11 +2480,13 @@ 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 {
-		globalReplicationPool.queueReplicaTask(ri)
+		globalReplicationPool.Get().queueReplicaTask(ri)
 	}
 }
 
@@ -2501,9 +2614,9 @@ func proxyGetTaggingToRepTarget(ctx context.Context, bucket, object string, opts
 }
 
 func scheduleReplicationDelete(ctx context.Context, dv DeletedObjectReplicationInfo, o ObjectLayer) {
-	globalReplicationPool.queueReplicaDeleteTask(dv)
+	globalReplicationPool.Get().queueReplicaDeleteTask(dv)
 	for arn := range dv.ReplicationState.Targets {
-		globalReplicationStats.Update(dv.Bucket, replicatedTargetInfo{Arn: arn, Size: 0, Duration: 0, OpType: replication.DeleteReplicationType}, replication.Pending, replication.StatusType(""))
+		globalReplicationStats.Load().Update(dv.Bucket, replicatedTargetInfo{Arn: arn, Size: 0, Duration: 0, OpType: replication.DeleteReplicationType}, replication.Pending, replication.StatusType(""))
 	}
 }
 
@@ -2849,17 +2962,19 @@ func (s *replicationResyncer) resyncBucket(ctx context.Context, objectAPI Object
 						ReplicationProxyRequest: "false",
 					},
 				})
+				sz := roi.Size
 				if err != nil {
 					if roi.DeleteMarker && isErrMethodNotAllowed(ErrorRespToObjectError(err, opts.bucket, roi.Name)) {
 						st.ReplicatedCount++
 					} else {
 						st.FailedCount++
 					}
+					sz = 0
 				} else {
 					st.ReplicatedCount++
 					st.ReplicatedSize += roi.Size
 				}
-				traceFn(err)
+				traceFn(sz, err)
 				select {
 				case <-ctx.Done():
 					return
@@ -2931,9 +3046,9 @@ func (s *replicationResyncer) start(ctx context.Context, objAPI ObjectLayer, opt
 	if len(tgtArns) == 0 {
 		return fmt.Errorf("arn %s specified for resync not found in replication config", opts.arn)
 	}
-	globalReplicationPool.resyncer.RLock()
-	data, ok := globalReplicationPool.resyncer.statusMap[opts.bucket]
-	globalReplicationPool.resyncer.RUnlock()
+	globalReplicationPool.Get().resyncer.RLock()
+	data, ok := globalReplicationPool.Get().resyncer.statusMap[opts.bucket]
+	globalReplicationPool.Get().resyncer.RUnlock()
 	if !ok {
 		data, err = loadBucketResyncMetadata(ctx, opts.bucket, objAPI)
 		if err != nil {
@@ -2959,9 +3074,9 @@ func (s *replicationResyncer) start(ctx context.Context, objAPI ObjectLayer, opt
 		return err
 	}
 
-	globalReplicationPool.resyncer.Lock()
-	defer globalReplicationPool.resyncer.Unlock()
-	brs, ok := globalReplicationPool.resyncer.statusMap[opts.bucket]
+	globalReplicationPool.Get().resyncer.Lock()
+	defer globalReplicationPool.Get().resyncer.Unlock()
+	brs, ok := globalReplicationPool.Get().resyncer.statusMap[opts.bucket]
 	if !ok {
 		brs = BucketReplicationResyncStatus{
 			Version:    resyncMetaVersion,
@@ -2969,22 +3084,22 @@ func (s *replicationResyncer) start(ctx context.Context, objAPI ObjectLayer, opt
 		}
 	}
 	brs.TargetsMap[opts.arn] = status
-	globalReplicationPool.resyncer.statusMap[opts.bucket] = brs
-	go globalReplicationPool.resyncer.resyncBucket(GlobalContext, objAPI, false, opts)
+	globalReplicationPool.Get().resyncer.statusMap[opts.bucket] = brs
+	go globalReplicationPool.Get().resyncer.resyncBucket(GlobalContext, objAPI, false, opts)
 	return nil
 }
 
-func (s *replicationResyncer) trace(resyncID string, path string) func(err error) {
+func (s *replicationResyncer) trace(resyncID string, path string) func(sz int64, err error) {
 	startTime := time.Now()
-	return func(err error) {
+	return func(sz int64, err error) {
 		duration := time.Since(startTime)
 		if globalTrace.NumSubscribers(madmin.TraceReplicationResync) > 0 {
-			globalTrace.Publish(replicationResyncTrace(resyncID, startTime, duration, path, err))
+			globalTrace.Publish(replicationResyncTrace(resyncID, startTime, duration, path, err, sz))
 		}
 	}
 }
 
-func replicationResyncTrace(resyncID string, startTime time.Time, duration time.Duration, path string, err error) madmin.TraceInfo {
+func replicationResyncTrace(resyncID string, startTime time.Time, duration time.Duration, path string, err error, sz int64) madmin.TraceInfo {
 	var errStr string
 	if err != nil {
 		errStr = err.Error()
@@ -2998,6 +3113,7 @@ func replicationResyncTrace(resyncID string, startTime time.Time, duration time.
 		Duration:  duration,
 		Path:      path,
 		Error:     errStr,
+		Bytes:     sz,
 	}
 }
 
@@ -3014,7 +3130,7 @@ func (p *ReplicationPool) deleteResyncMetadata(ctx context.Context, bucket strin
 }
 
 // initResync - initializes bucket replication resync for all buckets.
-func (p *ReplicationPool) initResync(ctx context.Context, buckets []BucketInfo, objAPI ObjectLayer) error {
+func (p *ReplicationPool) initResync(ctx context.Context, buckets []string, objAPI ObjectLayer) error {
 	if objAPI == nil {
 		return errServerNotInitialized
 	}
@@ -3023,7 +3139,7 @@ func (p *ReplicationPool) initResync(ctx context.Context, buckets []BucketInfo,
 	return nil
 }
 
-func (p *ReplicationPool) startResyncRoutine(ctx context.Context, buckets []BucketInfo, objAPI ObjectLayer) {
+func (p *ReplicationPool) startResyncRoutine(ctx context.Context, buckets []string, objAPI ObjectLayer) {
 	r := rand.New(rand.NewSource(time.Now().UnixNano()))
 	// Run the replication resync in a loop
 	for {
@@ -3041,13 +3157,13 @@ func (p *ReplicationPool) startResyncRoutine(ctx context.Context, buckets []Buck
 }
 
 // Loads bucket replication resync statuses into memory.
-func (p *ReplicationPool) loadResync(ctx context.Context, buckets []BucketInfo, objAPI ObjectLayer) error {
+func (p *ReplicationPool) loadResync(ctx context.Context, buckets []string, objAPI ObjectLayer) error {
 	// Make sure only one node running resync on the cluster.
 	ctx, cancel := globalLeaderLock.GetLock(ctx)
 	defer cancel()
 
 	for index := range buckets {
-		bucket := buckets[index].Name
+		bucket := buckets[index]
 
 		meta, err := loadBucketResyncMetadata(ctx, bucket, objAPI)
 		if err != nil {
@@ -3301,7 +3417,7 @@ func queueReplicationHeal(ctx context.Context, bucket string, oi ObjectInfo, rcf
 		if roi.ReplicationStatus == replication.Pending ||
 			roi.ReplicationStatus == replication.Failed ||
 			roi.VersionPurgeStatus == Failed || roi.VersionPurgeStatus == Pending {
-			globalReplicationPool.queueReplicaDeleteTask(dv)
+			globalReplicationPool.Get().queueReplicaDeleteTask(dv)
 			return
 		}
 		// if replication status is Complete on DeleteMarker and existing object resync required
@@ -3317,12 +3433,12 @@ func queueReplicationHeal(ctx context.Context, bucket string, oi ObjectInfo, rcf
 	switch roi.ReplicationStatus {
 	case replication.Pending, replication.Failed:
 		roi.EventType = ReplicateHeal
-		globalReplicationPool.queueReplicaTask(roi)
+		globalReplicationPool.Get().queueReplicaTask(roi)
 		return
 	}
 	if roi.ExistingObjResync.mustResync() {
 		roi.EventType = ReplicateExisting
-		globalReplicationPool.queueReplicaTask(roi)
+		globalReplicationPool.Get().queueReplicaTask(roi)
 	}
 	return
 }
@@ -3387,8 +3503,8 @@ func (p *ReplicationPool) queueMRFSave(entry MRFReplicateEntry) {
 		return
 	}
 	if entry.RetryCount > mrfRetryLimit { // let scanner catch up if retry count exceeded
-		atomic.AddUint64(&globalReplicationStats.mrfStats.TotalDroppedCount, 1)
-		atomic.AddUint64(&globalReplicationStats.mrfStats.TotalDroppedBytes, uint64(entry.sz))
+		atomic.AddUint64(&p.stats.mrfStats.TotalDroppedCount, 1)
+		atomic.AddUint64(&p.stats.mrfStats.TotalDroppedBytes, uint64(entry.sz))
 		return
 	}
 
@@ -3401,8 +3517,8 @@ func (p *ReplicationPool) queueMRFSave(entry MRFReplicateEntry) {
 		select {
 		case p.mrfSaveCh <- entry:
 		default:
-			atomic.AddUint64(&globalReplicationStats.mrfStats.TotalDroppedCount, 1)
-			atomic.AddUint64(&globalReplicationStats.mrfStats.TotalDroppedBytes, uint64(entry.sz))
+			atomic.AddUint64(&p.stats.mrfStats.TotalDroppedCount, 1)
+			atomic.AddUint64(&p.stats.mrfStats.TotalDroppedBytes, uint64(entry.sz))
 		}
 	}
 }
@@ -3433,7 +3549,7 @@ func (p *ReplicationPool) persistToDrive(ctx context.Context, v MRFReplicateEntr
 	}
 
 	globalLocalDrivesMu.RLock()
-	localDrives := cloneDrives(globalLocalDrives)
+	localDrives := cloneDrives(globalLocalDrivesMap)
 	globalLocalDrivesMu.RUnlock()
 
 	for _, localDrive := range localDrives {
@@ -3451,7 +3567,7 @@ func (p *ReplicationPool) saveMRFEntries(ctx context.Context, entries map[string
 	if !p.initialized() {
 		return
 	}
-	atomic.StoreUint64(&globalReplicationStats.mrfStats.LastFailedCount, uint64(len(entries)))
+	atomic.StoreUint64(&p.stats.mrfStats.LastFailedCount, uint64(len(entries)))
 	if len(entries) == 0 {
 		return
 	}
@@ -3500,7 +3616,7 @@ func (p *ReplicationPool) loadMRF() (mrfRec MRFReplicateEntries, err error) {
 	}
 
 	globalLocalDrivesMu.RLock()
-	localDrives := cloneDrives(globalLocalDrives)
+	localDrives := cloneDrives(globalLocalDrivesMap)
 	globalLocalDrivesMu.RUnlock()
 
 	for _, localDrive := range localDrives {
@@ -3621,3 +3737,12 @@ func (p *ReplicationPool) getMRF(ctx context.Context, bucket string) (ch <-chan
 
 	return mrfCh, nil
 }
+
+// validateReplicationDestinationOptions is used to configure the validation of the replication destination.
+// validateReplicationDestination uses this to configure the validation.
+type validateReplicationDestinationOptions struct {
+	CheckRemoteBucket bool
+	CheckReady        bool
+
+	checkReadyErr sync.Map
+}

cmd/bucket-stats.go

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

cmd/bucket-targets.go

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

cmd/common-main.go

@@ -402,6 +402,29 @@ func buildServerCtxt(ctx *cli.Context, ctxt *serverCtxt) (err error) {
 		ctxt.certsDirSet = true
 	}
 
+	memAvailable := availableMemory()
+	if ctx.IsSet("memlimit") || ctx.GlobalIsSet("memlimit") {
+		memlimit := ctx.String("memlimit")
+		if memlimit == "" {
+			memlimit = ctx.GlobalString("memlimit")
+		}
+		mlimit, err := humanize.ParseBytes(memlimit)
+		if err != nil {
+			return err
+		}
+		if mlimit > memAvailable {
+			logger.Info("WARNING: maximum memory available (%s) smaller than specified --memlimit=%s, ignoring --memlimit value",
+				humanize.IBytes(memAvailable), memlimit)
+		}
+		ctxt.MemLimit = mlimit
+	} else {
+		ctxt.MemLimit = memAvailable
+	}
+
+	if memAvailable < ctxt.MemLimit {
+		ctxt.MemLimit = memAvailable
+	}
+
 	ctxt.FTP = ctx.StringSlice("ftp")
 	ctxt.SFTP = ctx.StringSlice("sftp")
 	ctxt.Interface = ctx.String("interface")
@@ -410,7 +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)
@@ -663,16 +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) {
@@ -746,7 +758,7 @@ func serverHandleEnvVars() {
 	if len(domains) != 0 {
 		for _, domainName := range strings.Split(domains, config.ValueSeparator) {
 			if _, ok := dns2.IsDomainName(domainName); !ok {
-				logger.Fatal(config.ErrInvalidDomainValue(nil).Msg("Unknown value `%s`", domainName),
+				logger.Fatal(config.ErrInvalidDomainValue(nil).Msgf("Unknown value `%s`", domainName),
 					"Invalid MINIO_DOMAIN value in environment variable")
 			}
 			globalDomainNames = append(globalDomainNames, domainName)
@@ -755,7 +767,7 @@ func serverHandleEnvVars() {
 		lcpSuf := lcpSuffix(globalDomainNames)
 		for _, domainName := range globalDomainNames {
 			if domainName == lcpSuf && len(globalDomainNames) > 1 {
-				logger.Fatal(config.ErrOverlappingDomainValue(nil).Msg("Overlapping domains `%s` not allowed", globalDomainNames),
+				logger.Fatal(config.ErrOverlappingDomainValue(nil).Msgf("Overlapping domains `%s` not allowed", globalDomainNames),
 					"Invalid MINIO_DOMAIN value in environment variable")
 			}
 		}
@@ -812,7 +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() {
@@ -821,6 +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, "")
@@ -829,6 +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
@@ -837,8 +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"+
@@ -852,6 +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{
@@ -240,6 +240,11 @@ func initHelp() {
 			Description: "manage Browser HTTP specific features, such as Security headers, etc.",
 			Optional:    true,
 		},
+		config.HelpKV{
+			Key:         config.ILMSubSys,
+			Description: "manage ILM settings for expiration and transition workers",
+			Optional:    true,
+		},
 	}
 
 	if globalIsErasure {
@@ -288,6 +293,7 @@ func initHelp() {
 		config.DriveSubSys:          drive.HelpDrive,
 		config.CacheSubSys:          cache.Help,
 		config.BrowserSubSys:        browser.Help,
+		config.ILMSubSys:            ilm.Help,
 	}
 
 	config.RegisterHelpSubSys(helpMap)

cmd/data-scanner.go

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

cmd/data-usage_test.go

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

cmd/dummy-data-generator_test.go

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

cmd/dummy-handlers.go

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

cmd/encryption-v1.go

@@ -134,11 +134,16 @@ func DecryptETags(ctx context.Context, k *kms.KMS, objects []ObjectInfo) error {
 		SSES3SinglePartObjects := make(map[int]bool)
 		for i, object := range batch {
 			if kind, ok := crypto.IsEncrypted(object.UserDefined); ok && kind == crypto.S3 && !crypto.IsMultiPart(object.UserDefined) {
-				SSES3SinglePartObjects[i] = true
-
-				metadata = append(metadata, object.UserDefined)
-				buckets = append(buckets, object.Bucket)
-				names = append(names, object.Name)
+				ETag, err := etag.Parse(object.ETag)
+				if err != nil {
+					continue
+				}
+				if ETag.IsEncrypted() {
+					SSES3SinglePartObjects[i] = true
+					metadata = append(metadata, object.UserDefined)
+					buckets = append(buckets, object.Bucket)
+					names = append(names, object.Name)
+				}
 			}
 		}
 
@@ -190,7 +195,7 @@ func DecryptETags(ctx context.Context, k *kms.KMS, objects []ObjectInfo) error {
 				if err != nil {
 					return err
 				}
-				if SSES3SinglePartObjects[i] && ETag.IsEncrypted() {
+				if SSES3SinglePartObjects[i] {
 					ETag, err = etag.Decrypt(keys[0][:], ETag)
 					if err != nil {
 						return err
@@ -596,28 +601,44 @@ func DecryptBlocksRequestR(inputReader io.Reader, h http.Header, seqNumber uint3
 	partEncRelOffset := int64(seqNumber) * (SSEDAREPackageBlockSize + SSEDAREPackageMetaSize)
 
 	w := &DecryptBlocksReader{
-		reader:            inputReader,
-		startSeqNum:       seqNumber,
-		partDecRelOffset:  partDecRelOffset,
-		partEncRelOffset:  partEncRelOffset,
-		parts:             oi.Parts,
-		partIndex:         partStart,
-		header:            h,
-		bucket:            bucket,
-		object:            object,
-		customerKeyHeader: h.Get(xhttp.AmzServerSideEncryptionCustomerKey),
-		copySource:        copySource,
-		metadata:          cloneMSS(oi.UserDefined),
+		reader:           inputReader,
+		startSeqNum:      seqNumber,
+		partDecRelOffset: partDecRelOffset,
+		partEncRelOffset: partEncRelOffset,
+		parts:            oi.Parts,
+		partIndex:        partStart,
 	}
 
