Add Full Object Checksums and CRC64-NVME (#20855 )

Backport of AIStor PR 247. Add support for full object checksums as described here: https://docs.aws.amazon.com/AmazonS3/latest/userguide/checking-object-integrity.html New checksum types are fully supported. Mint tests from https://github.com/minio/minio-go/pull/2026 are now passing. Includes fixes from https://github.com/minio/minio/pull/20743 for mint tests. Add using checksums as validation for object content. Fixes #20845 #20849 Fixes checksum replication (downstream PR 250)
do not list buckets without local quorum (#20852 )
2026-10-14 21:55:58 +03:00 · 2025-01-20 06:49:07 -08:00 · 2025-01-19 15:13:17 -08:00 · 2025-01-18 01:38:52 +00:00 · 2025-01-17 16:31:37 -08:00 · 2025-01-17 13:12:21 -08:00
392 changed files with 24460 additions and 13238 deletions
@@ -20,7 +20,7 @@ jobs:
    runs-on: ${{ matrix.os }}
    strategy:
      matrix:
-        go-version: [1.22.x]
+        go-version: [1.23.x]
        os: [ubuntu-latest]
    steps:
      - uses: actions/checkout@v4
@@ -20,7 +20,7 @@ jobs:
    runs-on: ${{ matrix.os }}
    strategy:
      matrix:
-        go-version: [1.22.x]
+        go-version: [1.23.x]
        os: [ubuntu-latest]
    steps:
      - uses: actions/checkout@v4
@@ -20,7 +20,7 @@ jobs:
    runs-on: ${{ matrix.os }}
    strategy:
      matrix:
-        go-version: [1.22.x]
+        go-version: [1.23.x]
        os: [ubuntu-latest]
    steps:
      - uses: actions/checkout@v4
@@ -20,24 +20,14 @@ jobs:
    runs-on: ${{ matrix.os }}
    strategy:
      matrix:
-        go-version: [1.22.x]
-        os: [ubuntu-latest, Windows]
+        go-version: [1.23.x]
+        os: [ubuntu-latest]
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-go@v5
        with:
          go-version: ${{ matrix.go-version }}
          check-latest: true
-      - name: Build on ${{ matrix.os }}
-        if: matrix.os == 'Windows'
-        env:
-          CGO_ENABLED: 0
-          GO111MODULE: on
-        run: |
-          Set-MpPreference -DisableRealtimeMonitoring $true
-          netsh int ipv4 set dynamicport tcp start=60000 num=61000
-          go build --ldflags="-s -w" -o %GOPATH%\bin\minio.exe
-          go test -v --timeout 120m ./...
      - name: Build on ${{ matrix.os }}
        if: matrix.os == 'ubuntu-latest'
        env:
@@ -0,0 +1,39 @@
+name: Resiliency Functional Tests
+
+on:
+  pull_request:
+    branches:
+      - master
+
+# This ensures that previous jobs for the PR are canceled when the PR is
+# updated.
+concurrency:
+  group: ${{ github.workflow }}-${{ github.head_ref }}
+  cancel-in-progress: true
+
+permissions:
+  contents: read
+
+jobs:
+  build:
+    name: Go ${{ matrix.go-version }} on ${{ matrix.os }}
+    runs-on: ${{ matrix.os }}
+    strategy:
+      matrix:
+        go-version: [1.23.x]
+        os: [ubuntu-latest]
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-go@v5
+        with:
+          go-version: ${{ matrix.go-version }}
+          check-latest: true
+      - name: Build on ${{ matrix.os }}
+        if: matrix.os == 'ubuntu-latest'
+        env:
+          CGO_ENABLED: 0
+          GO111MODULE: on
+        run: |
+          sudo sysctl net.ipv6.conf.all.disable_ipv6=0
+          sudo sysctl net.ipv6.conf.default.disable_ipv6=0
+          make test-resiliency
@@ -20,7 +20,7 @@ jobs:
    runs-on: ${{ matrix.os }}
    strategy:
      matrix:
-        go-version: [1.22.x]
+        go-version: [1.23.x]
        os: [ubuntu-latest]
    steps:
      - uses: actions/checkout@v4
@@ -39,3 +39,4 @@ jobs:
          sudo sysctl net.ipv6.conf.all.disable_ipv6=0
          sudo sysctl net.ipv6.conf.default.disable_ipv6=0
          make verify
+          make test-timeout
@@ -61,7 +61,7 @@ jobs:
      # are turned off - i.e. if ldap="", then ldap server is not enabled for
      # the tests.
      matrix:
-        go-version: [1.22.x]
+        go-version: [1.23.x]
        ldap: ["", "localhost:389"]
        etcd: ["", "http://localhost:2379"]
        openid: ["", "http://127.0.0.1:5556/dex"]
@@ -125,3 +125,37 @@ jobs:
        if: matrix.openid == 'http://127.0.0.1:5556/dex'
        run: |
          make test-site-replication-oidc
+  iam-import-with-missing-entities:
+    name: Test IAM import in new cluster with missing entities
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-go@v5
+        with:
+          go-version: ${{ matrix.go-version }}
+          check-latest: true
+      - name: Checkout minio-iam-testing
+        uses: actions/checkout@v4
+        with:
+          repository: minio/minio-iam-testing
+          path: minio-iam-testing
+      - name: Test import of IAM artifacts when in fresh cluster there are missing groups etc
+        run: |
+          make test-iam-import-with-missing-entities
+  iam-import-with-openid:
+    name: Test IAM import in new cluster with opendid configurations
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-go@v5
+        with:
+          go-version: ${{ matrix.go-version }}
+          check-latest: true
+      - name: Checkout minio-iam-testing
+        uses: actions/checkout@v4
+        with:
+          repository: minio/minio-iam-testing
+          path: minio-iam-testing
+      - name: Test import of IAM artifacts when in fresh cluster with openid configurations
+        run: |
+          make test-iam-import-with-openid
@@ -29,7 +29,7 @@ jobs:
      - name: setup-go-step
        uses: actions/setup-go@v5
        with:
-          go-version: 1.22.x
+          go-version: 1.23.x

      - name: github sha short
        id: vars
@@ -21,7 +21,7 @@ jobs:

    strategy:
      matrix:
-        go-version: [1.22.x]
+        go-version: [1.23.x]

    steps:
      - uses: actions/checkout@v4
@@ -40,6 +40,7 @@ jobs:
          sudo sysctl net.ipv6.conf.all.disable_ipv6=0
          sudo sysctl net.ipv6.conf.default.disable_ipv6=0
          make test-ilm
+          make test-ilm-transition