-	if w.copySource {
-		w.customerKeyHeader = h.Get(xhttp.AmzServerSideEncryptionCopyCustomerKey)
+	// In case of SSE-C, we have to decrypt the OEK using the client-provided key.
+	// In case of a SSE-C server-side copy, the client might provide two keys,
+	// one for the source and one for the target. This reader is the source.
+	var ssecClientKey []byte
+	if crypto.SSEC.IsEncrypted(oi.UserDefined) {
+		if copySource && crypto.SSECopy.IsRequested(h) {
+			key, err := crypto.SSECopy.ParseHTTP(h)
+			if err != nil {
+				return nil, err
+			}
+			ssecClientKey = key[:]
+		} else {
+			key, err := crypto.SSEC.ParseHTTP(h)
+			if err != nil {
+				return nil, err
+			}
+			ssecClientKey = key[:]
+		}
 	}
 
+	// Decrypt the OEK once and reuse it for all subsequent parts.
+	objectEncryptionKey, err := decryptObjectMeta(ssecClientKey, bucket, object, oi.UserDefined)
+	if err != nil {
+		return nil, err
+	}
+	w.objectEncryptionKey = objectEncryptionKey
+
 	if err := w.buildDecrypter(w.parts[w.partIndex].Number); err != nil {
 		return nil, err
 	}
-
 	return w, nil
 }
 
@@ -633,48 +654,17 @@ type DecryptBlocksReader struct {
 	// Current part index
 	partIndex int
 	// Parts information
-	parts          []ObjectPartInfo
-	header         http.Header
-	bucket, object string
-	metadata       map[string]string
+	parts []ObjectPartInfo
 
+	objectEncryptionKey                []byte
 	partDecRelOffset, partEncRelOffset int64
-
-	copySource bool
-	// Customer Key
-	customerKeyHeader string
 }
 
 func (d *DecryptBlocksReader) buildDecrypter(partID int) error {
-	m := cloneMSS(d.metadata)
-	// Initialize the first decrypter; new decrypters will be
-	// initialized in Read() operation as needed.
-	var key []byte
-	var err error
-	if d.copySource {
-		if crypto.SSEC.IsEncrypted(d.metadata) {
-			d.header.Set(xhttp.AmzServerSideEncryptionCopyCustomerKey, d.customerKeyHeader)
-			key, err = ParseSSECopyCustomerRequest(d.header, d.metadata)
-		}
-	} else {
-		if crypto.SSEC.IsEncrypted(d.metadata) {
-			d.header.Set(xhttp.AmzServerSideEncryptionCustomerKey, d.customerKeyHeader)
-			key, err = ParseSSECustomerHeader(d.header)
-		}
-	}
-	if err != nil {
-		return err
-	}
-
-	objectEncryptionKey, err := decryptObjectMeta(key, d.bucket, d.object, m)
-	if err != nil {
-		return err
-	}
-
 	var partIDbin [4]byte
 	binary.LittleEndian.PutUint32(partIDbin[:], uint32(partID)) // marshal part ID
 
-	mac := hmac.New(sha256.New, objectEncryptionKey) // derive part encryption key from part ID and object key
+	mac := hmac.New(sha256.New, d.objectEncryptionKey) // derive part encryption key from part ID and object key
 	mac.Write(partIDbin[:])
 	partEncryptionKey := mac.Sum(nil)
 
@@ -738,8 +728,9 @@ func (d *DecryptBlocksReader) Read(p []byte) (int, error) {
 // but has an invalid size.
 func (o ObjectInfo) DecryptedSize() (int64, error) {
 	if _, ok := crypto.IsEncrypted(o.UserDefined); !ok {
-		return 0, errors.New("Cannot compute decrypted size of an unencrypted object")
+		return -1, errors.New("Cannot compute decrypted size of an unencrypted object")
 	}
+
 	if !o.isMultipart() {
 		size, err := sio.DecryptedSize(uint64(o.Size))
 		if err != nil {
@@ -752,7 +743,7 @@ func (o ObjectInfo) DecryptedSize() (int64, error) {
 	for _, part := range o.Parts {
 		partSize, err := sio.DecryptedSize(uint64(part.Size))
 		if err != nil {
-			return 0, errObjectTampered
+			return -1, errObjectTampered
 		}
 		size += int64(partSize)
 	}
@@ -1025,7 +1016,9 @@ 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 +1070,16 @@ 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
 		}
@@ -1093,15 +1089,15 @@ func (o *ObjectInfo) metadataDecrypter() objectMetaDecryptFn {
 	}
 }
 
-// decryptChecksums will attempt to decode checksums and return it/them if set.
+// decryptPartsChecksums will attempt to decode checksums and return it/them if set.
 // if part > 0, and we have the checksum for the part that will be returned.
-func (o *ObjectInfo) decryptPartsChecksums() {
+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 +1153,25 @@ 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-ellipses.go

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

cmd/endpoint.go

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

cmd/erasure-common.go

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

cmd/erasure-decode.go

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

cmd/erasure-healing-common.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 (
@@ -220,21 +226,28 @@ func (er *erasureObjects) auditHealObject(ctx context.Context, bucket, object, v
 	opts := AuditLogOptions{
 		Event:     "HealObject",
 		Bucket:    bucket,
-		Object:    object,
+		Object:    decodeDirObject(object),
 		VersionID: versionID,
 	}
 	if err != nil {
 		opts.Error = err.Error()
 	}
 
-	opts.Tags = map[string]interface{}{
-		"healResult": result,
-		"objectLocation": auditObjectOp{
-			Name:   decodeDirObject(object),
-			Pool:   er.poolIndex + 1,
-			Set:    er.setIndex + 1,
-			Drives: er.getEndpointStrings(),
-		},
+	b, a := result.GetCorruptedCounts()
+	if b == a {
+		opts.Error = fmt.Sprintf("unable to heal %d corrupted blocks on drives", b)
+	}
+	b, a = result.GetMissingCounts()
+	if b == a {
+		opts.Error = fmt.Sprintf("unable to heal %d missing blocks on drives", b)
+	}
+
+	opts.Tags = map[string]string{
+		"healObject": auditObjectOp{
+			Name: opts.Object,
+			Pool: er.poolIndex + 1,
+			Set:  er.setIndex + 1,
+		}.String(),
 	}
 
 	auditLogInternal(ctx, opts)
@@ -332,7 +345,7 @@ 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 +368,20 @@ 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 +398,6 @@ 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 +424,7 @@ 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 +636,7 @@ func (er *erasureObjects) healObject(ctx context.Context, bucket string, object
 		}
 
 		for i, v := range result.Before.Drives {
-			if v.Endpoint == disk.String() {
+			if v.Endpoint == disk.Endpoint().String() {
 				result.After.Drives[i].State = madmin.DriveStateOk
 			}
 		}
@@ -908,35 +920,52 @@ 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 +977,7 @@ 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
@@ -1065,6 +1094,7 @@ func healTrace(funcName healingMetric, startTime time.Time, bucket, object strin
 		if result != nil {
 			tr.Custom["version-id"] = result.VersionID
 			tr.Custom["disks"] = strconv.Itoa(result.DiskCount)
+			tr.Bytes = result.ObjectSize
 		}
 	}
 	if err != nil {

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

cmd/erasure-metadata_test.go

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

cmd/erasure-multipart.go

@@ -80,6 +80,14 @@ func (er erasureObjects) checkUploadIDExists(ctx context.Context, bucket, object
 		return fi, nil, err
 	}
 
+	if readQuorum < 0 {
+		return fi, nil, errErasureReadQuorum
+	}
+
+	if writeQuorum < 0 {
+		return fi, nil, errErasureWriteQuorum
+	}
+
 	quorum := readQuorum
 	if write {
 		quorum = writeQuorum
@@ -88,14 +96,13 @@ func (er erasureObjects) checkUploadIDExists(ctx context.Context, bucket, object
 	// List all online disks.
 	_, modTime, etag := listOnlineDisks(storageDisks, partsMetadata, errs, quorum)
 
-	var reducedErr error
 	if write {
-		reducedErr = reduceWriteQuorumErrs(ctx, errs, objectOpIgnoredErrs, writeQuorum)
+		err = reduceWriteQuorumErrs(ctx, errs, objectOpIgnoredErrs, writeQuorum)
 	} else {
-		reducedErr = reduceReadQuorumErrs(ctx, errs, objectOpIgnoredErrs, readQuorum)
+		err = reduceReadQuorumErrs(ctx, errs, objectOpIgnoredErrs, readQuorum)
 	}
-	if reducedErr != nil {
-		return fi, nil, reducedErr
+	if err != nil {
+		return fi, nil, err
 	}
 
 	// Pick one from the first valid metadata.
@@ -103,10 +110,9 @@ func (er erasureObjects) checkUploadIDExists(ctx context.Context, bucket, object
 	return fi, partsMetadata, err
 }
 
-// Removes part.meta given by partName belonging to a multipart upload from minioMetaBucket
-func (er erasureObjects) removePartMeta(bucket, object, uploadID, dataDir string, partNumber int) {
-	uploadIDPath := er.getUploadIDDir(bucket, object, uploadID)
-	curpartPath := pathJoin(uploadIDPath, dataDir, fmt.Sprintf("part.%d", partNumber))
+// cleanupMultipartPath removes all extraneous files and parts from the multipart folder, this is used per CompleteMultipart.
+// do not use this function outside of completeMultipartUpload()
+func (er erasureObjects) cleanupMultipartPath(ctx context.Context, paths ...string) {
 	storageDisks := er.getDisks()
 
 	g := errgroup.WithNErrs(len(storageDisks))
@@ -116,42 +122,7 @@ func (er erasureObjects) removePartMeta(bucket, object, uploadID, dataDir string
 		}
 		index := index
 		g.Go(func() error {
-			_ = storageDisks[index].Delete(context.TODO(), minioMetaMultipartBucket, curpartPath+".meta", DeleteOptions{
-				Recursive: false,
-				Immediate: false,
-			})
-
-			return nil
-		}, index)
-	}
-	g.Wait()
-}
-
-// Removes part given by partName belonging to a multipart upload from minioMetaBucket
-func (er erasureObjects) removeObjectPart(bucket, object, uploadID, dataDir string, partNumber int) {
-	uploadIDPath := er.getUploadIDDir(bucket, object, uploadID)
-	curpartPath := pathJoin(uploadIDPath, dataDir, fmt.Sprintf("part.%d", partNumber))
-	storageDisks := er.getDisks()
-
-	g := errgroup.WithNErrs(len(storageDisks))
-	for index, disk := range storageDisks {
-		if disk == nil {
-			continue
-		}
-		index := index
-		g.Go(func() error {
-			// Ignoring failure to remove parts that weren't present in CompleteMultipartUpload
-			// requests. xl.meta is the authoritative source of truth on which parts constitute
-			// the object. The presence of parts that don't belong in the object doesn't affect correctness.
-			_ = storageDisks[index].Delete(context.TODO(), minioMetaMultipartBucket, curpartPath, DeleteOptions{
-				Recursive: false,
-				Immediate: false,
-			})
-			_ = storageDisks[index].Delete(context.TODO(), minioMetaMultipartBucket, curpartPath+".meta", DeleteOptions{
-				Recursive: false,
-				Immediate: false,
-			})
-
+			_ = storageDisks[index].DeleteBulk(ctx, minioMetaMultipartBucket, paths...)
 			return nil
 		}, index)
 	}
@@ -159,7 +130,7 @@ func (er erasureObjects) removeObjectPart(bucket, object, uploadID, dataDir stri
 }
 
 // Clean-up the old multipart uploads. Should be run in a Go routine.
-func (er erasureObjects) cleanupStaleUploads(ctx context.Context, expiry time.Duration) {
+func (er erasureObjects) cleanupStaleUploads(ctx context.Context) {
 	// run multiple cleanup's local to this server.
 	var wg sync.WaitGroup
 	for _, disk := range er.getLocalDisks() {
@@ -167,7 +138,7 @@ func (er erasureObjects) cleanupStaleUploads(ctx context.Context, expiry time.Du
 			wg.Add(1)
 			go func(disk StorageAPI) {
 				defer wg.Done()
-				er.cleanupStaleUploadsOnDisk(ctx, disk, expiry)
+				er.cleanupStaleUploadsOnDisk(ctx, disk)
 			}(disk)
 		}
 	}
@@ -193,7 +164,7 @@ func (er erasureObjects) deleteAll(ctx context.Context, bucket, prefix string) {
 }
 
 // Remove the old multipart uploads on the given disk.
-func (er erasureObjects) cleanupStaleUploadsOnDisk(ctx context.Context, disk StorageAPI, expiry time.Duration) {
+func (er erasureObjects) cleanupStaleUploadsOnDisk(ctx context.Context, disk StorageAPI) {
 	drivePath := disk.Endpoint().Path
 
 	readDirFn(pathJoin(drivePath, minioMetaMultipartBucket), func(shaDir string, typ os.FileMode) error {
@@ -219,7 +190,7 @@ func (er erasureObjects) cleanupStaleUploadsOnDisk(ctx context.Context, disk Sto
 				modTime = fi.ModTime
 				wait()
 			}
-			if time.Since(modTime) < expiry {
+			if time.Since(modTime) < globalAPIConfig.getStaleUploadsExpiry() {
 				return nil
 			}
 			w := xioutil.NewDeadlineWorker(globalDriveConfig.GetMaxTimeout())
@@ -238,7 +209,7 @@ func (er erasureObjects) cleanupStaleUploadsOnDisk(ctx context.Context, disk Sto
 			return nil
 		}
 		// Modtime is returned in the Created field. See (*xlStorage).StatVol
-		if time.Since(vi.Created) < expiry {
+		if time.Since(vi.Created) < globalAPIConfig.getStaleUploadsExpiry() {
 			return nil
 		}
 		w := xioutil.NewDeadlineWorker(globalDriveConfig.GetMaxTimeout())
@@ -267,7 +238,7 @@ func (er erasureObjects) cleanupStaleUploadsOnDisk(ctx context.Context, disk Sto
 		w := xioutil.NewDeadlineWorker(globalDriveConfig.GetMaxTimeout())
 		return w.Run(func() error {
 			wait := deleteMultipartCleanupSleeper.Timer(ctx)
-			if time.Since(vi.Created) > expiry {
+			if time.Since(vi.Created) > globalAPIConfig.getStaleUploadsExpiry() {
 				pathUUID := mustGetUUID()
 				targetPath := pathJoin(drivePath, minioMetaTmpDeletedBucket, pathUUID)
 
@@ -287,7 +258,7 @@ func (er erasureObjects) cleanupStaleUploadsOnDisk(ctx context.Context, disk Sto
 // towards simplification of multipart APIs.
 // The resulting ListMultipartsInfo structure is unmarshalled directly as XML.
 func (er erasureObjects) ListMultipartUploads(ctx context.Context, bucket, object, keyMarker, uploadIDMarker, delimiter string, maxUploads int) (result ListMultipartsInfo, err error) {
-	auditObjectErasureSet(ctx, object, &er)
+	auditObjectErasureSet(ctx, "ListMultipartUploads", object, &er)
 
 	result.MaxUploads = maxUploads
 	result.KeyMarker = keyMarker
@@ -483,6 +454,11 @@ 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
@@ -521,9 +497,10 @@ func (er erasureObjects) newMultipartUpload(ctx context.Context, bucket string,
 	uploadIDPath := er.getUploadIDDir(bucket, object, uploadUUID)
 
 	// Write updated `xl.meta` to all disks.
-	if _, err := writeUniqueFileInfo(ctx, onlineDisks, bucket, minioMetaMultipartBucket, uploadIDPath, partsMetadata, writeQuorum); err != nil {
+	if _, err := writeAllMetadata(ctx, onlineDisks, bucket, minioMetaMultipartBucket, uploadIDPath, partsMetadata, writeQuorum); err != nil {
 		return nil, toObjectErr(err, bucket, object)
 	}
+
 	return &NewMultipartUploadResult{
 		UploadID:     uploadID,
 		ChecksumAlgo: userDefined[hash.MinIOMultipartChecksum],
@@ -537,14 +514,14 @@ func (er erasureObjects) newMultipartUpload(ctx context.Context, bucket string,
 // Implements S3 compatible initiate multipart API.
 func (er erasureObjects) NewMultipartUpload(ctx context.Context, bucket, object string, opts ObjectOptions) (*NewMultipartUploadResult, error) {
 	if !opts.NoAuditLog {
-		auditObjectErasureSet(ctx, object, &er)
+		auditObjectErasureSet(ctx, "NewMultipartUpload", object, &er)
 	}
 
 	return er.newMultipartUpload(ctx, bucket, object, opts)
 }
 
 // renamePart - renames multipart part to its relevant location under uploadID.
-func renamePart(ctx context.Context, disks []StorageAPI, srcBucket, srcEntry, dstBucket, dstEntry string, writeQuorum int) ([]StorageAPI, error) {
+func (er erasureObjects) renamePart(ctx context.Context, disks []StorageAPI, srcBucket, srcEntry, dstBucket, dstEntry string, optsMeta []byte, writeQuorum int) ([]StorageAPI, error) {
 	g := errgroup.WithNErrs(len(disks))
 