      - name: Test PBAC
        run: |
@@ -70,3 +71,9 @@ jobs:
          sudo sysctl net.ipv6.conf.all.disable_ipv6=0
          sudo sysctl net.ipv6.conf.default.disable_ipv6=0
          make test-versioning
+
+      - name: Test Multipart upload with failures
+        run: |
+          sudo sysctl net.ipv6.conf.all.disable_ipv6=0
+          sudo sysctl net.ipv6.conf.default.disable_ipv6=0
+          make test-multipart
@@ -20,7 +20,7 @@ jobs:
    runs-on: ${{ matrix.os }}
    strategy:
      matrix:
-        go-version: [1.22.x]
+        go-version: [1.23.x]
        os: [ubuntu-latest]

    steps:
@@ -15,6 +15,9 @@ docker volume rm $(docker volume ls -f dangling=true) || true
 ## change working directory
 cd .github/workflows/mint

+## always pull latest
+docker pull docker.io/minio/mint:edge
+
 docker-compose -f minio-${MODE}.yaml up -d
 sleep 1m

@@ -20,7 +20,7 @@ jobs:
    runs-on: ${{ matrix.os }}
    strategy:
      matrix:
-        go-version: [1.22.x]
+        go-version: [1.23.x]
        os: [ubuntu-latest]

    steps:
@@ -21,7 +21,7 @@ jobs:
      - name: Set up Go
        uses: actions/setup-go@v5
        with:
-          go-version: 1.22.5
+          go-version: 1.23.2
      - name: Get official govulncheck
        run: go install golang.org/x/vuln/cmd/govulncheck@latest
        shell: bash
@@ -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``
@@ -1,5 +1,7 @@
 FROM minio/minio:latest

+RUN chmod -R 777 /usr/bin
+
 COPY ./minio /usr/bin/minio
 COPY dockerscripts/docker-entrypoint.sh /usr/bin/docker-entrypoint.sh

@@ -1,12 +0,0 @@
-FROM minio/minio:latest
-
-ENV PATH=/opt/bin:$PATH
-
-COPY ./minio /opt/bin/minio
-COPY dockerscripts/docker-entrypoint.sh /usr/bin/docker-entrypoint.sh
-
-ENTRYPOINT ["/usr/bin/docker-entrypoint.sh"]
-
-VOLUME ["/data"]
-
-CMD ["minio"]
@@ -1,24 +1,26 @@
-FROM golang:1.21-alpine as build
+FROM golang:1.23-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 \
@@ -51,9 +53,11 @@ ENV MINIO_ACCESS_KEY_FILE=access_key \
    MINIO_CONFIG_ENV_FILE=config.env \
    MC_CONFIG_DIR=/tmp/.mc

+RUN chmod -R 777 /usr/bin
+
 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
@@ -1,35 +1,39 @@
-FROM golang:1.21-alpine as build
+FROM golang:1.23-alpine AS build

 ARG TARGETARCH
 ARG RELEASE

-ENV GOPATH /go
-ENV CGO_ENABLED 0
+ENV GOPATH=/go
+ENV CGO_ENABLED=0
+
+WORKDIR /build

 # Install curl and minisign
 RUN apk add -U --no-cache ca-certificates && \
    apk add -U --no-cache curl && \
+    apk add -U --no-cache bash && \
    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 [ "$TARGETARCH" = "amd64" ]; then \
-       curl -L -s -q https://github.com/moparisthebest/static-curl/releases/latest/download/curl-${TARGETARCH} -o /go/bin/curl; \
-       chmod +x /go/bin/curl; \
-    fi
-
 # Verify binary signature using public key "RWTx5Zr1tiHQLwG9keckT0c45M3AGeHD6IvimQHpyRywVWGbP1aVSGavRUN"
 RUN minisign -Vqm /go/bin/minio -x /go/bin/minio.minisig -P RWTx5Zr1tiHQLwG9keckT0c45M3AGeHD6IvimQHpyRywVWGbP1aVSGav && \
    minisign -Vqm /go/bin/mc -x /go/bin/mc.minisig -P RWTx5Zr1tiHQLwG9keckT0c45M3AGeHD6IvimQHpyRywVWGbP1aVSGav

+COPY dockerscripts/download-static-curl.sh /build/download-static-curl
+RUN chmod +x /build/download-static-curl && \
+    /build/download-static-curl
+
 FROM registry.access.redhat.com/ubi9/ubi-micro:latest

 ARG RELEASE
@@ -51,10 +55,12 @@ ENV MINIO_ACCESS_KEY_FILE=access_key \
    MINIO_CONFIG_ENV_FILE=config.env \
    MC_CONFIG_DIR=/tmp/.mc

+RUN chmod -R 777 /usr/bin
+
 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/cur* /usr/bin/
+COPY --from=build /go/bin/minio* /usr/bin/
+COPY --from=build /go/bin/mc* /usr/bin/
+COPY --from=build /go/bin/curl* /usr/bin/

 COPY CREDITS /licenses/CREDITS
 COPY LICENSE /licenses/LICENSE
@@ -1,21 +1,28 @@
-FROM golang:1.21-alpine as build
+FROM golang:1.23-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; \
@@ -44,8 +51,11 @@ ENV MINIO_ACCESS_KEY_FILE=access_key \
    MINIO_UPDATE_MINISIGN_PUBKEY="RWTx5Zr1tiHQLwG9keckT0c45M3AGeHD6IvimQHpyRywVWGbP1aVSGav" \
    MINIO_CONFIG_ENV_FILE=config.env

+RUN chmod -R 777 /usr/bin
+
 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
@@ -1,24 +1,26 @@
-FROM golang:1.21-alpine as build
+FROM golang:1.23-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 \
@@ -51,9 +53,11 @@ ENV MINIO_ACCESS_KEY_FILE=access_key \
    MINIO_CONFIG_ENV_FILE=config.env \
    MC_CONFIG_DIR=/tmp/.mc

+RUN chmod -R 777 /usr/bin
+
 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
@@ -24,7 +24,7 @@ help: ## print this help
 getdeps: ## fetch necessary dependencies
 	@mkdir -p ${GOPATH}/bin
 	@echo "Installing golangci-lint" && curl -sSfL https://raw.githubusercontent.com/golangci/golangci-lint/master/install.sh | sh -s -- -b $(GOLANGCI_DIR)
-	@echo "Installing msgp" && go install -v github.com/tinylib/msgp@v1.1.10-0.20240227114326-6d6f813fff1b
+	@echo "Installing msgp" && go install -v github.com/tinylib/msgp@v1.2.5
 	@echo "Installing stringer" && go install -v golang.org/x/tools/cmd/stringer@latest

 crosscompile: ## cross compile minio
@@ -60,6 +60,10 @@ test-ilm: install-race
 	@echo "Running ILM tests"
 	@env bash $(PWD)/docs/bucket/replication/setup_ilm_expiry_replication.sh

+test-ilm-transition: install-race
+	@echo "Running ILM tiering tests with healing"
+	@env bash $(PWD)/docs/bucket/lifecycle/setup_ilm_transition.sh
+
 test-pbac: install-race
 	@echo "Running bucket policies tests"
 	@env bash $(PWD)/docs/iam/policies/pbac-tests.sh
@@ -97,6 +101,14 @@ 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