 	// Rename file on all underlying storage disks.
@@ -554,60 +531,25 @@ func renamePart(ctx context.Context, disks []StorageAPI, srcBucket, srcEntry, ds
 			if disks[index] == nil {
 				return errDiskNotFound
 			}
-			return disks[index].RenameFile(ctx, srcBucket, srcEntry, dstBucket, dstEntry)
+			return disks[index].RenamePart(ctx, srcBucket, srcEntry, dstBucket, dstEntry, optsMeta)
 		}, index)
 	}
 
 	// Wait for all renames to finish.
 	errs := g.Wait()
 
-	// Do not need to undo partial successful operation since those will be cleaned up
-	// in 24hrs via multipart cleaner, never rename() back to `.minio.sys/tmp` as there
-	// is no way to clean them.
-
-	// We can safely allow RenameFile errors up to len(er.getDisks()) - writeQuorum
-	// otherwise return failure. Cleanup successful renames.
-	return evalDisks(disks, errs), reduceWriteQuorumErrs(ctx, errs, objectOpIgnoredErrs, writeQuorum)
-}
-
-// writeAllDisks - writes 'b' to all provided disks.
-// If write cannot reach quorum, the files will be deleted from all disks.
-func writeAllDisks(ctx context.Context, disks []StorageAPI, dstBucket, dstEntry string, b []byte, writeQuorum int) ([]StorageAPI, error) {
-	g := errgroup.WithNErrs(len(disks))
-
-	// Write file to all underlying storage disks.
-	for index := range disks {
-		index := index
-		g.Go(func() error {
-			if disks[index] == nil {
-				return errDiskNotFound
-			}
-			return disks[index].WriteAll(ctx, dstBucket, dstEntry, b)
-		}, index)
+	paths := []string{
+		dstEntry,
+		dstEntry + ".meta",
 	}
 
-	// Wait for all renames to finish.
-	errs := g.Wait()
-
-	// We can safely allow RenameFile errors up to len(er.getDisks()) - writeQuorum
-	// otherwise return failure. Cleanup successful renames.
 	err := reduceWriteQuorumErrs(ctx, errs, objectOpIgnoredErrs, writeQuorum)
-	if errors.Is(err, errErasureWriteQuorum) {
-		// Remove all written
-		g := errgroup.WithNErrs(len(disks))
-		for index := range disks {
-			if disks[index] == nil || errs[index] != nil {
-				continue
-			}
-			index := index
-			g.Go(func() error {
-				return disks[index].Delete(ctx, dstBucket, dstEntry, DeleteOptions{Immediate: true})
-			}, index)
-		}
-		// Ignore these errors.
-		g.WaitErr()
+	if err != nil {
+		er.cleanupMultipartPath(ctx, paths...)
 	}
 
+	// We can safely allow RenameFile errors up to len(er.getDisks()) - writeQuorum
+	// otherwise return failure. Cleanup successful renames.
 	return evalDisks(disks, errs), err
 }
 
@@ -618,7 +560,7 @@ func writeAllDisks(ctx context.Context, disks []StorageAPI, dstBucket, dstEntry
 // Implements S3 compatible Upload Part API.
 func (er erasureObjects) PutObjectPart(ctx context.Context, bucket, object, uploadID string, partID int, r *PutObjReader, opts ObjectOptions) (pi PartInfo, err error) {
 	if !opts.NoAuditLog {
-		auditObjectErasureSet(ctx, object, &er)
+		auditObjectErasureSet(ctx, "PutObjectPart", object, &er)
 	}
 
 	data := r.Reader
@@ -763,19 +705,6 @@ func (er erasureObjects) PutObjectPart(ctx context.Context, bucket, object, uplo
 
 	// Rename temporary part file to its final location.
 	partPath := pathJoin(uploadIDPath, fi.DataDir, partSuffix)
-	onlineDisks, err = renamePart(ctx, onlineDisks, minioMetaTmpBucket, tmpPartPath, minioMetaMultipartBucket, partPath, writeQuorum)
-	if err != nil {
-		if errors.Is(err, errFileNotFound) {
-			// An in-quorum errFileNotFound means that client stream
-			// prematurely closed and we do not find any xl.meta or
-			// part.1's - in such a scenario we must return as if client
-			// disconnected. This means that erasure.Encode() CreateFile()
-			// did not do anything.
-			return pi, IncompleteBody{Bucket: bucket, Object: object}
-		}
-
-		return pi, toObjectErr(err, minioMetaMultipartBucket, partPath)
-	}
 
 	md5hex := r.MD5CurrentHexString()
 	if opts.PreserveETag != "" {
@@ -797,15 +726,22 @@ func (er erasureObjects) PutObjectPart(ctx context.Context, bucket, object, uplo
 		Checksums:  r.ContentCRC(),
 	}
 
-	fi.Parts = []ObjectPartInfo{partInfo}
-	partFI, err := fi.MarshalMsg(nil)
+	partFI, err := partInfo.MarshalMsg(nil)
 	if err != nil {
 		return pi, toObjectErr(err, minioMetaMultipartBucket, partPath)
 	}
 
-	// Write part metadata to all disks.
-	onlineDisks, err = writeAllDisks(ctx, onlineDisks, minioMetaMultipartBucket, partPath+".meta", partFI, writeQuorum)
+	onlineDisks, err = er.renamePart(ctx, onlineDisks, minioMetaTmpBucket, tmpPartPath, minioMetaMultipartBucket, partPath, partFI, writeQuorum)
 	if err != nil {
+		if errors.Is(err, errFileNotFound) {
+			// An in-quorum errFileNotFound means that client stream
+			// prematurely closed and we do not find any xl.meta or
+			// part.1's - in such a scenario we must return as if client
+			// disconnected. This means that erasure.Encode() CreateFile()
+			// did not do anything.
+			return pi, IncompleteBody{Bucket: bucket, Object: object}
+		}
+
 		return pi, toObjectErr(err, minioMetaMultipartBucket, partPath)
 	}
 
@@ -830,7 +766,7 @@ func (er erasureObjects) PutObjectPart(ctx context.Context, bucket, object, uplo
 // Does not contain currently uploaded parts by design.
 func (er erasureObjects) GetMultipartInfo(ctx context.Context, bucket, object, uploadID string, opts ObjectOptions) (MultipartInfo, error) {
 	if !opts.NoAuditLog {
-		auditObjectErasureSet(ctx, object, &er)
+		auditObjectErasureSet(ctx, "GetMultipartInfo", object, &er)
 	}
 
 	result := MultipartInfo{
@@ -868,7 +804,7 @@ func (er erasureObjects) GetMultipartInfo(ctx context.Context, bucket, object, u
 // replied back to the client.
 func (er erasureObjects) ListObjectParts(ctx context.Context, bucket, object, uploadID string, partNumberMarker, maxParts int, opts ObjectOptions) (result ListPartsInfo, err error) {
 	if !opts.NoAuditLog {
-		auditObjectErasureSet(ctx, object, &er)
+		auditObjectErasureSet(ctx, "ListObjectParts", object, &er)
 	}
 
 	uploadIDLock := er.NewNSLock(bucket, pathJoin(object, uploadID))
@@ -948,7 +884,7 @@ func (er erasureObjects) ListObjectParts(ctx context.Context, bucket, object, up
 
 	g := errgroup.WithNErrs(len(req.Files)).WithConcurrency(32)
 
-	partsInfo := make([]ObjectPartInfo, len(req.Files))
+	partsInfo := make([]*ObjectPartInfo, len(req.Files))
 	for i, file := range req.Files {
 		file := file
 		partN := i + start
@@ -960,21 +896,17 @@ func (er erasureObjects) ListObjectParts(ctx context.Context, bucket, object, up
 				return err
 			}
 
-			var pfi FileInfo
-			_, err = pfi.UnmarshalMsg(buf)
+			pinfo := &ObjectPartInfo{}
+			_, err = pinfo.UnmarshalMsg(buf)
 			if err != nil {
 				return err
 			}
 
-			if len(pfi.Parts) != 1 {
-				return errors.New("invalid number of parts expected 1, got 0")
+			if partN != pinfo.Number {
+				return fmt.Errorf("part.%d.meta has incorrect corresponding part number: expected %d, got %d", partN, partN, pinfo.Number)
 			}
 
-			if partN != pfi.Parts[0].Number {
-				return fmt.Errorf("part.%d.meta has incorrect corresponding part number: expected %d, got %d", partN, partN, pfi.Parts[0].Number)
-			}
-
-			partsInfo[i] = pfi.Parts[0]
+			partsInfo[i] = pinfo
 			return nil
 		}, i)
 	}
@@ -982,7 +914,7 @@ func (er erasureObjects) ListObjectParts(ctx context.Context, bucket, object, up
 	g.Wait()
 
 	for _, part := range partsInfo {
-		if part.Number != 0 && !part.ModTime.IsZero() {
+		if part != nil && part.Number != 0 && !part.ModTime.IsZero() {
 			fi.AddObjectPart(part.Number, part.ETag, part.Size, part.ActualSize, part.ModTime, part.Index, part.Checksums)
 		}
 	}
@@ -1018,6 +950,106 @@ func (er erasureObjects) ListObjectParts(ctx context.Context, bucket, object, up
 	return result, nil
 }
 
+func readParts(ctx context.Context, disks []StorageAPI, bucket string, partMetaPaths []string, partNumbers []int, readQuorum int) ([]*ObjectPartInfo, error) {
+	g := errgroup.WithNErrs(len(disks))
+
+	objectPartInfos := make([][]*ObjectPartInfo, len(disks))
+	// Rename file on all underlying storage disks.
+	for index := range disks {
+		index := index
+		g.Go(func() (err error) {
+			if disks[index] == nil {
+				return errDiskNotFound
+			}
+			objectPartInfos[index], err = disks[index].ReadParts(ctx, bucket, partMetaPaths...)
+			return err
+		}, index)
+	}
+
+	if err := reduceReadQuorumErrs(ctx, g.Wait(), objectOpIgnoredErrs, readQuorum); err != nil {
+		return nil, err
+	}
+
+	partInfosInQuorum := make([]*ObjectPartInfo, len(partMetaPaths))
+	partMetaQuorumMap := make(map[string]int, len(partNumbers))
+	for pidx := range partMetaPaths {
+		var pinfos []*ObjectPartInfo
+		for idx := range disks {
+			if len(objectPartInfos[idx]) == 0 {
+				partMetaQuorumMap[partMetaPaths[pidx]]++
+				continue
+			}
+
+			pinfo := objectPartInfos[idx][pidx]
+			if pinfo == nil {
+				partMetaQuorumMap[partMetaPaths[pidx]]++
+				continue
+			}
+
+			if pinfo.ETag == "" {
+				partMetaQuorumMap[partMetaPaths[pidx]]++
+			} else {
+				pinfos = append(pinfos, pinfo)
+				partMetaQuorumMap[pinfo.ETag]++
+			}
+		}
+
+		var maxQuorum int
+		var maxETag string
+		var maxPartMeta string
+		for etag, quorum := range partMetaQuorumMap {
+			if maxQuorum < quorum {
+				maxQuorum = quorum
+				maxETag = etag
+				maxPartMeta = etag
+			}
+		}
+
+		var pinfo *ObjectPartInfo
+		for _, pinfo = range pinfos {
+			if pinfo != nil && maxETag != "" && pinfo.ETag == maxETag {
+				break
+			}
+			if maxPartMeta != "" && path.Base(maxPartMeta) == fmt.Sprintf("part.%d.meta", pinfo.Number) {
+				break
+			}
+		}
+
+		if pinfo != nil && pinfo.ETag != "" && partMetaQuorumMap[maxETag] >= readQuorum {
+			partInfosInQuorum[pidx] = pinfo
+			continue
+		}
+
+		if partMetaQuorumMap[maxPartMeta] == len(disks) {
+			if pinfo != nil && pinfo.Error != "" {
+				partInfosInQuorum[pidx] = &ObjectPartInfo{Error: pinfo.Error}
+			} else {
+				partInfosInQuorum[pidx] = &ObjectPartInfo{
+					Error: InvalidPart{
+						PartNumber: partNumbers[pidx],
+					}.Error(),
+				}
+			}
+		} else {
+			partInfosInQuorum[pidx] = &ObjectPartInfo{Error: errErasureReadQuorum.Error()}
+		}
+	}
+	return partInfosInQuorum, nil
+}
+
+func errStrToPartErr(errStr string) error {
+	if strings.Contains(errStr, "file not found") {
+		return InvalidPart{}
+	}
+	if strings.Contains(errStr, "Specified part could not be found") {
+		return InvalidPart{}
+	}
+	if strings.Contains(errStr, errErasureReadQuorum.Error()) {
+		return errErasureReadQuorum
+	}
+	return errors.New(errStr)
+}
+
 // CompleteMultipartUpload - completes an ongoing multipart
 // transaction after receiving all the parts indicated by the client.
 // Returns an md5sum calculated by concatenating all the individual
@@ -1026,7 +1058,7 @@ func (er erasureObjects) ListObjectParts(ctx context.Context, bucket, object, up
 // Implements S3 compatible Complete multipart API.
 func (er erasureObjects) CompleteMultipartUpload(ctx context.Context, bucket string, object string, uploadID string, parts []CompletePart, opts ObjectOptions) (oi ObjectInfo, err error) {
 	if !opts.NoAuditLog {
-		auditObjectErasureSet(ctx, object, &er)
+		auditObjectErasureSet(ctx, "CompleteMultipartUpload", object, &er)
 	}
 
 	if opts.CheckPrecondFn != nil {
@@ -1071,24 +1103,22 @@ func (er erasureObjects) CompleteMultipartUpload(ctx context.Context, bucket str
 	uploadIDPath := er.getUploadIDDir(bucket, object, uploadID)
 	onlineDisks := er.getDisks()
 	writeQuorum := fi.WriteQuorum(er.defaultWQuorum())
+	readQuorum := fi.ReadQuorum(er.defaultRQuorum())
 
 	// Read Part info for all parts
 	partPath := pathJoin(uploadIDPath, fi.DataDir) + "/"
-	req := ReadMultipleReq{
-		Bucket:       minioMetaMultipartBucket,
-		Prefix:       partPath,
-		MaxSize:      1 << 20, // Each part should realistically not be > 1MiB.
-		Files:        make([]string, 0, len(parts)),
-		AbortOn404:   true,
-		MetadataOnly: true,
+	partMetaPaths := make([]string, len(parts))
+	partNumbers := make([]int, len(parts))
+	for idx, part := range parts {
+		partMetaPaths[idx] = pathJoin(partPath, fmt.Sprintf("part.%d.meta", part.PartNumber))
+		partNumbers[idx] = part.PartNumber
 	}
-	for _, part := range parts {
-		req.Files = append(req.Files, fmt.Sprintf("part.%d.meta", part.PartNumber))
-	}
-	partInfoFiles, err := readMultipleFiles(ctx, onlineDisks, req, writeQuorum)
+
+	partInfoFiles, err := readParts(ctx, onlineDisks, minioMetaMultipartBucket, partMetaPaths, partNumbers, readQuorum)
 	if err != nil {
 		return oi, err
 	}
+
 	if len(partInfoFiles) != len(parts) {
 		// Should only happen through internal error
 		err := fmt.Errorf("unexpected part result count: %d, want %d", len(partInfoFiles), len(parts))
@@ -1150,35 +1180,22 @@ func (er erasureObjects) CompleteMultipartUpload(ctx context.Context, bucket str
 		opts.EncryptFn = metadataEncrypter(key)
 	}
 
-	for i, part := range partInfoFiles {
-		partID := parts[i].PartNumber
-		if part.Error != "" || !part.Exists {
-			return oi, InvalidPart{
-				PartNumber: partID,
-			}
-		}
-
-		var pfi FileInfo
-		_, err := pfi.UnmarshalMsg(part.Data)
-		if err != nil {
-			// Maybe crash or similar.
+	for idx, part := range partInfoFiles {
+		if part.Error != "" {
+			err = errStrToPartErr(part.Error)
 			bugLogIf(ctx, err)
-			return oi, InvalidPart{
-				PartNumber: partID,
-			}
+			return oi, err
 		}
 
-		partI := pfi.Parts[0]
-		partNumber := partI.Number
-		if partID != partNumber {
-			internalLogIf(ctx, fmt.Errorf("part.%d.meta has incorrect corresponding part number: expected %d, got %d", partID, partID, partI.Number))
+		if parts[idx].PartNumber != part.Number {
+			internalLogIf(ctx, fmt.Errorf("part.%d.meta has incorrect corresponding part number: expected %d, got %d", parts[idx].PartNumber, parts[idx].PartNumber, part.Number))
 			return oi, InvalidPart{
-				PartNumber: partID,
+				PartNumber: part.Number,
 			}
 		}
 
 		// Add the current part.
-		fi.AddObjectPart(partI.Number, partI.ETag, partI.Size, partI.ActualSize, partI.ModTime, partI.Index, partI.Checksums)
+		fi.AddObjectPart(part.Number, part.ETag, part.Size, part.ActualSize, part.ModTime, part.Index, part.Checksums)
 	}
 
 	// Calculate full object size.
@@ -1294,9 +1311,18 @@ 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)
@@ -1323,7 +1349,9 @@ func (er erasureObjects) CompleteMultipartUpload(ctx context.Context, bucket str
 
 	// Save the consolidated actual size.
 	if opts.ReplicationRequest {
-		fi.Metadata[ReservedMetadataPrefix+"actual-size"] = opts.UserDefined["X-Minio-Internal-Actual-Object-Size"]
+		if v := opts.UserDefined[ReservedMetadataPrefix+"Actual-Object-Size"]; v != "" {
+			fi.Metadata[ReservedMetadataPrefix+"actual-size"] = v
+		}
 	} else {
 		fi.Metadata[ReservedMetadataPrefix+"actual-size"] = strconv.FormatInt(objectActualSize, 10)
 	}
@@ -1345,10 +1373,10 @@ func (er erasureObjects) CompleteMultipartUpload(ctx context.Context, bucket str
 		}
 	}
 
+	paths := make([]string, 0, len(currentFI.Parts))
 	// Remove parts that weren't present in CompleteMultipartUpload request.
 	for _, curpart := range currentFI.Parts {
-		// Remove part.meta which is not needed anymore.
-		er.removePartMeta(bucket, object, uploadID, currentFI.DataDir, curpart.Number)
+		paths = append(paths, pathJoin(uploadIDPath, currentFI.DataDir, fmt.Sprintf("part.%d.meta", curpart.Number)))
 
 		if objectPartIndex(fi.Parts, curpart.Number) == -1 {
 			// Delete the missing part files. e.g,
@@ -1357,10 +1385,12 @@ func (er erasureObjects) CompleteMultipartUpload(ctx context.Context, bucket str
 			// Request 3: PutObjectPart 2
 			// Request 4: CompleteMultipartUpload --part 2
 			// N.B. 1st part is not present. This part should be removed from the storage.
-			er.removeObjectPart(bucket, object, uploadID, currentFI.DataDir, curpart.Number)
+			paths = append(paths, pathJoin(uploadIDPath, currentFI.DataDir, fmt.Sprintf("part.%d", curpart.Number)))
 		}
 	}
 
+	er.cleanupMultipartPath(ctx, paths...) // cleanup all part.N.meta, and skipped part.N's before final rename().
+
 	defer func() {
 		if err == nil {
 			er.deleteAll(context.Background(), minioMetaMultipartBucket, uploadIDPath)
@@ -1379,13 +1409,13 @@ func (er erasureObjects) CompleteMultipartUpload(ctx context.Context, bucket str
 	}
 
 	if !opts.Speedtest && len(versions) > 0 {
-		globalMRFState.addPartialOp(partialOperation{
-			bucket:    bucket,
-			object:    object,
-			queued:    time.Now(),
-			versions:  versions,
-			setIndex:  er.setIndex,
-			poolIndex: er.poolIndex,
+		globalMRFState.addPartialOp(PartialOperation{
+			Bucket:    bucket,
+			Object:    object,
+			Queued:    time.Now(),
+			Versions:  versions,
+			SetIndex:  er.setIndex,
+			PoolIndex: er.poolIndex,
 		})
 	}
 
@@ -1425,7 +1455,7 @@ func (er erasureObjects) CompleteMultipartUpload(ctx context.Context, bucket str
 // operation.
 func (er erasureObjects) AbortMultipartUpload(ctx context.Context, bucket, object, uploadID string, opts ObjectOptions) (err error) {
 	if !opts.NoAuditLog {
-		auditObjectErasureSet(ctx, object, &er)
+		auditObjectErasureSet(ctx, "AbortMultipartUpload", object, &er)
 	}
 
 	lk := er.NewNSLock(bucket, pathJoin(object, uploadID))

cmd/erasure-object.go

@@ -20,6 +20,7 @@ package cmd
 import (
 	"bytes"
 	"context"
+	"encoding/base64"
 	"errors"
 	"fmt"
 	"io"
@@ -69,7 +70,7 @@ func countOnlineDisks(onlineDisks []StorageAPI) (online int) {
 // update metadata.
 func (er erasureObjects) CopyObject(ctx context.Context, srcBucket, srcObject, dstBucket, dstObject string, srcInfo ObjectInfo, srcOpts, dstOpts ObjectOptions) (oi ObjectInfo, err error) {
 	if !dstOpts.NoAuditLog {
-		auditObjectErasureSet(ctx, dstObject, &er)
+		auditObjectErasureSet(ctx, "CopyObject", dstObject, &er)
 	}
 
 	// This call shouldn't be used for anything other than metadata updates or adding self referential versions.
@@ -96,7 +97,7 @@ func (er erasureObjects) CopyObject(ctx context.Context, srcBucket, srcObject, d
 	if srcOpts.VersionID != "" {
 		metaArr, errs = readAllFileInfo(ctx, storageDisks, "", srcBucket, srcObject, srcOpts.VersionID, true, false)
 	} else {
-		metaArr, errs = readAllXL(ctx, storageDisks, srcBucket, srcObject, true, false, true)
+		metaArr, errs = readAllXL(ctx, storageDisks, srcBucket, srcObject, true, false)
 	}
 
 	readQuorum, writeQuorum, err := objectQuorumFromMeta(ctx, metaArr, errs, er.defaultParityCount)
@@ -198,7 +199,7 @@ func (er erasureObjects) CopyObject(ctx context.Context, srcBucket, srcObject, d
 // Read(Closer). When err != nil, the returned reader is always nil.
 func (er erasureObjects) GetObjectNInfo(ctx context.Context, bucket, object string, rs *HTTPRangeSpec, h http.Header, opts ObjectOptions) (gr *GetObjectReader, err error) {
 	if !opts.NoAuditLog {
-		auditObjectErasureSet(ctx, object, &er)
+		auditObjectErasureSet(ctx, "GetObject", object, &er)
 	}
 
 	var unlockOnDefer bool
@@ -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()
@@ -394,24 +395,16 @@ func (er erasureObjects) getObjectWithFileInfo(ctx context.Context, bucket, obje
 			// that we have some parts or data blocks missing or corrupted
 			// - attempt a heal to successfully heal them for future calls.
 			if written == partLength {
-				var scan madmin.HealScanMode
-				switch {
-				case errors.Is(err, errFileNotFound):
-					scan = madmin.HealNormalScan
-				case errors.Is(err, errFileCorrupt):
-					scan = madmin.HealDeepScan
-				}
-				switch scan {
-				case madmin.HealNormalScan, madmin.HealDeepScan:
+				if errors.Is(err, errFileNotFound) || errors.Is(err, errFileCorrupt) {
 					healOnce.Do(func() {
-						globalMRFState.addPartialOp(partialOperation{
-							bucket:    bucket,
-							object:    object,
-							versionID: fi.VersionID,
-							queued:    time.Now(),
-							setIndex:  er.setIndex,
-							poolIndex: er.poolIndex,
-							scanMode:  scan,
+						globalMRFState.addPartialOp(PartialOperation{
+							Bucket:     bucket,
+							Object:     object,
+							VersionID:  fi.VersionID,
+							Queued:     time.Now(),
+							SetIndex:   er.setIndex,
+							PoolIndex:  er.poolIndex,
+							BitrotScan: errors.Is(err, errFileCorrupt),
 						})
 					})
 					// Healing is triggered and we have written
@@ -438,7 +431,7 @@ func (er erasureObjects) getObjectWithFileInfo(ctx context.Context, bucket, obje
 // GetObjectInfo - reads object metadata and replies back ObjectInfo.
 func (er erasureObjects) GetObjectInfo(ctx context.Context, bucket, object string, opts ObjectOptions) (info ObjectInfo, err error) {
 	if !opts.NoAuditLog {
-		auditObjectErasureSet(ctx, object, &er)
+		auditObjectErasureSet(ctx, "GetObjectInfo", object, &er)
 	}
 
 	if !opts.NoLock {
@@ -455,7 +448,7 @@ func (er erasureObjects) GetObjectInfo(ctx context.Context, bucket, object strin
 	return er.getObjectInfo(ctx, bucket, object, opts)
 }
 
-func auditDanglingObjectDeletion(ctx context.Context, bucket, object, versionID string, tags map[string]interface{}) {
+func auditDanglingObjectDeletion(ctx context.Context, bucket, object, versionID string, tags map[string]string) {
 	if len(logger.AuditTargets()) == 0 {
 		return
 	}
@@ -471,50 +464,63 @@ func auditDanglingObjectDeletion(ctx context.Context, bucket, object, versionID
 	auditLogInternal(ctx, opts)
 }
 
-func joinErrs(errs []error) []string {
-	s := make([]string, len(errs))
+func joinErrs(errs []error) string {
+	var s string
 	for i := range s {
+		if s != "" {
+			s += ","
+		}
 		if errs[i] == nil {
-			s[i] = "<nil>"
+			s += "<nil>"
 		} else {
-			s[i] = errs[i].Error()
+			s += errs[i].Error()
 		}
 	}
 	return s
 }
 
-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.
 		return FileInfo{}, errErasureReadQuorum
 	}
-	tags := make(map[string]interface{}, 4)
-	tags["set"] = er.setIndex
-	tags["pool"] = er.poolIndex
+	tags := make(map[string]string, 16)
+	tags["set"] = strconv.Itoa(er.setIndex)
+	tags["pool"] = strconv.Itoa(er.poolIndex)
 	tags["merrs"] = joinErrs(errs)
-	tags["derrs"] = joinErrs(dataErrs)
+	tags["derrs"] = fmt.Sprintf("%v", dataErrsByPart)
 	if m.IsValid() {
-		tags["size"] = m.Size
-		tags["mtime"] = m.ModTime.Format(http.TimeFormat)
-		tags["data"] = m.Erasure.DataBlocks
-		tags["parity"] = m.Erasure.ParityBlocks
+		tags["sz"] = strconv.FormatInt(m.Size, 10)
+		tags["mt"] = m.ModTime.Format(iso8601Format)
+		tags["d:p"] = fmt.Sprintf("%d:%d", m.Erasure.DataBlocks, m.Erasure.ParityBlocks)
 	} else {
-		tags["invalid-meta"] = true
-		tags["data"] = er.setDriveCount - er.defaultParityCount
-		tags["parity"] = er.defaultParityCount
+		tags["invalid"] = "1"
+		tags["d:p"] = fmt.Sprintf("%d:%d", er.setDriveCount-er.defaultParityCount, er.defaultParityCount)
 	}
 
 	// count the number of offline disks
 	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++
 		}
 	}
 	if offline > 0 {
-		tags["offline"] = offline
+		tags["offline"] = strconv.Itoa(offline)
 	}
 
 	_, file, line, cok := runtime.Caller(1)
@@ -543,55 +549,24 @@ func (er erasureObjects) deleteIfDangling(ctx context.Context, bucket, object st
 		}, index)
 	}
 
-	rmDisks := make(map[string]string, len(disks))
 	for index, err := range g.Wait() {
-		var errStr, diskName string
+		var errStr string
 		if err != nil {
 			errStr = err.Error()
 		} else {
 			errStr = "<nil>"
 		}
-		if disks[index] != nil {
-			diskName = disks[index].String()
-		} else {
-			diskName = fmt.Sprintf("disk-%d", index)
-		}
-		rmDisks[diskName] = errStr
+		tags[fmt.Sprintf("ddisk-%d", index)] = errStr
 	}
-	tags["cleanupResult"] = rmDisks
+
 	return m, nil
 }
 
-func fileInfoFromRaw(ri RawFileInfo, bucket, object string, readData, inclFreeVers, allParts bool) (FileInfo, error) {
-	var xl xlMetaV2
-	if err := xl.LoadOrConvert(ri.Buf); err != nil {
-		return FileInfo{}, errFileCorrupt
-	}
-
-	fi, err := xl.ToFileInfo(bucket, object, "", inclFreeVers, allParts)
-	if err != nil {
-		return FileInfo{}, err
-	}
-
-	if !fi.IsValid() {
-		return FileInfo{}, errFileCorrupt
-	}
-
-	versionID := fi.VersionID
-	if versionID == "" {
-		versionID = nullVersionID
-	}
-
-	fileInfo, err := xl.ToFileInfo(bucket, object, versionID, inclFreeVers, allParts)
-	if err != nil {
-		return FileInfo{}, err
-	}
-
-	if readData {
-		fileInfo.Data = xl.data.find(versionID)
-	}
-
-	return fileInfo, nil
+func fileInfoFromRaw(ri RawFileInfo, bucket, object string, readData, inclFreeVers bool) (FileInfo, error) {
+	return getFileInfo(ri.Buf, bucket, object, "", fileInfoOpts{
+		Data:             readData,
+		InclFreeVersions: inclFreeVers,
+	})
 }
 
 func readAllRawFileInfo(ctx context.Context, disks []StorageAPI, bucket, object string, readData bool) ([]RawFileInfo, []error) {
@@ -615,7 +590,7 @@ func readAllRawFileInfo(ctx context.Context, disks []StorageAPI, bucket, object
 	return rawFileInfos, g.Wait()
 }
 
-func pickLatestQuorumFilesInfo(ctx context.Context, rawFileInfos []RawFileInfo, errs []error, bucket, object string, readData, inclFreeVers, allParts bool) ([]FileInfo, []error) {
+func pickLatestQuorumFilesInfo(ctx context.Context, rawFileInfos []RawFileInfo, errs []error, bucket, object string, readData, inclFreeVers bool) ([]FileInfo, []error) {
 	metadataArray := make([]*xlMetaV2, len(rawFileInfos))
 	metaFileInfos := make([]FileInfo, len(rawFileInfos))
 	metadataShallowVersions := make([][]xlMetaV2ShallowVersion, len(rawFileInfos))
@@ -652,7 +627,7 @@ func pickLatestQuorumFilesInfo(ctx context.Context, rawFileInfos []RawFileInfo,
 
 	readQuorum := (len(rawFileInfos) + 1) / 2
 	meta := &xlMetaV2{versions: mergeXLV2Versions(readQuorum, false, 1, metadataShallowVersions...)}
-	lfi, err := meta.ToFileInfo(bucket, object, "", inclFreeVers, allParts)
+	lfi, err := meta.ToFileInfo(bucket, object, "", inclFreeVers, true)
 	if err != nil {
 		for i := range errs {
 			if errs[i] == nil {
@@ -681,7 +656,7 @@ func pickLatestQuorumFilesInfo(ctx context.Context, rawFileInfos []RawFileInfo,
 		}
 
 		// make sure to preserve this for diskmtime based healing bugfix.
-		metaFileInfos[index], errs[index] = metadataArray[index].ToFileInfo(bucket, object, versionID, inclFreeVers, allParts)
+		metaFileInfos[index], errs[index] = metadataArray[index].ToFileInfo(bucket, object, versionID, inclFreeVers, true)
 		if errs[index] != nil {
 			continue
 		}
@@ -726,9 +701,9 @@ func shouldCheckForDangling(err error, errs []error, bucket string) bool {
 	return false
 }
 
-func readAllXL(ctx context.Context, disks []StorageAPI, bucket, object string, readData, inclFreeVers, allParts bool) ([]FileInfo, []error) {
+func readAllXL(ctx context.Context, disks []StorageAPI, bucket, object string, readData, inclFreeVers bool) ([]FileInfo, []error) {
 	rawFileInfos, errs := readAllRawFileInfo(ctx, disks, bucket, object, readData)
-	return pickLatestQuorumFilesInfo(ctx, rawFileInfos, errs, bucket, object, readData, inclFreeVers, allParts)
+	return pickLatestQuorumFilesInfo(ctx, rawFileInfos, errs, bucket, object, readData, inclFreeVers)
 }
 
 func (er erasureObjects) getObjectFileInfo(ctx context.Context, bucket, object string, opts ObjectOptions, readData bool) (FileInfo, []FileInfo, []StorageAPI, error) {
@@ -743,8 +718,9 @@ func (er erasureObjects) getObjectFileInfo(ctx context.Context, bucket, object s
 	disks := er.getDisks()
 
 	ropts := ReadOptions{
-		ReadData: readData,
-		Healing:  false,
+		ReadData:         readData,
+		InclFreeVersions: opts.InclFreeVersions,
+		Healing:          false,
 	}
 
 	mrfCheck := make(chan FileInfo)
@@ -784,7 +760,7 @@ func (er erasureObjects) getObjectFileInfo(ctx context.Context, bucket, object s
 					// Read the latest version
 					rfi, err = disk.ReadXL(ctx, bucket, object, readData)
 					if err == nil {
-						fi, err = fileInfoFromRaw(rfi, bucket, object, readData, opts.InclFreeVersions, true)
+						fi, err = fileInfoFromRaw(rfi, bucket, object, readData, opts.InclFreeVersions)
 					}
 				}
 
@@ -830,13 +806,13 @@ func (er erasureObjects) getObjectFileInfo(ctx context.Context, bucket, object s
 		// additionally do not heal delete markers inline, let them be
 		// healed upon regular heal process.
 		if missingBlocks > 0 && missingBlocks < fi.Erasure.DataBlocks {
-			globalMRFState.addPartialOp(partialOperation{
-				bucket:    fi.Volume,
-				object:    fi.Name,
-				versionID: fi.VersionID,
-				queued:    time.Now(),
-				setIndex:  er.setIndex,
-				poolIndex: er.poolIndex,
+			globalMRFState.addPartialOp(PartialOperation{
+				Bucket:    fi.Volume,
+				Object:    fi.Name,
+				VersionID: fi.VersionID,
+				Queued:    time.Now(),
+				SetIndex:  er.setIndex,
+				PoolIndex: er.poolIndex,
 			})
 		}
 