+test-iam-import-with-missing-entities: install-race ## test import of external iam config withg missing entities
+	@echo "Test IAM import configurations with missing entities"
+	@env bash $(PWD)/docs/distributed/iam-import-with-missing-entities.sh
+
+test-iam-import-with-openid: install-race
+	@echo "Test IAM import configurations with openid"
+	@env bash $(PWD)/docs/distributed/iam-import-with-openid.sh
+
 test-sio-error:
 	@(env bash $(PWD)/docs/bucket/replication/sio-error.sh)

@@ -133,6 +145,14 @@ test-site-replication-minio: install-race ## verify automatic site replication
 	@echo "Running tests for automatic site replication of SSE-C objects with compression enabled for site"
 	@(env bash $(PWD)/docs/site-replication/run-ssec-object-replication-with-compression.sh)

+test-multipart: install-race ## test multipart
+	@echo "Test multipart behavior when part files are missing"
+	@(env bash $(PWD)/buildscripts/multipart-quorum-test.sh)
+
+test-timeout: install-race ## test multipart
+	@echo "Test server timeout"
+	@(env bash $(PWD)/buildscripts/test-timeout.sh)
+
 verify: install-race ## verify minio various setups
 	@echo "Verifying build with race"
 	@(env bash $(PWD)/buildscripts/verify-build.sh)
@@ -197,6 +217,10 @@ docker: build ## builds minio docker container
 	@echo "Building minio docker image '$(TAG)'"
 	@docker build -q --no-cache -t $(TAG) . -f Dockerfile

+test-resiliency: build
+	@echo "Running resiliency tests"
+	@(DOCKER_COMPOSE_FILE=$(PWD)/docs/resiliency/docker-compose.yaml env bash $(PWD)/docs/resiliency/resiliency-tests.sh)
+
 install-race: checks build-debugging ## builds minio to $(PWD)
 	@echo "Building minio binary with -race to './minio'"
 	@GORACE=history_size=7 CGO_ENABLED=1 go build -tags kqueue,dev -race -trimpath --ldflags "$(LDFLAGS)" -o $(PWD)/minio 1>/dev/null
@@ -4,7 +4,7 @@

 [![MinIO](https://raw.githubusercontent.com/minio/minio/master/.github/logo.svg?sanitize=true)](https://min.io)

-MinIO is a High Performance Object Storage released under GNU Affero General Public License v3.0. It is API compatible with Amazon S3 cloud storage service. Use MinIO to build high performance infrastructure for machine learning, analytics and application data workloads.
+MinIO is a High Performance Object Storage released under GNU Affero General Public License v3.0. It is API compatible with Amazon S3 cloud storage service. Use MinIO to build high performance infrastructure for machine learning, analytics and application data workloads. To learn more about what MinIO is doing for AI storage, go to [AI storage documentation](https://min.io/solutions/object-storage-for-ai).

 This README provides quickstart instructions on running MinIO on bare metal hardware, including container-based installations. For Kubernetes environments, use the [MinIO Kubernetes Operator](https://github.com/minio/operator/blob/master/README.md).

@@ -93,7 +93,6 @@ The following table lists supported architectures. Replace the `wget` URL with t
 | 64-bit Intel/AMD               | <https://dl.min.io/server/minio/release/linux-amd64/minio>   |
 | 64-bit ARM                     | <https://dl.min.io/server/minio/release/linux-arm64/minio>   |
 | 64-bit PowerPC LE (ppc64le)    | <https://dl.min.io/server/minio/release/linux-ppc64le/minio> |
-| IBM Z-Series (S390X)           | <https://dl.min.io/server/minio/release/linux-s390x/minio>   |

 The MinIO deployment starts using default root credentials `minioadmin:minioadmin`. You can test the deployment using the MinIO Console, an embedded web-based object browser built into MinIO Server. Point a web browser running on the host machine to <http://127.0.0.1:9000> and log in with the root credentials. You can use the Browser to create buckets, upload objects, and browse the contents of the MinIO server.

@@ -8,7 +8,7 @@
 #
 # This script assumes that LDAP server is at:
 #
-#   `localhost:1389`
+#   `localhost:389`
 #
 # if this is not the case, set the environment variable
 # `_MINIO_LDAP_TEST_SERVER`.
@@ -41,7 +41,7 @@ __init__() {
 	fi

 	if [ -z "$_MINIO_LDAP_TEST_SERVER" ]; then
-		export _MINIO_LDAP_TEST_SERVER=localhost:1389
+		export _MINIO_LDAP_TEST_SERVER=localhost:389
 		echo "Using default LDAP endpoint: $_MINIO_LDAP_TEST_SERVER"
 	fi