@@ -918,14 +894,16 @@ func (er erasureObjects) getObjectFileInfo(ctx context.Context, bucket, object s
 		}
 
 		rw.Lock()
+		// when its a versioned bucket and empty versionID - at totalResp == setDriveCount
+		// we must use rawFileInfo to resolve versions to figure out the latest version.
 		if opts.VersionID == "" && totalResp == er.setDriveCount {
 			fi, onlineMeta, onlineDisks, modTime, etag, err = calcQuorum(pickLatestQuorumFilesInfo(ctx,
-				rawArr, errs, bucket, object, readData, opts.InclFreeVersions, true))
+				rawArr, errs, bucket, object, readData, opts.InclFreeVersions))
 		} else {
 			fi, onlineMeta, onlineDisks, modTime, etag, err = calcQuorum(metaArr, errs)
 		}
 		rw.Unlock()
-		if err == nil && fi.InlineData() {
+		if err == nil && (fi.InlineData() || len(fi.Data) > 0) {
 			break
 		}
 	}
@@ -1260,7 +1238,7 @@ func (er erasureObjects) PutObject(ctx context.Context, bucket string, object st
 // putObject wrapper for erasureObjects PutObject
 func (er erasureObjects) putObject(ctx context.Context, bucket string, object string, r *PutObjReader, opts ObjectOptions) (objInfo ObjectInfo, err error) {
 	if !opts.NoAuditLog {
-		auditObjectErasureSet(ctx, object, &er)
+		auditObjectErasureSet(ctx, "PutObject", object, &er)
 	}
 
 	data := r.Reader
@@ -1355,6 +1333,12 @@ 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 +1387,7 @@ func (er erasureObjects) putObject(ctx context.Context, bucket string, object st
 	writers := make([]io.Writer, len(onlineDisks))
 	var inlineBuffers []*bytes.Buffer
 	if shardFileSize >= 0 {
-		if !opts.Versioned && shardFileSize < inlineBlock {
+		if !opts.Versioned && shardFileSize <= inlineBlock {
 			inlineBuffers = make([]*bytes.Buffer, len(onlineDisks))
 		} else if shardFileSize < inlineBlock/8 {
 			inlineBuffers = make([]*bytes.Buffer, len(onlineDisks))
@@ -1411,7 +1395,7 @@ func (er erasureObjects) putObject(ctx context.Context, bucket string, object st
 	} else {
 		// If compressed, use actual size to determine.
 		if sz := erasure.ShardFileSize(data.ActualSize()); sz > 0 {
-			if !opts.Versioned && sz < inlineBlock {
+			if !opts.Versioned && sz <= inlineBlock {
 				inlineBuffers = make([]*bytes.Buffer, len(onlineDisks))
 			} else if sz < inlineBlock/8 {
 				inlineBuffers = make([]*bytes.Buffer, len(onlineDisks))
@@ -1580,13 +1564,13 @@ func (er erasureObjects) putObject(ctx context.Context, bucket string, object st
 				break
 			}
 		} else {
-			globalMRFState.addPartialOp(partialOperation{
-				bucket:    bucket,
-				object:    object,
-				queued:    time.Now(),
-				versions:  versions,
-				setIndex:  er.setIndex,
-				poolIndex: er.poolIndex,
+			globalMRFState.addPartialOp(PartialOperation{
+				Bucket:    bucket,
+				Object:    object,
+				Queued:    time.Now(),
+				Versions:  versions,
+				SetIndex:  er.setIndex,
+				PoolIndex: er.poolIndex,
 			})
 		}
 	}
@@ -1629,7 +1613,7 @@ func (er erasureObjects) deleteObjectVersion(ctx context.Context, bucket, object
 func (er erasureObjects) DeleteObjects(ctx context.Context, bucket string, objects []ObjectToDelete, opts ObjectOptions) ([]DeletedObject, []error) {
 	if !opts.NoAuditLog {
 		for _, obj := range objects {
-			auditObjectErasureSet(ctx, obj.ObjectV.ObjectName, &er)
+			auditObjectErasureSet(ctx, "DeleteObjects", obj.ObjectV.ObjectName, &er)
 		}
 	}
 
@@ -1849,7 +1833,7 @@ func (er erasureObjects) deletePrefix(ctx context.Context, bucket, prefix string
 // response to the client request.
 func (er erasureObjects) DeleteObject(ctx context.Context, bucket, object string, opts ObjectOptions) (objInfo ObjectInfo, err error) {
 	if !opts.NoAuditLog {
-		auditObjectErasureSet(ctx, object, &er)
+		auditObjectErasureSet(ctx, "DeleteObject", object, &er)
 	}
 
 	var lc *lifecycle.Lifecycle
@@ -2115,11 +2099,11 @@ func (er erasureObjects) DeleteObject(ctx context.Context, bucket, object string
 // Send the successful but partial upload/delete, however ignore
 // if the channel is blocked by other items.
 func (er erasureObjects) addPartial(bucket, object, versionID string) {
-	globalMRFState.addPartialOp(partialOperation{
-		bucket:    bucket,
-		object:    object,
-		versionID: versionID,
-		queued:    time.Now(),
+	globalMRFState.addPartialOp(PartialOperation{
+		Bucket:    bucket,
+		Object:    object,
+		VersionID: versionID,
+		Queued:    time.Now(),
 	})
 }
 
@@ -2144,7 +2128,7 @@ func (er erasureObjects) PutObjectMetadata(ctx context.Context, bucket, object s
 	if opts.VersionID != "" {
 		metaArr, errs = readAllFileInfo(ctx, disks, "", bucket, object, opts.VersionID, false, false)
 	} else {
-		metaArr, errs = readAllXL(ctx, disks, bucket, object, false, false, true)
+		metaArr, errs = readAllXL(ctx, disks, bucket, object, false, false)
 	}
 
 	readQuorum, _, err := objectQuorumFromMeta(ctx, metaArr, errs, er.defaultParityCount)
@@ -2221,7 +2205,7 @@ func (er erasureObjects) PutObjectTags(ctx context.Context, bucket, object strin
 	if opts.VersionID != "" {
 		metaArr, errs = readAllFileInfo(ctx, disks, "", bucket, object, opts.VersionID, false, false)
 	} else {
-		metaArr, errs = readAllXL(ctx, disks, bucket, object, false, false, true)
+		metaArr, errs = readAllXL(ctx, disks, bucket, object, false, false)
 	}
 
 	readQuorum, _, err := objectQuorumFromMeta(ctx, metaArr, errs, er.defaultParityCount)

cmd/erasure-server-pool-decom.go

@@ -358,7 +358,7 @@ func (p *poolMeta) validate(pools []*erasureSets) (bool, error) {
 			update = true
 		}
 		if ok && pi.completed {
-			return false, fmt.Errorf("pool(%s) = %s is decommissioned, please remove from server command line", humanize.Ordinal(pi.position+1), k)
+			logger.LogIf(GlobalContext, "decommission", fmt.Errorf("pool(%s) = %s is decommissioned, please remove from server command line", humanize.Ordinal(pi.position+1), k))
 		}
 	}
 
@@ -608,7 +608,7 @@ func (z *erasureServerPools) IsDecommissionRunning() bool {
 	return false
 }
 
-func (z *erasureServerPools) decommissionObject(ctx context.Context, bucket string, gr *GetObjectReader) (err error) {
+func (z *erasureServerPools) decommissionObject(ctx context.Context, idx int, bucket string, gr *GetObjectReader) (err error) {
 	objInfo := gr.ObjInfo
 
 	defer func() {
@@ -623,9 +623,11 @@ func (z *erasureServerPools) decommissionObject(ctx context.Context, bucket stri
 
 	if objInfo.isMultipart() {
 		res, err := z.NewMultipartUpload(ctx, bucket, objInfo.Name, ObjectOptions{
-			VersionID:   objInfo.VersionID,
-			UserDefined: objInfo.UserDefined,
-			NoAuditLog:  true,
+			VersionID:    objInfo.VersionID,
+			UserDefined:  objInfo.UserDefined,
+			NoAuditLog:   true,
+			SrcPoolIdx:   idx,
+			DataMovement: true,
 		})
 		if err != nil {
 			return fmt.Errorf("decommissionObject: NewMultipartUpload() %w", err)
@@ -660,6 +662,7 @@ func (z *erasureServerPools) decommissionObject(ctx context.Context, bucket stri
 			}
 		}
 		_, err = z.CompleteMultipartUpload(ctx, bucket, objInfo.Name, res.UploadID, parts, ObjectOptions{
+			SrcPoolIdx:   idx,
 			DataMovement: true,
 			MTime:        objInfo.ModTime,
 			NoAuditLog:   true,
@@ -681,6 +684,7 @@ func (z *erasureServerPools) decommissionObject(ctx context.Context, bucket stri
 		NewPutObjReader(hr),
 		ObjectOptions{
 			DataMovement: true,
+			SrcPoolIdx:   idx,
 			VersionID:    objInfo.VersionID,
 			MTime:        objInfo.ModTime,
 			UserDefined:  objInfo.UserDefined,
@@ -836,7 +840,7 @@ func (z *erasureServerPools) decommissionPool(ctx context.Context, idx int, pool
 				if filterLifecycle(bi.Name, version.Name, version) {
 					expired++
 					decommissioned++
-					stopFn(errors.New("ILM expired object/version will be skipped"))
+					stopFn(version.Size, errors.New("ILM expired object/version will be skipped"))
 					continue
 				}
 
@@ -846,7 +850,7 @@ func (z *erasureServerPools) decommissionPool(ctx context.Context, idx int, pool
 				remainingVersions := len(fivs.Versions) - expired
 				if version.Deleted && remainingVersions == 1 {
 					decommissioned++
-					stopFn(errors.New("DELETE marked object with no other non-current versions will be skipped"))
+					stopFn(version.Size, errors.New("DELETE marked object with no other non-current versions will be skipped"))
 					continue
 				}
 
@@ -855,6 +859,7 @@ func (z *erasureServerPools) decommissionPool(ctx context.Context, idx int, pool
 					versionID = nullVersionID
 				}
 
+				var failure, ignore bool
 				if version.Deleted {
 					_, err := z.DeleteObject(ctx,
 						bi.Name,
@@ -867,17 +872,22 @@ func (z *erasureServerPools) decommissionPool(ctx context.Context, idx int, pool
 							VersionID:          versionID,
 							MTime:              version.ModTime,
 							DeleteReplication:  version.ReplicationState,
+							SrcPoolIdx:         idx,
+							DataMovement:       true,
 							DeleteMarker:       true, // make sure we create a delete marker
 							SkipDecommissioned: true, // make sure we skip the decommissioned pool
 							NoAuditLog:         true,
 						})
-					var failure bool
 					if err != nil {
-						if isErrObjectNotFound(err) || isErrVersionNotFound(err) {
-							err = nil
+						// This can happen when rebalance stop races with ongoing rebalance workers.
+						// These rebalance failures can be ignored.
+						if isErrObjectNotFound(err) || isErrVersionNotFound(err) || isDataMovementOverWriteErr(err) {
+							ignore = true
+							stopFn(0, nil)
+							continue
 						}
 					}
-					stopFn(err)
+					stopFn(version.Size, err)
 					if err != nil {
 						decomLogIf(ctx, err)
 						failure = true
@@ -893,22 +903,26 @@ func (z *erasureServerPools) decommissionPool(ctx context.Context, idx int, pool
 					continue
 				}
 
-				var failure, ignore bool
 				// gr.Close() is ensured by decommissionObject().
 				for try := 0; try < 3; try++ {
 					if version.IsRemote() {
 						if err := z.DecomTieredObject(ctx, bi.Name, version.Name, version, ObjectOptions{
-							VersionID:   versionID,
-							MTime:       version.ModTime,
-							UserDefined: version.Metadata,
+							VersionID:    versionID,
+							MTime:        version.ModTime,
+							UserDefined:  version.Metadata,
+							SrcPoolIdx:   idx,
+							DataMovement: true,
 						}); err != nil {
-							stopFn(err)
-							failure = true
-							decomLogIf(ctx, err)
-							continue
+							if isErrObjectNotFound(err) || isErrVersionNotFound(err) || isDataMovementOverWriteErr(err) {
+								ignore = true
+								stopFn(0, nil)
+							}
+						}
+						if !ignore {
+							stopFn(version.Size, err)
+							failure = err != nil
+							decomLogIf(ctx, err)
 						}
-						stopFn(nil)
-						failure = false
 						break
 					}
 					gr, err := set.GetObjectNInfo(ctx,
@@ -925,14 +939,14 @@ func (z *erasureServerPools) decommissionPool(ctx context.Context, idx int, pool
 					if isErrObjectNotFound(err) || isErrVersionNotFound(err) {
 						// object deleted by the application, nothing to do here we move on.
 						ignore = true
-						stopFn(nil)
+						stopFn(0, nil)
 						break
 					}
 					if err != nil && !ignore {
 						// if usage-cache.bin is not readable log and ignore it.
 						if bi.Name == minioMetaBucket && strings.Contains(version.Name, dataUsageCacheName) {
 							ignore = true
-							stopFn(err)
+							stopFn(version.Size, err)
 							decomLogIf(ctx, err)
 							break
 						}
@@ -940,16 +954,21 @@ func (z *erasureServerPools) decommissionPool(ctx context.Context, idx int, pool
 					if err != nil {
 						failure = true
 						decomLogIf(ctx, err)
-						stopFn(err)
+						stopFn(version.Size, err)
 						continue
 					}
-					if err = z.decommissionObject(ctx, bi.Name, gr); err != nil {
-						stopFn(err)
+					if err = z.decommissionObject(ctx, idx, bi.Name, gr); err != nil {
+						if isErrObjectNotFound(err) || isErrVersionNotFound(err) || isDataMovementOverWriteErr(err) {
+							ignore = true
+							stopFn(0, nil)
+							break
+						}
+						stopFn(version.Size, err)
 						failure = true
 						decomLogIf(ctx, err)
 						continue
 					}
-					stopFn(nil)
+					stopFn(version.Size, nil)
 					failure = false
 					break
 				}
@@ -977,7 +996,7 @@ func (z *erasureServerPools) decommissionPool(ctx context.Context, idx int, pool
 						NoAuditLog:         true,
 					},
 				)
-				stopFn(err)
+				stopFn(0, err)
 				auditLogDecom(ctx, "DecomDeleteObject", bi.Name, entry.name, "", err)
 				if err != nil {
 					decomLogIf(ctx, err)
@@ -1038,7 +1057,7 @@ const (
 	decomMetricDecommissionRemoveObject
 )
 
-func decomTrace(d decomMetric, poolIdx int, startTime time.Time, duration time.Duration, path string, err error) madmin.TraceInfo {
+func decomTrace(d decomMetric, poolIdx int, startTime time.Time, duration time.Duration, path string, err error, sz int64) madmin.TraceInfo {
 	var errStr string
 	if err != nil {
 		errStr = err.Error()
@@ -1051,15 +1070,16 @@ func decomTrace(d decomMetric, poolIdx int, startTime time.Time, duration time.D
 		Duration:  duration,
 		Path:      path,
 		Error:     errStr,
+		Bytes:     sz,
 	}
 }
 
-func (m *decomMetrics) log(d decomMetric, poolIdx int, paths ...string) func(err error) {
+func (m *decomMetrics) log(d decomMetric, poolIdx int, paths ...string) func(z int64, err error) {
 	startTime := time.Now()
-	return func(err error) {
+	return func(sz int64, err error) {
 		duration := time.Since(startTime)
 		if globalTrace.NumSubscribers(madmin.TraceDecommission) > 0 {
-			globalTrace.Publish(decomTrace(d, poolIdx, startTime, duration, strings.Join(paths, " "), err))
+			globalTrace.Publish(decomTrace(d, poolIdx, startTime, duration, strings.Join(paths, " "), err, sz))
 		}
 	}
 }
@@ -1092,10 +1112,10 @@ func (z *erasureServerPools) decommissionInBackground(ctx context.Context, idx i
 		}
 		stopFn := globalDecommissionMetrics.log(decomMetricDecommissionBucket, idx, bucket.Name)
 		if err := z.decommissionPool(ctx, idx, pool, bucket); err != nil {
-			stopFn(err)
+			stopFn(0, err)
 			return err
 		}
-		stopFn(nil)
+		stopFn(0, nil)
 
 		z.poolMetaMutex.Lock()
 		if z.poolMeta.BucketDone(idx, bucket) {
@@ -1326,11 +1346,7 @@ 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
@@ -427,7 +431,7 @@ func (z *erasureServerPools) rebalanceBuckets(ctx context.Context, poolIdx int)
 
 			stopFn := globalRebalanceMetrics.log(rebalanceMetricSaveMetadata, poolIdx, traceMsg)
 			err := z.saveRebalanceStats(GlobalContext, poolIdx, rebalSaveStats)
-			stopFn(err)
+			stopFn(0, err)
 			rebalanceLogIf(GlobalContext, err)
 
 			if quit {
@@ -456,7 +460,7 @@ func (z *erasureServerPools) rebalanceBuckets(ctx context.Context, poolIdx int)
 
 		stopFn := globalRebalanceMetrics.log(rebalanceMetricRebalanceBucket, poolIdx, bucket)
 		if err = z.rebalanceBucket(ctx, bucket, poolIdx); err != nil {
-			stopFn(err)
+			stopFn(0, err)
 			if errors.Is(err, errServerNotInitialized) || errors.Is(err, errBucketMetadataNotInitialized) {
 				continue
 			}
@@ -464,7 +468,7 @@ func (z *erasureServerPools) rebalanceBuckets(ctx context.Context, poolIdx int)
 			doneCh <- err
 			return
 		}
-		stopFn(nil)
+		stopFn(0, nil)
 		z.bucketRebalanceDone(bucket, poolIdx)
 	}
 
@@ -622,14 +626,18 @@ func (z *erasureServerPools) rebalanceBucket(ctx context.Context, bucket string,
 
 			var rebalanced, expired int
 			for _, version := range fivs.Versions {
+				stopFn := globalRebalanceMetrics.log(rebalanceMetricRebalanceObject, poolIdx, bucket, version.Name, version.VersionID)
+
 				// Skip transitioned objects for now. TBD
 				if version.IsRemote() {
+					stopFn(version.Size, errors.New("ILM Tiered version will be skipped for now"))
 					continue
 				}
 
 				// Apply lifecycle rules on the objects that are expired.
 				if filterLifecycle(bucket, version.Name, version) {
 					expired++
+					stopFn(version.Size, errors.New("ILM expired object/version will be skipped"))
 					continue
 				}
 
@@ -639,6 +647,7 @@ func (z *erasureServerPools) rebalanceBucket(ctx context.Context, bucket string,
 				remainingVersions := len(fivs.Versions) - expired
 				if version.Deleted && remainingVersions == 1 {
 					rebalanced++
+					stopFn(version.Size, errors.New("DELETE marked object with no other non-current versions will be skipped"))
 					continue
 				}
 
@@ -647,6 +656,7 @@ func (z *erasureServerPools) rebalanceBucket(ctx context.Context, bucket string,
 					versionID = nullVersionID
 				}
 
+				var failure, ignore bool
 				if version.Deleted {
 					_, err := z.DeleteObject(ctx,
 						bucket,
@@ -656,16 +666,26 @@ func (z *erasureServerPools) rebalanceBucket(ctx context.Context, bucket string,
 							VersionID:         versionID,
 							MTime:             version.ModTime,
 							DeleteReplication: version.ReplicationState,
+							SrcPoolIdx:        poolIdx,
+							DataMovement:      true,
 							DeleteMarker:      true, // make sure we create a delete marker
 							SkipRebalancing:   true, // make sure we skip the decommissioned pool
 							NoAuditLog:        true,
 						})
-					var failure bool
-					if err != nil && !isErrObjectNotFound(err) && !isErrVersionNotFound(err) {
-						rebalanceLogIf(ctx, err)
-						failure = true
+					// This can happen when rebalance stop races with ongoing rebalance workers.
+					// These rebalance failures can be ignored.
+					if err != nil {
+						// This can happen when rebalance stop races with ongoing rebalance workers.
+						// These rebalance failures can be ignored.
+						if isErrObjectNotFound(err) || isErrVersionNotFound(err) || isDataMovementOverWriteErr(err) {
+							ignore = true
+							stopFn(0, nil)
+							continue
+						}
 					}
-
+					stopFn(version.Size, err)
+					rebalanceLogIf(ctx, err)
+					failure = err != nil
 					if !failure {
 						z.updatePoolStats(poolIdx, bucket, version)
 						rebalanced++
@@ -674,10 +694,8 @@ func (z *erasureServerPools) rebalanceBucket(ctx context.Context, bucket string,
 					continue
 				}
 
-				var failure, ignore bool
 				for try := 0; try < 3; try++ {
 					// GetObjectReader.Close is called by rebalanceObject
-					stopFn := globalRebalanceMetrics.log(rebalanceMetricRebalanceObject, poolIdx, bucket, version.Name, version.VersionID)
 					gr, err := set.GetObjectNInfo(ctx,
 						bucket,
 						encodeDirObject(version.Name),
@@ -692,24 +710,31 @@ func (z *erasureServerPools) rebalanceBucket(ctx context.Context, bucket string,
 					if isErrObjectNotFound(err) || isErrVersionNotFound(err) {
 						// object deleted by the application, nothing to do here we move on.
 						ignore = true
-						stopFn(nil)
+						stopFn(0, nil)
 						break
 					}
 					if err != nil {
 						failure = true
 						rebalanceLogIf(ctx, err)
-						stopFn(err)
+						stopFn(0, err)
 						continue
 					}
 
-					if err = z.rebalanceObject(ctx, bucket, gr); err != nil {
+					if err = z.rebalanceObject(ctx, poolIdx, bucket, gr); err != nil {
+						// This can happen when rebalance stop races with ongoing rebalance workers.
+						// These rebalance failures can be ignored.
+						if isErrObjectNotFound(err) || isErrVersionNotFound(err) || isDataMovementOverWriteErr(err) {
+							ignore = true
+							stopFn(0, nil)
+							break
+						}
 						failure = true
 						rebalanceLogIf(ctx, err)
-						stopFn(err)
+						stopFn(version.Size, err)
 						continue
 					}
 
-					stopFn(nil)
+					stopFn(version.Size, nil)
 					failure = false
 					break
 				}
@@ -735,7 +760,7 @@ func (z *erasureServerPools) rebalanceBucket(ctx context.Context, bucket string,
 						NoAuditLog:         true,
 					},
 				)
-				stopFn(err)
+				stopFn(0, err)
 				auditLogRebalance(ctx, "Rebalance:DeleteObject", bucket, entry.name, "", err)
 				if err != nil {
 					rebalanceLogIf(ctx, err)
@@ -794,7 +819,9 @@ func (z *erasureServerPools) saveRebalanceStats(ctx context.Context, poolIdx int
 	case rebalSaveStoppedAt:
 		r.StoppedAt = time.Now()
 	case rebalSaveStats:
-		r.PoolStats[poolIdx] = z.rebalMeta.PoolStats[poolIdx]
+		if z.rebalMeta != nil {
+			r.PoolStats[poolIdx] = z.rebalMeta.PoolStats[poolIdx]
+		}
 	}
 	z.rebalMeta = r
 
@@ -816,7 +843,7 @@ func auditLogRebalance(ctx context.Context, apiName, bucket, object, versionID s
 	})
 }
 
-func (z *erasureServerPools) rebalanceObject(ctx context.Context, bucket string, gr *GetObjectReader) (err error) {
+func (z *erasureServerPools) rebalanceObject(ctx context.Context, poolIdx int, bucket string, gr *GetObjectReader) (err error) {
 	oi := gr.ObjInfo
 
 	defer func() {
@@ -831,9 +858,11 @@ func (z *erasureServerPools) rebalanceObject(ctx context.Context, bucket string,
 
 	if oi.isMultipart() {
 		res, err := z.NewMultipartUpload(ctx, bucket, oi.Name, ObjectOptions{
-			VersionID:   oi.VersionID,
-			UserDefined: oi.UserDefined,
-			NoAuditLog:  true,
+			VersionID:    oi.VersionID,
+			UserDefined:  oi.UserDefined,
+			NoAuditLog:   true,
+			DataMovement: true,
+			SrcPoolIdx:   poolIdx,
 		})
 		if err != nil {
 			return fmt.Errorf("rebalanceObject: NewMultipartUpload() %w", err)
@@ -885,6 +914,7 @@ func (z *erasureServerPools) rebalanceObject(ctx context.Context, bucket string,
 		oi.Name,
 		NewPutObjReader(hr),
 		ObjectOptions{
+			SrcPoolIdx:   poolIdx,
 			DataMovement: true,
 			VersionID:    oi.VersionID,
 			MTime:        oi.ModTime,
@@ -935,7 +965,7 @@ func (z *erasureServerPools) StartRebalance() {
 		go func(idx int) {
 			stopfn := globalRebalanceMetrics.log(rebalanceMetricRebalanceBuckets, idx)
 			err := z.rebalanceBuckets(ctx, idx)
-			stopfn(err)
+			stopfn(0, err)
 		}(poolIdx)
 	}
 }
@@ -975,7 +1005,9 @@ const (
 	rebalanceMetricSaveMetadata
 )
 
-func rebalanceTrace(r rebalanceMetric, poolIdx int, startTime time.Time, duration time.Duration, err error, path string) madmin.TraceInfo {
+var errDataMovementSrcDstPoolSame = errors.New("source and destination pool are the same")
+
+func rebalanceTrace(r rebalanceMetric, poolIdx int, startTime time.Time, duration time.Duration, err error, path string, sz int64) madmin.TraceInfo {
 	var errStr string
 	if err != nil {
 		errStr = err.Error()
@@ -988,15 +1020,16 @@ func rebalanceTrace(r rebalanceMetric, poolIdx int, startTime time.Time, duratio
 		Duration:  duration,
 		Path:      path,
 		Error:     errStr,
+		Bytes:     sz,
 	}
 }
 
-func (p *rebalanceMetrics) log(r rebalanceMetric, poolIdx int, paths ...string) func(err error) {
+func (p *rebalanceMetrics) log(r rebalanceMetric, poolIdx int, paths ...string) func(sz int64, err error) {
 	startTime := time.Now()
-	return func(err error) {
+	return func(sz int64, err error) {
 		duration := time.Since(startTime)
 		if globalTrace.NumSubscribers(madmin.TraceRebalance) > 0 {
-			globalTrace.Publish(rebalanceTrace(r, poolIdx, startTime, duration, err, strings.Join(paths, " ")))
+			globalTrace.Publish(rebalanceTrace(r, poolIdx, startTime, duration, err, strings.Join(paths, " "), sz))
 		}
 	}
 }

cmd/erasure-server-pool.go

@@ -104,7 +104,11 @@ func newErasureServerPools(ctx context.Context, endpointServerPools EndpointServ
 	// Initialize byte pool once for all sets, bpool size is set to
 	// setCount * setDriveCount with each memory upto blockSizeV2.
 	buffers := bpool.NewBytePoolCap(n, blockSizeV2, blockSizeV2*2)
-	buffers.Populate()
+	if n >= 16384 {
+		// pre-populate buffers only n >= 16384 which is (32Gi/2Mi)
+		// for all setups smaller than this avoid pre-alloc.
+		buffers.Populate()
+	}
 	globalBytePoolCap.Store(buffers)
 
 	var localDrives []StorageAPI
@@ -164,7 +168,10 @@ 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
+		}
 		globalLocalDrivesMu.Unlock()
 	}
 
@@ -180,19 +187,12 @@ func newErasureServerPools(ctx context.Context, endpointServerPools EndpointServ
 		globalLeaderLock = newSharedLock(GlobalContext, z, "leader.lock")
 	})
 
-	// Enable background operations on
-	//
-	// - Disk auto healing
-	// - MRF (most recently failed) healing
-	// - Background expiration routine for lifecycle policies
+	// Start self healing after the object initialization
+	// so various tasks will be useful
 	bootstrapTrace("initAutoHeal", func() {
 		initAutoHeal(GlobalContext, z)
 	})
 
-	bootstrapTrace("initHealMRF", func() {
-		go globalMRFState.healRoutine(z)
-	})
-
 	// initialize the object layer.
 	defer setObjectLayer(z)
 
@@ -1080,6 +1080,14 @@ func (z *erasureServerPools) PutObject(ctx context.Context, bucket string, objec
 		return ObjectInfo{}, err
 	}
 
+	if opts.DataMovement && idx == opts.SrcPoolIdx {
+		return ObjectInfo{}, DataMovementOverwriteErr{
+			Bucket:    bucket,
+			Object:    object,
+			VersionID: opts.VersionID,
+			Err:       errDataMovementSrcDstPoolSame,
+		}
+	}
 	// Overwrite the object at the right pool
 	return z.serverPools[idx].PutObject(ctx, bucket, object, data, opts)
 }
@@ -1132,6 +1140,16 @@ func (z *erasureServerPools) DeleteObject(ctx context.Context, bucket string, ob
 		return pinfo.ObjInfo, nil
 	}
 
+	// Datamovement must never be allowed on the same pool.
+	if opts.DataMovement && opts.SrcPoolIdx == pinfo.Index {
+		return pinfo.ObjInfo, DataMovementOverwriteErr{
+			Bucket:    bucket,
+			Object:    object,
+			VersionID: opts.VersionID,
+			Err:       errDataMovementSrcDstPoolSame,
+		}
+	}
+
 	// Delete concurrently in all server pools with read quorum error for unversioned objects.
 	if len(noReadQuorumPools) > 0 && !opts.Versioned && !opts.VersionSuspended {
 		return z.deleteObjectFromAllPools(ctx, bucket, object, opts, noReadQuorumPools)
@@ -1519,14 +1537,12 @@ func (z *erasureServerPools) listObjectsGeneric(ctx context.Context, bucket, pre
 			loi.NextMarker = last.Name
 		}
 
-		if merged.lastSkippedEntry != "" {
-			if merged.lastSkippedEntry > loi.NextMarker {
-				// An object hidden by ILM was found during listing. Since the number of entries
-				// fetched from drives is limited, set IsTruncated to true to ask the s3 client
-				// to continue listing if it wishes in order to find if there is more objects.
-				loi.IsTruncated = true
-				loi.NextMarker = merged.lastSkippedEntry
-			}
+		if loi.IsTruncated && merged.lastSkippedEntry > loi.NextMarker {
+			// An object hidden by ILM was found during a truncated listing. Since the number of entries
+			// fetched from drives is limited by max-keys, we should use the last ILM filtered entry
+			// as a continuation token if it is lexially higher than the last visible object so that the
+			// next call of WalkDir() with the max-keys can reach new objects not seen previously.
+			loi.NextMarker = merged.lastSkippedEntry
 		}
 
 		if loi.NextMarker != "" {
@@ -1535,7 +1551,7 @@ func (z *erasureServerPools) listObjectsGeneric(ctx context.Context, bucket, pre
 		return loi, nil
 	}
 	ri := logger.GetReqInfo(ctx)
-	hadoop := ri != nil && strings.Contains(ri.UserAgent, `Hadoop `) && strings.Contains(ri.UserAgent, "scala/")
+	hadoop := ri != nil && strings.Contains(ri.UserAgent, "Hadoop ") && strings.Contains(ri.UserAgent, "scala/")
 	matches := func() bool {
 		if prefix == "" {
 			return false
@@ -1600,7 +1616,7 @@ func (z *erasureServerPools) listObjectsGeneric(ctx context.Context, bucket, pre
 		// }
 		if matches() {
 			objInfo, err := z.GetObjectInfo(ctx, bucket, path.Dir(prefix), ObjectOptions{NoLock: true})
-			if err == nil {
+			if err == nil || objInfo.IsLatest && objInfo.DeleteMarker {
 				if opts.Lifecycle != nil {
 					evt := evalActionFromLifecycle(ctx, *opts.Lifecycle, opts.Retention, opts.Replication.Config, objInfo)
 					if evt.Action.Delete() {
@@ -1626,41 +1642,29 @@ func (z *erasureServerPools) listObjectsGeneric(ctx context.Context, bucket, pre
 		// and throw a 404 exception. This is especially a problem for spark jobs overwriting the same partition
 		// repeatedly. This workaround recursively lists the top 3 entries including delete markers to reflect the
 		// correct state of the directory in the list results.
-		opts.Recursive = true
-		opts.InclDeleted = true
-		opts.Limit = maxKeys + 1
-		li, err := listFn(ctx, opts, opts.Limit)
-		if err == nil {
-			switch {
-			case len(li.Objects) == 0 && len(li.Prefixes) == 0:
-				return loi, nil
-			case len(li.Objects) > 0 || len(li.Prefixes) > 0:
-				var o ObjectInfo
-				var pfx string
-				if len(li.Objects) > 0 {
-					o = li.Objects[0]
-					p := strings.TrimPrefix(o.Name, opts.Prefix)
-					if p != "" {
-						sidx := strings.Index(p, "/")
-						if sidx > 0 {
-							pfx = p[:sidx]
-						}
-					}
-				}
-				switch {
-				case o.DeleteMarker:
-					loi.Objects = append(loi.Objects, ObjectInfo{Bucket: bucket, IsDir: true, Name: prefix})
-					return loi, nil
-				case len(li.Objects) >= 1:
-					loi.Objects = append(loi.Objects, o)
-					if pfx != "" {
-						loi.Prefixes = append(loi.Prefixes, path.Join(opts.Prefix, pfx))
-					}
-				case len(li.Prefixes) > 0:
-					loi.Prefixes = append(loi.Prefixes, li.Prefixes...)
-				}
+		if strings.HasSuffix(opts.Prefix, SlashSeparator) {
+			li, err := listFn(ctx, opts, maxKeys)
+			if err != nil {
+				return loi, err
 			}
-			return loi, nil
+			if len(li.Objects) == 0 {
+				prefixes := li.Prefixes[:0]
+				for _, prefix := range li.Prefixes {
+					objInfo, _ := z.GetObjectInfo(ctx, bucket, pathJoin(prefix, "_SUCCESS"), ObjectOptions{NoLock: true})
+					if objInfo.IsLatest && objInfo.DeleteMarker {
+						continue
+					}
+					prefixes = append(prefixes, prefix)
+				}
+				if len(prefixes) > 0 {
+					objInfo, _ := z.GetObjectInfo(ctx, bucket, pathJoin(opts.Prefix, "_SUCCESS"), ObjectOptions{NoLock: true})
+					if objInfo.IsLatest && objInfo.DeleteMarker {
+						return loi, nil
+					}
+				}
+				li.Prefixes = prefixes
+			}
+			return li, nil
 		}
 	}
 