@@ -58,7 +58,7 @@ create_iam_content_in_old_minio() {
 	mc alias set old-minio http://localhost:9000 minioadmin minioadmin
 	mc ready old-minio
 	mc idp ldap add old-minio \
-		server_addr=localhost:1389 \
+		server_addr=localhost:389 \
 		server_insecure=on \
 		lookup_bind_dn=cn=admin,dc=min,dc=io \
 		lookup_bind_password=admin \
@@ -0,0 +1,126 @@
+#!/bin/bash
+
+if [ -n "$TEST_DEBUG" ]; then
+	set -x
+fi
+
+WORK_DIR="$PWD/.verify-$RANDOM"
+MINIO_CONFIG_DIR="$WORK_DIR/.minio"
+MINIO=("$PWD/minio" --config-dir "$MINIO_CONFIG_DIR" server)
+
+if [ ! -x "$PWD/minio" ]; then
+	echo "minio executable binary not found in current directory"
+	exit 1
+fi
+
+if [ ! -x "$PWD/minio" ]; then
+	echo "minio executable binary not found in current directory"
+	exit 1
+fi
+
+trap 'catch $LINENO' ERR
+
+function purge() {
+	rm -rf "$1"
+}
+
+# shellcheck disable=SC2120
+catch() {
+	if [ $# -ne 0 ]; then
+		echo "error on line $1"
+	fi
+
+	echo "Cleaning up instances of MinIO"
+	pkill minio || true
+	pkill -9 minio || true
+	purge "$WORK_DIR"
+	if [ $# -ne 0 ]; then
+		exit $#
+	fi
+}
+
+catch
+
+function start_minio_10drive() {
+	start_port=$1
+
+	export MINIO_ROOT_USER=minio
+	export MINIO_ROOT_PASSWORD=minio123
+	export MC_HOST_minio="http://minio:minio123@127.0.0.1:${start_port}/"
+	unset MINIO_KMS_AUTO_ENCRYPTION # do not auto-encrypt objects
+	export MINIO_CI_CD=1
+
+	mkdir ${WORK_DIR}
+	C_PWD=${PWD}
+	if [ ! -x "$PWD/mc" ]; then
+		MC_BUILD_DIR="mc-$RANDOM"
+		if ! git clone --quiet https://github.com/minio/mc "$MC_BUILD_DIR"; then
+			echo "failed to download https://github.com/minio/mc"
+			purge "${MC_BUILD_DIR}"
+			exit 1
+		fi
+
+		(cd "${MC_BUILD_DIR}" && go build -o "$C_PWD/mc")
+
+		# remove mc source.
+		purge "${MC_BUILD_DIR}"
+	fi
+
+	"${MINIO[@]}" --address ":$start_port" "${WORK_DIR}/disk{1...10}" >"${WORK_DIR}/server1.log" 2>&1 &
+	pid=$!
+	disown $pid
+	sleep 5
+
+	if ! ps -p ${pid} 1>&2 >/dev/null; then
+		echo "server1 log:"
+		cat "${WORK_DIR}/server1.log"
+		echo "FAILED"
+		purge "$WORK_DIR"
+		exit 1
+	fi
+
+	"${PWD}/mc" mb --with-versioning minio/bucket
+
+	export AWS_ACCESS_KEY_ID=minio
+	export AWS_SECRET_ACCESS_KEY=minio123
+	aws --endpoint-url http://localhost:"$start_port" s3api create-multipart-upload --bucket bucket --key obj-1 >upload-id.json
+	uploadId=$(jq -r '.UploadId' upload-id.json)
+
+	truncate -s 5MiB file-5mib
+	for i in {1..2}; do
+		aws --endpoint-url http://localhost:"$start_port" s3api upload-part \
+			--upload-id "$uploadId" --bucket bucket --key obj-1 \
+			--part-number "$i" --body ./file-5mib
+	done
+	for i in {1..6}; do
+		find ${WORK_DIR}/disk${i}/.minio.sys/multipart/ -type f -name "part.1" -delete
+	done
+	cat <<EOF >parts.json
+{
+    "Parts": [
+        {
+            "PartNumber": 1,
+            "ETag": "5f363e0e58a95f06cbe9bbc662c5dfb6"
+        },
+        {
+            "PartNumber": 2,
+            "ETag": "5f363e0e58a95f06cbe9bbc662c5dfb6"
+        }
+    ]
+}
+EOF
+	err=$(aws --endpoint-url http://localhost:"$start_port" s3api complete-multipart-upload --upload-id "$uploadId" --bucket bucket --key obj-1 --multipart-upload file://./parts.json 2>&1)
+	rv=$?
+	if [ $rv -eq 0 ]; then
+		echo "Failed to receive an error"
+		exit 1
+	fi
+	echo "Received an error during complete-multipart as expected: $err"
+}
+
+function main() {
+	start_port=$(shuf -i 10000-65000 -n 1)
+	start_minio_10drive ${start_port}
+}
+
+main "$@"
@@ -0,0 +1,137 @@
+#!/bin/bash
+
+if [ -n "$TEST_DEBUG" ]; then
+	set -x
+fi
+
+WORK_DIR="$PWD/.verify-$RANDOM"
+MINIO_CONFIG_DIR="$WORK_DIR/.minio"
+MINIO=("$PWD/minio" --config-dir "$MINIO_CONFIG_DIR" server)
+
+if [ ! -x "$PWD/minio" ]; then
+	echo "minio executable binary not found in current directory"
+	exit 1
+fi
+
+if [ ! -x "$PWD/minio" ]; then
+	echo "minio executable binary not found in current directory"
+	exit 1
+fi
+
+trap 'catch $LINENO' ERR
+
+function purge() {
+	rm -rf "$1"
+}
+
+# shellcheck disable=SC2120
+catch() {
+	if [ $# -ne 0 ]; then
+		echo "error on line $1"
+	fi
+
+	echo "Cleaning up instances of MinIO"
+	pkill minio || true
+	pkill -9 minio || true
+	purge "$WORK_DIR"
+	if [ $# -ne 0 ]; then
+		exit $#
+	fi
+}
+
+catch
+
+function gen_put_request() {
+	hdr_sleep=$1
+	body_sleep=$2
+
+	echo "PUT /testbucket/testobject HTTP/1.1"
+	sleep $hdr_sleep
+	echo "Host: foo-header"
+	echo "User-Agent: curl/8.2.1"
+	echo "Accept: */*"
+	echo "Content-Length: 30"
+	echo ""
+
+	sleep $body_sleep
+	echo "random line 0"
+	echo "random line 1"
+	echo ""
+	echo ""
+}
+
+function send_put_object_request() {
+	hdr_timeout=$1
+	body_timeout=$2
+
+	start=$(date +%s)
+	timeout 5m bash -c "gen_put_request $hdr_timeout $body_timeout | netcat 127.0.0.1 $start_port | read" || return -1
+	[ $(($(date +%s) - start)) -gt $((srv_hdr_timeout + srv_idle_timeout + 1)) ] && return -1
+	return 0
+}
+
+function test_minio_with_timeout() {
+	start_port=$1
+
+	export MINIO_ROOT_USER=minio
+	export MINIO_ROOT_PASSWORD=minio123
+	export MC_HOST_minio="http://minio:minio123@127.0.0.1:${start_port}/"
+	export MINIO_CI_CD=1
+
+	mkdir ${WORK_DIR}
+	C_PWD=${PWD}
+	if [ ! -x "$PWD/mc" ]; then
+		MC_BUILD_DIR="mc-$RANDOM"
+		if ! git clone --quiet https://github.com/minio/mc "$MC_BUILD_DIR"; then
+			echo "failed to download https://github.com/minio/mc"
+			purge "${MC_BUILD_DIR}"
+			exit 1
+		fi
+
+		(cd "${MC_BUILD_DIR}" && go build -o "$C_PWD/mc")
+
+		# remove mc source.
+		purge "${MC_BUILD_DIR}"
+	fi
+
+	"${MINIO[@]}" --address ":$start_port" --read-header-timeout ${srv_hdr_timeout}s --idle-timeout ${srv_idle_timeout}s "${WORK_DIR}/disk/" >"${WORK_DIR}/server1.log" 2>&1 &
+	pid=$!
+	disown $pid
+	sleep 1
+
+	if ! ps -p ${pid} 1>&2 >/dev/null; then
+		echo "server1 log:"
+		cat "${WORK_DIR}/server1.log"
+		echo "FAILED"
+		purge "$WORK_DIR"
+		exit 1
+	fi
+
+	set -e
+
+	"${PWD}/mc" mb minio/testbucket
+	"${PWD}/mc" anonymous set public minio/testbucket
+
+	# slow header writing
+	send_put_object_request 20 0 && exit -1
+	"${PWD}/mc" stat minio/testbucket/testobject && exit -1
+
+	# quick header write and slow bodywrite
+	send_put_object_request 0 40 && exit -1
+	"${PWD}/mc" stat minio/testbucket/testobject && exit -1
+
+	# quick header and body write
+	send_put_object_request 1 1 || exit -1
+	"${PWD}/mc" stat minio/testbucket/testobject || exit -1
+}
+
+function main() {
+	export start_port=$(shuf -i 10000-65000 -n 1)
+	export srv_hdr_timeout=5
+	export srv_idle_timeout=5
+	export -f gen_put_request
+
+	test_minio_with_timeout ${start_port}
+}
+
+main "$@"
@@ -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
 			}

@@ -190,7 +190,7 @@ func (a adminAPIHandlers) AttachDetachPolicyLDAP(w http.ResponseWriter, r *http.
 //
 // PUT /minio/admin/v3/idp/ldap/add-service-account
 func (a adminAPIHandlers) AddServiceAccountLDAP(w http.ResponseWriter, r *http.Request) {
-	ctx, cred, opts, createReq, targetUser, APIError := commonAddServiceAccount(r)
+	ctx, cred, opts, createReq, targetUser, APIError := commonAddServiceAccount(r, true)
 	if APIError.Code != "" {
 		writeErrorResponseJSON(ctx, w, APIError, r.URL)
 		return
@@ -499,7 +499,7 @@ func (a adminAPIHandlers) ListAccessKeysLDAPBulk(w http.ResponseWriter, r *http.

 	dnList := r.Form["userDNs"]
 	isAll := r.Form.Get("all") == "true"
-	onlySelf := !isAll && len(dnList) == 0
+	selfOnly := !isAll && len(dnList) == 0

 	if isAll && len(dnList) > 0 {
 		// This should be checked on client side, so return generic error
@@ -527,7 +527,7 @@ func (a adminAPIHandlers) ListAccessKeysLDAPBulk(w http.ResponseWriter, r *http.
 			dn = foundResult.NormDN
 		}
 		if dn == cred.ParentUser || dnList[0] == cred.ParentUser {
-			onlySelf = true
+			selfOnly = true
 		}
 	}

@@ -538,13 +538,13 @@ func (a adminAPIHandlers) ListAccessKeysLDAPBulk(w http.ResponseWriter, r *http.
 		ConditionValues: getConditionValues(r, "", cred),
 		IsOwner:         owner,
 		Claims:          cred.Claims,
-		DenyOnly:        onlySelf,
+		DenyOnly:        selfOnly,
 	}) {
 		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrAccessDenied), r.URL)
 		return
 	}

-	if onlySelf && len(dnList) == 0 {
+	if selfOnly && len(dnList) == 0 {
 		selfDN := cred.AccessKey
 		if cred.ParentUser != "" {
 			selfDN = cred.ParentUser
@@ -609,10 +609,9 @@ func (a adminAPIHandlers) ListAccessKeysLDAPBulk(w http.ResponseWriter, r *http.
 				return
 			}
 			for _, sts := range stsKeys {
-				expiryTime := sts.Expiration
 				accessKeys.STSKeys = append(accessKeys.STSKeys, madmin.ServiceAccountInfo{
 					AccessKey:  sts.AccessKey,
-					Expiration: &expiryTime,
+					Expiration: &sts.Expiration,
 				})
 			}
 			// if only STS keys, skip if user has no STS keys
@@ -628,10 +627,9 @@ func (a adminAPIHandlers) ListAccessKeysLDAPBulk(w http.ResponseWriter, r *http.
 				return
 			}
 			for _, svc := range serviceAccounts {
-				expiryTime := svc.Expiration
 				accessKeys.ServiceAccounts = append(accessKeys.ServiceAccounts, madmin.ServiceAccountInfo{
 					AccessKey:  svc.AccessKey,
-					Expiration: &expiryTime,
+					Expiration: &svc.Expiration,
 				})
 			}
 			// if only service accounts, skip if user has no service accounts
@@ -26,16 +26,18 @@ import (
 	"io"
 	"net/http"
 	"os"
+	"slices"
 	"sort"
 	"strconv"
+	"strings"
 	"time"
 	"unicode/utf8"

 	"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/minio/internal/logger"
 	"github.com/minio/mux"
 	xldap "github.com/minio/pkg/v3/ldap"
 	"github.com/minio/pkg/v3/policy"
@@ -64,6 +66,17 @@ func (a adminAPIHandlers) RemoveUser(w http.ResponseWriter, r *http.Request) {
 		return
 	}

+	// This API only supports removal of internal users not service accounts.
+	ok, _, err = globalIAMSys.IsServiceAccount(accessKey)
+	if err != nil {
+		writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
+		return
+	}
+	if ok {
+		writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, errIAMActionNotAllowed), r.URL)
+		return
+	}
+
 	// When the user is root credential you are not allowed to
 	// remove the root user. Also you cannot delete yourself.
 	if accessKey == globalActiveCred.AccessKey || accessKey == cred.AccessKey {
@@ -636,7 +649,7 @@ func (a adminAPIHandlers) TemporaryAccountInfo(w http.ResponseWriter, r *http.Re

 // AddServiceAccount - PUT /minio/admin/v3/add-service-account
 func (a adminAPIHandlers) AddServiceAccount(w http.ResponseWriter, r *http.Request) {
-	ctx, cred, opts, createReq, targetUser, APIError := commonAddServiceAccount(r)
+	ctx, cred, opts, createReq, targetUser, APIError := commonAddServiceAccount(r, false)
 	if APIError.Code != "" {
 		writeErrorResponseJSON(ctx, w, APIError, r.URL)
 		return
@@ -1166,6 +1179,172 @@ func (a adminAPIHandlers) DeleteServiceAccount(w http.ResponseWriter, r *http.Re
 	writeSuccessNoContent(w)
 }