@@ -1759,6 +1763,15 @@ func (z *erasureServerPools) NewMultipartUpload(ctx context.Context, bucket, obj
 		return nil, err
 	}
 
+	if opts.DataMovement && idx == opts.SrcPoolIdx {
+		return nil, DataMovementOverwriteErr{
+			Bucket:    bucket,
+			Object:    object,
+			VersionID: opts.VersionID,
+			Err:       errDataMovementSrcDstPoolSame,
+		}
+	}
+
 	return z.serverPools[idx].NewMultipartUpload(ctx, bucket, object, opts)
 }
 
@@ -1967,6 +1980,34 @@ func (z *erasureServerPools) DeleteBucket(ctx context.Context, bucket string, op
 		defer lk.Unlock(lkctx)
 	}
 
+	if !opts.Force {
+		results := make(chan itemOrErr[ObjectInfo])
+
+		ctx, cancel := context.WithTimeout(ctx, time.Minute)
+		defer cancel()
+		err := z.Walk(ctx, bucket, "", results, WalkOptions{Limit: 1})
+		if err != nil {
+			s3LogIf(ctx, fmt.Errorf("unable to verify if the bucket %s is empty: %w", bucket, err))
+			return toObjectErr(err, bucket)
+		}
+
+		select {
+		case <-ctx.Done():
+			return ctx.Err()
+		case r, found := <-results:
+			if found {
+				if r.Err != nil {
+					s3LogIf(ctx, fmt.Errorf("unable to verify if the bucket %s is empty: %w", bucket, r.Err))
+					return toObjectErr(r.Err, bucket)
+				}
+				return toObjectErr(errVolumeNotEmpty, bucket)
+			}
+		}
+
+		// Always pass force to the lower level
+		opts.Force = true
+	}
+
 	err := z.s3Peer.DeleteBucket(ctx, bucket, opts)
 	if err == nil || isErrBucketNotFound(err) {
 		// If site replication is configured, hold on to deleted bucket state until sites sync
@@ -2013,10 +2054,12 @@ func (z *erasureServerPools) ListBuckets(ctx context.Context, opts BucketOptions
 				if err != nil {
 					return nil, err
 				}
-				for i := range buckets {
-					createdAt, err := globalBucketMetadataSys.CreatedAt(buckets[i].Name)
-					if err == nil {
-						buckets[i].Created = createdAt
+				if !opts.NoMetadata {
+					for i := range buckets {
+						createdAt, err := globalBucketMetadataSys.CreatedAt(buckets[i].Name)
+						if err == nil {
+							buckets[i].Created = createdAt
+						}
 					}
 				}
 				return buckets, nil
@@ -2030,10 +2073,13 @@ func (z *erasureServerPools) ListBuckets(ctx context.Context, opts BucketOptions
 	if err != nil {
 		return nil, err
 	}
-	for i := range buckets {
-		createdAt, err := globalBucketMetadataSys.CreatedAt(buckets[i].Name)
-		if err == nil {
-			buckets[i].Created = createdAt
+
+	if !opts.NoMetadata {
+		for i := range buckets {
+			createdAt, err := globalBucketMetadataSys.CreatedAt(buckets[i].Name)
+			if err == nil {
+				buckets[i].Created = createdAt
+			}
 		}
 	}
 	return buckets, nil
@@ -2087,7 +2133,6 @@ func (z *erasureServerPools) HealBucket(ctx context.Context, bucket string, opts
 
 // Walk a bucket, optionally prefix recursively, until we have returned
 // all the contents of the provided bucket+prefix.
-// TODO: Note that most errors will result in a truncated listing.
 func (z *erasureServerPools) Walk(ctx context.Context, bucket, prefix string, results chan<- itemOrErr[ObjectInfo], opts WalkOptions) error {
 	if err := checkListObjsArgs(ctx, bucket, prefix, ""); err != nil {
 		xioutil.SafeClose(results)
@@ -2105,7 +2150,7 @@ func (z *erasureServerPools) Walk(ctx context.Context, bucket, prefix string, re
 			disks, infos, _ := set.getOnlineDisksWithHealingAndInfo(true)
 			if len(disks) == 0 {
 				xioutil.SafeClose(results)
-				err := fmt.Errorf("Walk: no online disks found in pool %d, set %d", setIdx, poolIdx)
+				err := fmt.Errorf("Walk: no online disks found in (set:%d pool:%d) %w", setIdx, poolIdx, errErasureReadQuorum)
 				cancelCause(err)
 				return err
 			}
@@ -2181,6 +2226,7 @@ func (z *erasureServerPools) Walk(ctx context.Context, bucket, prefix string, re
 					filterPrefix:   filterPrefix,
 					recursive:      true,
 					forwardTo:      opts.Marker,
+					perDiskLimit:   opts.Limit,
 					minDisks:       listingQuorum,
 					reportNotFound: false,
 					agreed:         send,
@@ -2204,9 +2250,8 @@ func (z *erasureServerPools) Walk(ctx context.Context, bucket, prefix string, re
 	// Convert and filter merged entries.
 	merged := make(chan metaCacheEntry, 100)
 	vcfg, _ := globalBucketVersioningSys.Get(bucket)
+	errCh := make(chan error, 1)
 	go func() {
-		defer cancelCause(nil)
-		defer xioutil.SafeClose(results)
 		sentErr := false
 		sendErr := func(err error) {
 			if !sentErr {
@@ -2217,6 +2262,15 @@ func (z *erasureServerPools) Walk(ctx context.Context, bucket, prefix string, re
 				}
 			}
 		}
+		defer func() {
+			select {
+			case <-ctx.Done():
+				sendErr(ctx.Err())
+			default:
+			}
+			xioutil.SafeClose(results)
+			cancelCause(nil)
+		}()
 		send := func(oi ObjectInfo) bool {
 			select {
 			case results <- itemOrErr[ObjectInfo]{Item: oi}:
@@ -2271,12 +2325,16 @@ func (z *erasureServerPools) Walk(ctx context.Context, bucket, prefix string, re
 				}
 			}
 		}
+		if err := <-errCh; err != nil {
+			sendErr(err)
+		}
 	}()
 	go func() {
+		defer close(errCh)
 		// Merge all entries from all disks.
 		// We leave quorum at 1, since entries are already resolved to have the desired quorum.
 		// mergeEntryChannels will close 'merged' channel upon completion or cancellation.
-		storageLogIf(ctx, mergeEntryChannels(ctx, entries, merged, 1))
+		errCh <- mergeEntryChannels(ctx, entries, merged, 1)
 	}()
 
 	return nil
@@ -2332,12 +2390,18 @@ func (z *erasureServerPools) HealObjects(ctx context.Context, bucket, prefix str
 
 	var poolErrs [][]error
 	for idx, erasureSet := range z.serverPools {
+		if opts.Pool != nil && *opts.Pool != idx {
+			continue
+		}
 		if z.IsSuspended(idx) {
 			continue
 		}
 		errs := make([]error, len(erasureSet.sets))
 		var wg sync.WaitGroup
 		for idx, set := range erasureSet.sets {
+			if opts.Set != nil && *opts.Set != idx {
+				continue
+			}
 			wg.Add(1)
 			go func(idx int, set *erasureObjects) {
 				defer wg.Done()
@@ -2430,6 +2494,7 @@ const (
 type HealthOptions struct {
 	Maintenance    bool
 	DeploymentType string
+	NoLogging      bool
 }
 
 // HealthResult returns the current state of the system, also
@@ -2454,6 +2519,24 @@ type HealthResult struct {
 	UsingDefaults bool
 }
 
+func (hr HealthResult) String() string {
+	var str strings.Builder
+	for i, es := range hr.ESHealth {
+		str.WriteString("(Pool: ")
+		str.WriteString(strconv.Itoa(es.PoolID))
+		str.WriteString(" Set: ")
+		str.WriteString(strconv.Itoa(es.SetID))
+		str.WriteString(" Healthy: ")
+		str.WriteString(strconv.FormatBool(es.Healthy))
+		if i == 0 {
+			str.WriteString(")")
+		} else {
+			str.WriteString(") | ")
+		}
+	}
+	return str.String()
+}
+
 // Health - returns current status of the object layer health,
 // provides if write access exists across sets, additionally
 // can be used to query scenarios if health may be lost
@@ -2571,18 +2654,18 @@ func (z *erasureServerPools) Health(ctx context.Context, opts HealthOptions) Hea
 			})
 
 			healthy := erasureSetUpCount[poolIdx][setIdx].online >= poolWriteQuorums[poolIdx]
-			if !healthy {
+			if !healthy && !opts.NoLogging {
 				storageLogIf(logger.SetReqInfo(ctx, reqInfo),
-					fmt.Errorf("Write quorum may be lost on pool: %d, set: %d, expected write quorum: %d",
-						poolIdx, setIdx, poolWriteQuorums[poolIdx]), logger.FatalKind)
+					fmt.Errorf("Write quorum could not be established on pool: %d, set: %d, expected write quorum: %d, drives-online: %d",
+						poolIdx, setIdx, poolWriteQuorums[poolIdx], erasureSetUpCount[poolIdx][setIdx].online), logger.FatalKind)
 			}
 			result.Healthy = result.Healthy && healthy
 
 			healthyRead := erasureSetUpCount[poolIdx][setIdx].online >= poolReadQuorums[poolIdx]
-			if !healthyRead {
+			if !healthyRead && !opts.NoLogging {
 				storageLogIf(logger.SetReqInfo(ctx, reqInfo),
-					fmt.Errorf("Read quorum may be lost on pool: %d, set: %d, expected read quorum: %d",
-						poolIdx, setIdx, poolReadQuorums[poolIdx]))
+					fmt.Errorf("Read quorum could not be established on pool: %d, set: %d, expected read quorum: %d, drives-online: %d",
+						poolIdx, setIdx, poolReadQuorums[poolIdx], erasureSetUpCount[poolIdx][setIdx].online))
 			}
 			result.HealthyRead = result.HealthyRead && healthyRead
 		}
@@ -2746,5 +2829,14 @@ func (z *erasureServerPools) DecomTieredObject(ctx context.Context, bucket, obje
 		return err
 	}
 
+	if opts.DataMovement && idx == opts.SrcPoolIdx {
+		return DataMovementOverwriteErr{
+			Bucket:    bucket,
+			Object:    object,
+			VersionID: opts.VersionID,
+			Err:       errDataMovementSrcDstPoolSame,
+		}
+	}
+
 	return z.serverPools[idx].DecomTieredObject(ctx, bucket, object, fi, opts)
 }

cmd/erasure-sets.go

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

cmd/erasure.go

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

cmd/event-notification.go

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

cmd/fmt-gen.go

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

cmd/ftp-server-driver.go

@@ -105,7 +105,7 @@ type ftpMetrics struct{}
 
 var globalFtpMetrics ftpMetrics
 
-func ftpTrace(s *ftp.Context, startTime time.Time, source, objPath string, err error) madmin.TraceInfo {
+func ftpTrace(s *ftp.Context, startTime time.Time, source, objPath string, err error, sz int64) madmin.TraceInfo {
 	var errStr string
 	if err != nil {
 		errStr = err.Error()
@@ -114,25 +114,33 @@ func ftpTrace(s *ftp.Context, startTime time.Time, source, objPath string, err e
 		TraceType: madmin.TraceFTP,
 		Time:      startTime,
 		NodeName:  globalLocalNodeName,
-		FuncName:  fmt.Sprintf("ftp USER=%s COMMAND=%s PARAM=%s ISLOGIN=%t, Source=%s", s.Sess.LoginUser(), s.Cmd, s.Param, s.Sess.IsLogin(), source),
+		FuncName:  s.Cmd,
 		Duration:  time.Since(startTime),
 		Path:      objPath,
 		Error:     errStr,
+		Bytes:     sz,
+		Custom: map[string]string{
+			"user":   s.Sess.LoginUser(),
+			"cmd":    s.Cmd,
+			"param":  s.Param,
+			"login":  fmt.Sprintf("%t", s.Sess.IsLogin()),
+			"source": source,
+		},
 	}
 }
 
-func (m *ftpMetrics) log(s *ftp.Context, paths ...string) func(err error) {
+func (m *ftpMetrics) log(s *ftp.Context, paths ...string) func(sz int64, err error) {
 	startTime := time.Now()
 	source := getSource(2)
-	return func(err error) {
-		globalTrace.Publish(ftpTrace(s, startTime, source, strings.Join(paths, " "), err))
+	return func(sz int64, err error) {
+		globalTrace.Publish(ftpTrace(s, startTime, source, strings.Join(paths, " "), err, sz))
 	}
 }
 
 // Stat implements ftpDriver
 func (driver *ftpDriver) Stat(ctx *ftp.Context, objPath string) (fi os.FileInfo, err error) {
 	stopFn := globalFtpMetrics.log(ctx, objPath)
-	defer stopFn(err)
+	defer stopFn(0, err)
 
 	if objPath == SlashSeparator {
 		return &minioFileInfo{
@@ -190,7 +198,7 @@ func (driver *ftpDriver) Stat(ctx *ftp.Context, objPath string) (fi os.FileInfo,
 // ListDir implements ftpDriver
 func (driver *ftpDriver) ListDir(ctx *ftp.Context, objPath string, callback func(os.FileInfo) error) (err error) {
 	stopFn := globalFtpMetrics.log(ctx, objPath)
-	defer stopFn(err)
+	defer stopFn(0, err)
 
 	clnt, err := driver.getMinIOClient(ctx)
 	if err != nil {
@@ -252,7 +260,7 @@ func (driver *ftpDriver) ListDir(ctx *ftp.Context, objPath string, callback func
 
 func (driver *ftpDriver) CheckPasswd(c *ftp.Context, username, password string) (ok bool, err error) {
 	stopFn := globalFtpMetrics.log(c, username)
-	defer stopFn(err)
+	defer stopFn(0, err)
 
 	if globalIAMSys.LDAPConfig.Enabled() {
 		sa, _, err := globalIAMSys.getServiceAccount(context.Background(), username)
@@ -376,7 +384,7 @@ func (driver *ftpDriver) getMinIOClient(ctx *ftp.Context) (*minio.Client, error)
 // DeleteDir implements ftpDriver
 func (driver *ftpDriver) DeleteDir(ctx *ftp.Context, objPath string) (err error) {
 	stopFn := globalFtpMetrics.log(ctx, objPath)
-	defer stopFn(err)
+	defer stopFn(0, err)
 
 	bucket, prefix := path2BucketObject(objPath)
 	if bucket == "" {
@@ -426,7 +434,7 @@ func (driver *ftpDriver) DeleteDir(ctx *ftp.Context, objPath string) (err error)
 // DeleteFile implements ftpDriver
 func (driver *ftpDriver) DeleteFile(ctx *ftp.Context, objPath string) (err error) {
 	stopFn := globalFtpMetrics.log(ctx, objPath)
-	defer stopFn(err)
+	defer stopFn(0, err)
 
 	bucket, object := path2BucketObject(objPath)
 	if bucket == "" {
@@ -444,7 +452,7 @@ func (driver *ftpDriver) DeleteFile(ctx *ftp.Context, objPath string) (err error
 // Rename implements ftpDriver
 func (driver *ftpDriver) Rename(ctx *ftp.Context, fromObjPath string, toObjPath string) (err error) {
 	stopFn := globalFtpMetrics.log(ctx, fromObjPath, toObjPath)
-	defer stopFn(err)
+	defer stopFn(0, err)
 
 	return NotImplemented{}
 }
@@ -452,7 +460,7 @@ func (driver *ftpDriver) Rename(ctx *ftp.Context, fromObjPath string, toObjPath
 // MakeDir implements ftpDriver
 func (driver *ftpDriver) MakeDir(ctx *ftp.Context, objPath string) (err error) {
 	stopFn := globalFtpMetrics.log(ctx, objPath)
-	defer stopFn(err)
+	defer stopFn(0, err)
 
 	bucket, prefix := path2BucketObject(objPath)
 	if bucket == "" {
@@ -479,7 +487,7 @@ func (driver *ftpDriver) MakeDir(ctx *ftp.Context, objPath string) (err error) {
 // GetFile implements ftpDriver
 func (driver *ftpDriver) GetFile(ctx *ftp.Context, objPath string, offset int64) (n int64, rc io.ReadCloser, err error) {
 	stopFn := globalFtpMetrics.log(ctx, objPath)
-	defer stopFn(err)
+	defer stopFn(n, err)
 
 	bucket, object := path2BucketObject(objPath)
 	if bucket == "" {
@@ -511,14 +519,14 @@ func (driver *ftpDriver) GetFile(ctx *ftp.Context, objPath string, offset int64)
 	if err != nil {
 		return 0, nil, err
 	}
-
-	return info.Size - offset, obj, nil
+	n = info.Size - offset
+	return n, obj, nil
 }
 