+// ListAccessKeysBulk - GET /minio/admin/v3/list-access-keys-bulk
+func (a adminAPIHandlers) ListAccessKeysBulk(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
+	}
+
+	users := r.Form["users"]
+	isAll := r.Form.Get("all") == "true"
+	selfOnly := !isAll && len(users) == 0
+
+	if isAll && len(users) > 0 {
+		// This should be checked on client side, so return generic error
+		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrInvalidRequest), r.URL)
+		return
+	}
+
+	// Empty user 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(users) == 1 {
+		if users[0] == cred.AccessKey || users[0] == cred.ParentUser {
+			selfOnly = 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:        selfOnly,
+	}) {
+		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrAccessDenied), r.URL)
+		return
+	}
+
+	if selfOnly && len(users) == 0 {
+		selfUser := cred.AccessKey
+		if cred.ParentUser != "" {
+			selfUser = cred.ParentUser
+		}
+		users = append(users, selfUser)
+	}
+
+	var checkedUserList []string
+	if isAll {
+		users, err := globalIAMSys.ListUsers(ctx)
+		if err != nil {
+			writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
+			return
+		}
+		for user := range users {
+			checkedUserList = append(checkedUserList, user)
+		}
+		checkedUserList = append(checkedUserList, globalActiveCred.AccessKey)
+	} else {
+		for _, user := range users {
+			// Validate the user
+			_, ok := globalIAMSys.GetUser(ctx, user)
+			if !ok {
+				continue
+			}
+			checkedUserList = append(checkedUserList, user)
+		}
+	}
+
+	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.ListAccessKeysResp)
+	for _, user := range checkedUserList {
+		accessKeys := madmin.ListAccessKeysResp{}
+		if listSTSKeys {
+			stsKeys, err := globalIAMSys.ListSTSAccounts(ctx, user)
+			if err != nil {
+				writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
+				return
+			}
+			for _, sts := range stsKeys {
+				accessKeys.STSKeys = append(accessKeys.STSKeys, madmin.ServiceAccountInfo{
+					AccessKey:  sts.AccessKey,
+					Expiration: &sts.Expiration,
+				})
+			}
+			// if only STS keys, skip if user has no STS keys
+			if !listServiceAccounts && len(stsKeys) == 0 {
+				continue
+			}
+		}
+
+		if listServiceAccounts {
+			serviceAccounts, err := globalIAMSys.ListServiceAccounts(ctx, user)
+			if err != nil {
+				writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
+				return
+			}
+			for _, svc := range serviceAccounts {
+				accessKeys.ServiceAccounts = append(accessKeys.ServiceAccounts, madmin.ServiceAccountInfo{
+					AccessKey:  svc.AccessKey,
+					Expiration: &svc.Expiration,
+				})
+			}
+			// if only service accounts, skip if user has no service accounts
+			if !listSTSKeys && len(serviceAccounts) == 0 {
+				continue
+			}
+		}
+		accessKeyMap[user] = 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)
+}
+
 // AccountInfoHandler returns usage, permissions and other bucket metadata for incoming us
 func (a adminAPIHandlers) AccountInfoHandler(w http.ResponseWriter, r *http.Request) {
 	ctx := r.Context()
@@ -1235,18 +1414,6 @@ 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
@@ -1287,7 +1454,12 @@ func (a adminAPIHandlers) AccountInfoHandler(w http.ResponseWriter, r *http.Requ

 	var buf []byte
 	switch {
-	case accountName == globalActiveCred.AccessKey:
+	case accountName == globalActiveCred.AccessKey || newGlobalAuthZPluginFn() != nil:
+		// For owner account and when plugin authZ is configured always set
+		// effective policy as `consoleAdmin`.
+		//
+		// In the latter case, we let the UI render everything, but individual
+		// actions would fail if not permitted by the external authZ service.
 		for _, policy := range policy.DefaultPolicies {
 			if policy.Name == "consoleAdmin" {
 				effectivePolicy = policy.Definition
@@ -1333,15 +1505,12 @@ 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 {
@@ -1411,6 +1580,7 @@ func (a adminAPIHandlers) InfoCannedPolicy(w http.ResponseWriter, r *http.Reques
 		writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, errTooManyPolicies), r.URL)
 		return
 	}
+	setReqInfoPolicyName(ctx, name)

 	policyDoc, err := globalIAMSys.InfoPolicy(name)
 	if err != nil {
@@ -1514,6 +1684,7 @@ func (a adminAPIHandlers) RemoveCannedPolicy(w http.ResponseWriter, r *http.Requ

 	vars := mux.Vars(r)
 	policyName := vars["name"]
+	setReqInfoPolicyName(ctx, policyName)

 	if err := globalIAMSys.DeletePolicy(ctx, policyName, true); err != nil {
 		writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
@@ -1546,6 +1717,13 @@ func (a adminAPIHandlers) AddCannedPolicy(w http.ResponseWriter, r *http.Request
 		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrAdminResourceInvalidArgument), r.URL)
 		return
 	}
+	setReqInfoPolicyName(ctx, policyName)
+
+	// Reject policy names with commas.
+	if strings.Contains(policyName, ",") {
+		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrPolicyInvalidName), r.URL)
+		return
+	}

 	// Error out if Content-Length is missing.
 	if r.ContentLength <= 0 {
@@ -1611,6 +1789,7 @@ func (a adminAPIHandlers) SetPolicyForUserOrGroup(w http.ResponseWriter, r *http
 	policyName := vars["policyName"]
 	entityName := vars["userOrGroup"]
 	isGroup := vars["isGroup"] == "true"
+	setReqInfoPolicyName(ctx, policyName)

 	if !isGroup {
 		ok, _, err := globalIAMSys.IsTempUser(entityName)
@@ -1696,7 +1875,7 @@ func (a adminAPIHandlers) SetPolicyForUserOrGroup(w http.ResponseWriter, r *http
 	}))
 }

-// ListPolicyMappingEntities - GET /minio/admin/v3/idp/builtin/polciy-entities?policy=xxx&user=xxx&group=xxx
+// ListPolicyMappingEntities - GET /minio/admin/v3/idp/builtin/policy-entities?policy=xxx&user=xxx&group=xxx
 func (a adminAPIHandlers) ListPolicyMappingEntities(w http.ResponseWriter, r *http.Request) {
 	ctx := r.Context()

@@ -1798,6 +1977,7 @@ func (a adminAPIHandlers) AttachDetachPolicyBuiltin(w http.ResponseWriter, r *ht
 		writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
 		return
 	}
+	setReqInfoPolicyName(ctx, strings.Join(addedOrRemoved, ","))

 	respBody := madmin.PolicyAssociationResp{
 		UpdatedAt: updatedAt,
@@ -1852,6 +2032,7 @@ func (a adminAPIHandlers) ExportIAM(w http.ResponseWriter, r *http.Request) {
 	// Get current object layer instance.
 	objectAPI, _ := validateAdminReq(ctx, w, r, policy.ExportIAMAction)
 	if objectAPI == nil {
+		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrServerNotInitialized), r.URL)
 		return
 	}
 	// Initialize a zip writer which will provide a zipped content
@@ -2062,19 +2243,25 @@ func (a adminAPIHandlers) ExportIAM(w http.ResponseWriter, r *http.Request) {

 // ImportIAM - imports all IAM info into MinIO
 func (a adminAPIHandlers) ImportIAM(w http.ResponseWriter, r *http.Request) {
+	a.importIAM(w, r, "")
+}
+
+// ImportIAMV2 - imports all IAM info into MinIO
+func (a adminAPIHandlers) ImportIAMV2(w http.ResponseWriter, r *http.Request) {
+	a.importIAM(w, r, "v2")
+}
+
+// ImportIAM - imports all IAM info into MinIO
+func (a adminAPIHandlers) importIAM(w http.ResponseWriter, r *http.Request, apiVer string) {
 	ctx := r.Context()

-	// Get current object layer instance.
-	objectAPI := newObjectLayerFn()
+	// Validate signature, permissions and get current object layer instance.
+	objectAPI, _ := validateAdminReq(ctx, w, r, policy.ImportIAMAction)
 	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
-	}
+
 	data, err := io.ReadAll(r.Body)
 	if err != nil {
 		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrInvalidRequest), r.URL)
@@ -2086,6 +2273,10 @@ func (a adminAPIHandlers) ImportIAM(w http.ResponseWriter, r *http.Request) {
 		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrInvalidRequest), r.URL)
 		return
 	}
+
+	var skipped, removed, added madmin.IAMEntities
+	var failed madmin.IAMErrEntities
+
 	// import policies first
 	{

@@ -2111,8 +2302,10 @@ func (a adminAPIHandlers) ImportIAM(w http.ResponseWriter, r *http.Request) {
 			for policyName, policy := range allPolicies {
 				if policy.IsEmpty() {
 					err = globalIAMSys.DeletePolicy(ctx, policyName, true)
+					removed.Policies = append(removed.Policies, policyName)
 				} else {
 					_, err = globalIAMSys.SetPolicy(ctx, policyName, policy)
+					added.Policies = append(added.Policies, policyName)
 				}
 				if err != nil {
 					writeErrorResponseJSON(ctx, w, importError(ctx, err, allPoliciesFile, policyName), r.URL)
@@ -2158,41 +2351,16 @@ func (a adminAPIHandlers) ImportIAM(w http.ResponseWriter, r *http.Request) {
 					return
 				}

-				if (cred.IsTemp() || cred.IsServiceAccount()) && cred.ParentUser == accessKey {
-					// Incoming access key matches parent user then we should
-					// reject password change requests.
-					writeErrorResponseJSON(ctx, w, importErrorWithAPIErr(ctx, ErrAddUserInvalidArgument, err, allUsersFile, accessKey), r.URL)
-					return
-				}
-
 				// Check if accessKey has beginning and end space characters, this only applies to new users.
 				if !exists && hasSpaceBE(accessKey) {
 					writeErrorResponseJSON(ctx, w, importErrorWithAPIErr(ctx, ErrAdminResourceInvalidArgument, err, allUsersFile, accessKey), r.URL)
 					return
 				}

-				checkDenyOnly := false
-				if accessKey == cred.AccessKey {
-					// Check that there is no explicit deny - otherwise it's allowed
-					// to change one's own password.
-					checkDenyOnly = true
-				}
-
-				if !globalIAMSys.IsAllowed(policy.Args{
-					AccountName:     cred.AccessKey,
-					Groups:          cred.Groups,
-					Action:          policy.CreateUserAdminAction,
-					ConditionValues: getConditionValues(r, "", cred),
-					IsOwner:         owner,
-					Claims:          cred.Claims,
-					DenyOnly:        checkDenyOnly,
-				}) {
-					writeErrorResponseJSON(ctx, w, importErrorWithAPIErr(ctx, ErrAccessDenied, err, allUsersFile, accessKey), r.URL)
-					return
-				}
 				if _, err = globalIAMSys.CreateUser(ctx, accessKey, ureq); err != nil {
-					writeErrorResponseJSON(ctx, w, importErrorWithAPIErr(ctx, toAdminAPIErrCode(ctx, err), err, allUsersFile, accessKey), r.URL)
-					return
+					failed.Users = append(failed.Users, madmin.IAMErrEntity{Name: accessKey, Error: err})
+				} else {
+					added.Users = append(added.Users, accessKey)
 				}

 			}
@@ -2230,8 +2398,9 @@ func (a adminAPIHandlers) ImportIAM(w http.ResponseWriter, r *http.Request) {
 					}
 				}
 				if _, gerr := globalIAMSys.AddUsersToGroup(ctx, group, grpInfo.Members); gerr != nil {
-					writeErrorResponseJSON(ctx, w, importError(ctx, gerr, allGroupsFile, group), r.URL)
-					return
+					failed.Groups = append(failed.Groups, madmin.IAMErrEntity{Name: group, Error: err})
+				} else {
+					added.Groups = append(added.Groups, group)
 				}
 			}
 		}
@@ -2260,7 +2429,8 @@ func (a adminAPIHandlers) ImportIAM(w http.ResponseWriter, r *http.Request) {

 			// Validations for LDAP enabled deployments.
 			if globalIAMSys.LDAPConfig.Enabled() {
-				err := globalIAMSys.NormalizeLDAPAccessKeypairs(ctx, serviceAcctReqs)
+				skippedAccessKeys, err := globalIAMSys.NormalizeLDAPAccessKeypairs(ctx, serviceAcctReqs)
+				skipped.ServiceAccounts = append(skipped.ServiceAccounts, skippedAccessKeys...)
 				if err != nil {
 					writeErrorResponseJSON(ctx, w, importError(ctx, err, allSvcAcctsFile, ""), r.URL)
 					return
@@ -2268,6 +2438,9 @@ func (a adminAPIHandlers) ImportIAM(w http.ResponseWriter, r *http.Request) {
 			}

 			for user, svcAcctReq := range serviceAcctReqs {
+				if slices.Contains(skipped.ServiceAccounts, user) {
+					continue
+				}
 				var sp *policy.Policy
 				var err error
 				if len(svcAcctReq.SessionPolicy) > 0 {
@@ -2283,17 +2456,6 @@ func (a adminAPIHandlers) ImportIAM(w http.ResponseWriter, r *http.Request) {
 					writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrAdminResourceInvalidArgument), r.URL)
 					return
 				}
-				if !globalIAMSys.IsAllowed(policy.Args{
-					AccountName:     cred.AccessKey,
-					Groups:          cred.Groups,
-					Action:          policy.CreateServiceAccountAdminAction,
-					ConditionValues: getConditionValues(r, "", cred),
-					IsOwner:         owner,
-					Claims:          cred.Claims,
-				}) {
-					writeErrorResponseJSON(ctx, w, importErrorWithAPIErr(ctx, ErrAccessDenied, err, allSvcAcctsFile, user), r.URL)
-					return
-				}
 				updateReq := true
 				_, _, err = globalIAMSys.GetServiceAccount(ctx, svcAcctReq.AccessKey)
 				if err != nil {
@@ -2308,7 +2470,7 @@ 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
 					}
@@ -2325,10 +2487,10 @@ func (a adminAPIHandlers) ImportIAM(w http.ResponseWriter, r *http.Request) {
 				}