 // PutFile implements ftpDriver
 func (driver *ftpDriver) PutFile(ctx *ftp.Context, objPath string, data io.Reader, offset int64) (n int64, err error) {
 	stopFn := globalFtpMetrics.log(ctx, objPath)
-	defer stopFn(err)
+	defer stopFn(n, err)
 
 	bucket, object := path2BucketObject(objPath)
 	if bucket == "" {
@@ -539,5 +547,6 @@ func (driver *ftpDriver) PutFile(ctx *ftp.Context, objPath string, data io.Reade
 		ContentType:          mimedb.TypeByExtension(path.Ext(object)),
 		DisableContentSha256: true,
 	})
-	return info.Size, err
+	n = info.Size
+	return n, err
 }

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

@@ -150,6 +150,11 @@ type healEntryResult struct {
 
 // healErasureSet lists and heals all objects in a specific erasure set
 func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string, tracker *healingTracker) error {
+	bgSeq, found := globalBackgroundHealState.getHealSequenceByToken(bgHealingUUID)
+	if !found {
+		return errors.New("no local healing sequence initialized, unable to heal the drive")
+	}
+
 	scanMode := madmin.HealNormalScan
 
 	// Make sure to copy since `buckets slice`
@@ -162,9 +167,21 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 		return errServerNotInitialized
 	}
 
+	started := tracker.Started
+	if started.IsZero() || started.Equal(timeSentinel) {
+		healingLogIf(ctx, fmt.Errorf("unexpected tracker healing start time found: %v", started))
+		started = time.Time{}
+	}
+
+	// Final tracer update before quitting
+	defer func() {
+		tracker.setObject("")
+		tracker.setBucket("")
+		healingLogIf(ctx, tracker.update(ctx))
+	}()
+
 	for _, bucket := range healBuckets {
-		_, err := objAPI.HealBucket(ctx, bucket, madmin.HealOpts{ScanMode: scanMode})
-		if err != nil {
+		if err := bgSeq.healBucket(objAPI, bucket, true); err != nil {
 			// Log bucket healing error if any, we shall retry again.
 			healingLogIf(ctx, err)
 		}
@@ -190,7 +207,7 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 		numHealers = uint64(v)
 	}
 
-	healingLogEvent(ctx, fmt.Sprintf("Healing drive '%s' - use %d parallel workers.", tracker.disk.String(), numHealers))
+	healingLogEvent(ctx, "Healing drive '%s' - use %d parallel workers.", tracker.disk.String(), numHealers)
 
 	jt, _ := workers.New(int(numHealers))
 
@@ -223,7 +240,11 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 
 	// Collect updates to tracker from concurrent healEntry calls
 	results := make(chan healEntryResult, 1000)
-	defer close(results)
+	quitting := make(chan struct{})
+	defer func() {
+		close(results)
+		<-quitting
+	}()
 
 	go func() {
 		for res := range results {
@@ -237,6 +258,9 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 
 			tracker.updateProgress(res.success, res.skipped, res.bytes)
 		}
+
+		healingLogIf(ctx, tracker.update(ctx))
+		close(quitting)
 	}()
 
 	var retErr error
@@ -261,9 +285,7 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 		tracker.setBucket(bucket)
 		// Heal current bucket again in case if it is failed
 		// in the beginning of erasure set healing
-		if _, err := objAPI.HealBucket(ctx, bucket, madmin.HealOpts{
-			ScanMode: scanMode,
-		}); err != nil {
+		if err := bgSeq.healBucket(objAPI, bucket, true); err != nil {
 			// Set this such that when we return this function
 			// we let the caller retry this disk again for the
 			// buckets that failed healing.
@@ -362,6 +384,7 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 				}
 				return false
 			case results <- result:
+				bgSeq.countScanned(madmin.HealItemObject)
 				return true
 			}
 		}
@@ -412,8 +435,10 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 						return
 					}
 					result = healEntryFailure(0)
+					bgSeq.countFailed(madmin.HealItemObject)
 					healingLogIf(ctx, fmt.Errorf("unable to heal object %s/%s: %w", bucket, entry.name, err))
 				} else {
+					bgSeq.countHealed(madmin.HealItemObject)
 					result = healEntrySuccess(uint64(res.ObjectSize))
 				}
 
@@ -423,13 +448,10 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 
 			var versionNotFound int
 			for _, version := range fivs.Versions {
-				// Ignore a version with a modtime newer than healing start time.
-				if version.ModTime.After(tracker.Started) {
-					continue
-				}
-
-				// Apply lifecycle rules on the objects that are expired.
-				if filterLifecycle(bucket, version.Name, version) {
+				// Ignore healing a version if:
+				// - It is uploaded after the drive healing is started
+				// - An object that is already expired by ILM rule.
+				if !started.IsZero() && version.ModTime.After(started) || filterLifecycle(bucket, version.Name, version) {
 					versionNotFound++
 					if !send(healEntrySkipped(uint64(version.Size))) {
 						return
@@ -437,6 +459,8 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 					continue
 				}
 
+				var versionHealed bool
+
 				res, err := er.HealObject(ctx, bucket, encodedEntryName,
 					version.VersionID, madmin.HealOpts{
 						ScanMode: scanMode,
@@ -449,15 +473,24 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 						versionNotFound++
 						continue
 					}
-					// If not deleted, assume they failed.
+				} else {
+					// Look for the healing results
+					if res.After.Drives[tracker.DiskIndex].State == madmin.DriveStateOk {
+						versionHealed = true
+					}
+				}
+
+				if versionHealed {
+					bgSeq.countHealed(madmin.HealItemObject)
+					result = healEntrySuccess(uint64(version.Size))
+				} else {
+					bgSeq.countFailed(madmin.HealItemObject)
 					result = healEntryFailure(uint64(version.Size))
 					if version.VersionID != "" {
 						healingLogIf(ctx, fmt.Errorf("unable to heal object %s/%s-v(%s): %w", bucket, version.Name, version.VersionID, err))
 					} else {
 						healingLogIf(ctx, fmt.Errorf("unable to heal object %s/%s: %w", bucket, version.Name, err))
 					}
-				} else {
-					result = healEntrySuccess(uint64(res.ObjectSize))
 				}
 
 				if !send(result) {
@@ -505,7 +538,11 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 				jt.Take()
 				go healEntry(bucket, *entry)
 			},
-			finished: nil,
+			finished: func(errs []error) {
+				if countErrs(errs, nil) != len(errs) {
+					retErr = fmt.Errorf("one or more errors reported during listing: %v", errors.Join(errs...))
+				}
+			},
 		})
 		jt.Wait() // synchronize all the concurrent heal jobs
 		if err != nil {
@@ -513,7 +550,10 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 			// we let the caller retry this disk again for the
 			// buckets it failed to list.
 			retErr = err
-			healingLogIf(ctx, err)
+		}
+
+		if retErr != nil {
+			healingLogIf(ctx, fmt.Errorf("listing failed with: %v on bucket: %v", retErr, bucket))
 			continue
 		}
 
@@ -526,11 +566,16 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 			healingLogIf(ctx, tracker.update(ctx))
 		}
 	}
+	if retErr != nil {
+		return retErr
+	}
 
-	tracker.setObject("")
-	tracker.setBucket("")
+	// Last sanity check
+	if len(tracker.QueuedBuckets) > 0 {
+		return fmt.Errorf("not all buckets were healed: %v", tracker.QueuedBuckets)
+	}
 
-	return retErr
+	return nil
 }
 
 func healBucket(bucket string, scan madmin.HealScanMode) error {

cmd/globals.go

@@ -160,8 +160,9 @@ type serverCtxt struct {
 	FTP  []string
 	SFTP []string
 
+	MemLimit uint64
+
 	UserTimeout         time.Duration
-	ShutdownTimeout     time.Duration
 	IdleTimeout         time.Duration
 	ReadHeaderTimeout   time.Duration
 	MaxIdleConnsPerHost int
@@ -309,6 +310,7 @@ var (
 	globalBootTime = UTCNow()
 
 	globalActiveCred         auth.Credentials
+	globalNodeAuthToken      string
 	globalSiteReplicatorCred siteReplicatorCred
 
 	// Captures if root credentials are set via ENV.
@@ -382,9 +384,7 @@ var (
 	globalBackgroundHealRoutine = newHealRoutine()
 	globalBackgroundHealState   = newHealState(GlobalContext, false)
 
-	globalMRFState = mrfState{
-		opCh: make(chan partialOperation, mrfOpsQueueSize),
-	}
+	globalMRFState = newMRFState()
 
 	// If writes to FS backend should be O_SYNC.
 	globalFSOSync bool
@@ -412,10 +412,9 @@ var (
 	globalServiceFreezeCnt int32
 	globalServiceFreezeMu  sync.Mutex // Updates.
 
-	// List of local drives to this node, this is only set during server startup,
-	// and is only mutated by HealFormat. Hold globalLocalDrivesMu to access.
-	globalLocalDrives    []StorageAPI
-	globalLocalDrivesMap = make(map[string]StorageAPI)
+	// Map of local drives to this node, this is set during server startup,
+	// disk reconnect and mutated by HealFormat. Hold globalLocalDrivesMu to access.
+	globalLocalDrivesMap map[string]StorageAPI
 	globalLocalDrivesMu  sync.RWMutex
 
 	globalDriveMonitoring = env.Get("_MINIO_DRIVE_ACTIVE_MONITORING", config.EnableOn) == config.EnableOn
@@ -448,8 +447,8 @@ var (
 	// dynamic sleeper for multipart expiration routine
 	deleteMultipartCleanupSleeper = newDynamicSleeper(5, 25*time.Millisecond, false)
 
-	// Is _MINIO_DISABLE_API_FREEZE_ON_BOOT set?
-	globalDisableFreezeOnBoot bool
+	// Is MINIO_SYNC_BOOT set?
+	globalEnableSyncBoot bool
 
 	// Contains NIC interface name used for internode communication
 	globalInternodeInterface     string

cmd/grid.go

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

cmd/handler-api.go

@@ -39,14 +39,14 @@ import (
 type apiConfig struct {
 	mu sync.RWMutex
 
-	requestsDeadline      time.Duration
-	requestsPool          chan struct{}
-	clusterDeadline       time.Duration
-	listQuorum            string
-	corsAllowOrigins      []string
-	replicationPriority   string
-	replicationMaxWorkers int
-	transitionWorkers     int
+	requestsPool           chan struct{}
+	clusterDeadline        time.Duration
+	listQuorum             string
+	corsAllowOrigins       []string
+	replicationPriority    string
+	replicationMaxWorkers  int
+	replicationMaxLWorkers int
+	transitionWorkers      int
 
 	staleUploadsExpiry          time.Duration
 	staleUploadsCleanupInterval time.Duration
@@ -91,11 +91,11 @@ func availableMemory() (available uint64) {
 	available = 2048 * blockSizeV2 * 2 // Default to 4 GiB when we can't find the limits.
 
 	if runtime.GOOS == "linux" {
-		// Useful in container mode
+		// Honor cgroup limits if set.
 		limit := cgroupMemLimit()
 		if limit > 0 {
-			// A valid value is found, return its 75%
-			available = (limit * 3) / 4
+			// A valid value is found, return its 90%
+			available = (limit * 9) / 10
 			return
 		}
 	} // for all other platforms limits are based on virtual memory.
@@ -104,8 +104,9 @@ func availableMemory() (available uint64) {
 	if err != nil {
 		return
 	}
-	// A valid value is available return its 75%
-	available = (memStats.Available * 3) / 4
+
+	// A valid value is available return its 90%
+	available = (memStats.Available * 9) / 10
 	return
 }
 
@@ -129,7 +130,7 @@ func (t *apiConfig) init(cfg api.Config, setDriveCounts []int, legacy bool) {
 		maxSetDrives := slices.Max(setDriveCounts)
 
 		// Returns 75% of max memory allowed
-		maxMem := availableMemory()
+		maxMem := globalServerCtxt.MemLimit
 
 		// max requests per node is calculated as
 		// total_ram / ram_per_request
@@ -162,18 +163,18 @@ func (t *apiConfig) init(cfg api.Config, setDriveCounts []int, legacy bool) {
 		// but this shouldn't last long.
 		t.requestsPool = make(chan struct{}, apiRequestsMaxPerNode)
 	}
-	t.requestsDeadline = cfg.RequestsDeadline
 	listQuorum := cfg.ListQuorum
 	if listQuorum == "" {
 		listQuorum = "strict"
 	}
 	t.listQuorum = listQuorum
-	if globalReplicationPool != nil &&
-		(cfg.ReplicationPriority != t.replicationPriority || cfg.ReplicationMaxWorkers != t.replicationMaxWorkers) {
-		globalReplicationPool.ResizeWorkerPriority(cfg.ReplicationPriority, cfg.ReplicationMaxWorkers)
+	if r := globalReplicationPool.GetNonBlocking(); r != nil &&
+		(cfg.ReplicationPriority != t.replicationPriority || cfg.ReplicationMaxWorkers != t.replicationMaxWorkers || cfg.ReplicationMaxLWorkers != t.replicationMaxLWorkers) {
+		r.ResizeWorkerPriority(cfg.ReplicationPriority, cfg.ReplicationMaxWorkers, cfg.ReplicationMaxLWorkers)
 	}
 	t.replicationPriority = cfg.ReplicationPriority
 	t.replicationMaxWorkers = cfg.ReplicationMaxWorkers
+	t.replicationMaxLWorkers = cfg.ReplicationMaxLWorkers
 
 	// N B api.transition_workers will be deprecated
 	if globalTransitionState != nil {
@@ -287,19 +288,15 @@ func (t *apiConfig) getRequestsPoolCapacity() int {
 	return cap(t.requestsPool)
 }
 
-func (t *apiConfig) getRequestsPool() (chan struct{}, time.Duration) {
+func (t *apiConfig) getRequestsPool() chan struct{} {
 	t.mu.RLock()
 	defer t.mu.RUnlock()
 
 	if t.requestsPool == nil {
-		return nil, 10 * time.Second
+		return nil
 	}
 
-	if t.requestsDeadline <= 0 {
-		return t.requestsPool, 10 * time.Second
-	}
-
-	return t.requestsPool, t.requestsDeadline
+	return t.requestsPool
 }
 
 // maxClients throttles the S3 API calls
@@ -322,17 +319,9 @@ func maxClients(f http.HandlerFunc) http.HandlerFunc {
 		}
 
 		globalHTTPStats.addRequestsInQueue(1)
-		defer globalHTTPStats.addRequestsInQueue(-1)
-
-		pool, deadline := globalAPIConfig.getRequestsPool()
+		pool := globalAPIConfig.getRequestsPool()
 		if pool == nil {
-			f.ServeHTTP(w, r)
-			return
-		}
-
-		// No deadline to wait, there is nothing to queue
-		// perform the API call immediately.
-		if deadline <= 0 {
+			globalHTTPStats.addRequestsInQueue(-1)
 			f.ServeHTTP(w, r)
 			return
 		}
@@ -341,19 +330,27 @@ func maxClients(f http.HandlerFunc) http.HandlerFunc {
 			tc.FuncName = "s3.MaxClients"
 		}
 
-		deadlineTimer := time.NewTimer(deadline)
-		defer deadlineTimer.Stop()
+		w.Header().Set("X-RateLimit-Limit", strconv.Itoa(cap(pool)))
+		w.Header().Set("X-RateLimit-Remaining", strconv.Itoa(cap(pool)-len(pool)))
 
 		ctx := r.Context()
 		select {
 		case pool <- struct{}{}:
 			defer func() { <-pool }()
+			globalHTTPStats.addRequestsInQueue(-1)
 			if contextCanceled(ctx) {
 				w.WriteHeader(499)
 				return
 			}
 			f.ServeHTTP(w, r)
-		case <-deadlineTimer.C:
+		case <-r.Context().Done():
+			globalHTTPStats.addRequestsInQueue(-1)
+			// When the client disconnects before getting the S3 handler
+			// status code response, set the status code to 499 so this request
+			// will be properly audited and traced.
+			w.WriteHeader(499)
+		default:
+			globalHTTPStats.addRequestsInQueue(-1)
 			if contextCanceled(ctx) {
 				w.WriteHeader(499)
 				return
@@ -362,11 +359,7 @@ func maxClients(f http.HandlerFunc) http.HandlerFunc {
 			writeErrorResponse(ctx, w,
 				errorCodes.ToAPIErr(ErrTooManyRequests),
 				r.URL)
-		case <-r.Context().Done():
-			// When the client disconnects before getting the S3 handler
-			// status code response, set the status code to 499 so this request
-			// will be properly audited and traced.
-			w.WriteHeader(499)
+
 		}
 	}
 }
@@ -377,14 +370,16 @@ func (t *apiConfig) getReplicationOpts() replicationPoolOpts {
 
 	if t.replicationPriority == "" {
 		return replicationPoolOpts{
-			Priority:   "auto",
-			MaxWorkers: WorkerMaxLimit,
+			Priority:    "auto",
+			MaxWorkers:  WorkerMaxLimit,
+			MaxLWorkers: LargeWorkerCount,
 		}
 	}
 
 	return replicationPoolOpts{
-		Priority:   t.replicationPriority,
-		MaxWorkers: t.replicationMaxWorkers,
+		Priority:    t.replicationPriority,
+		MaxWorkers:  t.replicationMaxWorkers,
+		MaxLWorkers: t.replicationMaxLWorkers,
 	}
 }
 

cmd/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, ",")
 			}