 				if _, _, err = globalIAMSys.NewServiceAccount(ctx, svcAcctReq.Parent, svcAcctReq.Groups, opts); err != nil {
-					writeErrorResponseJSON(ctx, w, importError(ctx, err, allSvcAcctsFile, user), r.URL)
-					return
+					failed.ServiceAccounts = append(failed.ServiceAccounts, madmin.IAMErrEntity{Name: user, Error: err})
+				} else {
+					added.ServiceAccounts = append(added.ServiceAccounts, user)
 				}
-
 			}
 		}
 	}
@@ -2365,8 +2527,15 @@ func (a adminAPIHandlers) ImportIAM(w http.ResponseWriter, r *http.Request) {
 					return
 				}
 				if _, err := globalIAMSys.PolicyDBSet(ctx, u, pm.Policies, regUser, false); err != nil {
-					writeErrorResponseJSON(ctx, w, importError(ctx, err, userPolicyMappingsFile, u), r.URL)
-					return
+					failed.UserPolicies = append(
+						failed.UserPolicies,
+						madmin.IAMErrPolicyEntity{
+							Name:     u,
+							Policies: strings.Split(pm.Policies, ","),
+							Error:    err,
+						})
+				} else {
+					added.UserPolicies = append(added.UserPolicies, map[string][]string{u: strings.Split(pm.Policies, ",")})
 				}
 			}
 		}
@@ -2396,7 +2565,8 @@ func (a adminAPIHandlers) ImportIAM(w http.ResponseWriter, r *http.Request) {
 			// Validations for LDAP enabled deployments.
 			if globalIAMSys.LDAPConfig.Enabled() {
 				isGroup := true
-				err := globalIAMSys.NormalizeLDAPMappingImport(ctx, isGroup, grpPolicyMap)
+				skippedDN, err := globalIAMSys.NormalizeLDAPMappingImport(ctx, isGroup, grpPolicyMap)
+				skipped.Groups = append(skipped.Groups, skippedDN...)
 				if err != nil {
 					writeErrorResponseJSON(ctx, w, importError(ctx, err, groupPolicyMappingsFile, ""), r.URL)
 					return
@@ -2404,9 +2574,19 @@ func (a adminAPIHandlers) ImportIAM(w http.ResponseWriter, r *http.Request) {
 			}

 			for g, pm := range grpPolicyMap {
+				if slices.Contains(skipped.Groups, g) {
+					continue
+				}
 				if _, err := globalIAMSys.PolicyDBSet(ctx, g, pm.Policies, unknownIAMUserType, true); err != nil {
-					writeErrorResponseJSON(ctx, w, importError(ctx, err, groupPolicyMappingsFile, g), r.URL)
-					return
+					failed.GroupPolicies = append(
+						failed.GroupPolicies,
+						madmin.IAMErrPolicyEntity{
+							Name:     g,
+							Policies: strings.Split(pm.Policies, ","),
+							Error:    err,
+						})
+				} else {
+					added.GroupPolicies = append(added.GroupPolicies, map[string][]string{g: strings.Split(pm.Policies, ",")})
 				}
 			}
 		}
@@ -2436,13 +2616,17 @@ func (a adminAPIHandlers) ImportIAM(w http.ResponseWriter, r *http.Request) {
 			// Validations for LDAP enabled deployments.
 			if globalIAMSys.LDAPConfig.Enabled() {
 				isGroup := true
-				err := globalIAMSys.NormalizeLDAPMappingImport(ctx, !isGroup, userPolicyMap)
+				skippedDN, err := globalIAMSys.NormalizeLDAPMappingImport(ctx, !isGroup, userPolicyMap)
+				skipped.Users = append(skipped.Users, skippedDN...)
 				if err != nil {
 					writeErrorResponseJSON(ctx, w, importError(ctx, err, stsUserPolicyMappingsFile, ""), r.URL)
 					return
 				}
 			}
 			for u, pm := range userPolicyMap {
+				if slices.Contains(skipped.Users, u) {
+					continue
+				}
 				// disallow setting policy mapping if user is a temporary user
 				ok, _, err := globalIAMSys.IsTempUser(u)
 				if err != nil && err != errNoSuchUser {
@@ -2455,12 +2639,36 @@ func (a adminAPIHandlers) ImportIAM(w http.ResponseWriter, r *http.Request) {
 				}

 				if _, err := globalIAMSys.PolicyDBSet(ctx, u, pm.Policies, stsUser, false); err != nil {
-					writeErrorResponseJSON(ctx, w, importError(ctx, err, stsUserPolicyMappingsFile, u), r.URL)
-					return
+					failed.STSPolicies = append(
+						failed.STSPolicies,
+						madmin.IAMErrPolicyEntity{
+							Name:     u,
+							Policies: strings.Split(pm.Policies, ","),
+							Error:    err,
+						})
+				} else {
+					added.STSPolicies = append(added.STSPolicies, map[string][]string{u: strings.Split(pm.Policies, ",")})
 				}
 			}
 		}
 	}
+
+	if apiVer == "v2" {
+		iamr := madmin.ImportIAMResult{
+			Skipped: skipped,
+			Removed: removed,
+			Added:   added,
+			Failed:  failed,
+		}
+
+		b, err := json.Marshal(iamr)
+		if err != nil {
+			writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
+			return
+		}
+
+		writeSuccessResponseJSON(w, b)
+	}
 }

 func addExpirationToCondValues(exp *time.Time, condValues map[string][]string) error {
@@ -2475,7 +2683,7 @@ func addExpirationToCondValues(exp *time.Time, condValues map[string][]string) e
 	return nil
 }

-func commonAddServiceAccount(r *http.Request) (context.Context, auth.Credentials, newServiceAccountOpts, madmin.AddServiceAccountReq, string, APIError) {
+func commonAddServiceAccount(r *http.Request, ldap bool) (context.Context, auth.Credentials, newServiceAccountOpts, madmin.AddServiceAccountReq, string, APIError) {
 	ctx := r.Context()

 	// Get current object layer instance.
@@ -2542,6 +2750,14 @@ func commonAddServiceAccount(r *http.Request) (context.Context, auth.Credentials
 		return ctx, auth.Credentials{}, newServiceAccountOpts{}, madmin.AddServiceAccountReq{}, "", toAdminAPIErr(ctx, err)
 	}

+	denyOnly := (targetUser == cred.AccessKey || targetUser == cred.ParentUser)
+	if ldap && !denyOnly {
+		res, _ := globalIAMSys.LDAPConfig.GetValidatedDNForUsername(targetUser)
+		if res.NormDN == cred.ParentUser {
+			denyOnly = true
+		}
+	}
+
 	// Check if action is allowed if creating access key for another user
 	// Check if action is explicitly denied if for self
 	if !globalIAMSys.IsAllowed(policy.Args{
@@ -2551,7 +2767,7 @@ func commonAddServiceAccount(r *http.Request) (context.Context, auth.Credentials
 		ConditionValues: condValues,
 		IsOwner:         owner,
 		Claims:          cred.Claims,
-		DenyOnly:        (targetUser == cred.AccessKey || targetUser == cred.ParentUser),
+		DenyOnly:        denyOnly,
 	}) {
 		return ctx, auth.Credentials{}, newServiceAccountOpts{}, madmin.AddServiceAccountReq{}, "", errorCodes.ToAPIErr(ErrAccessDenied)
 	}
@@ -2568,3 +2784,10 @@ func commonAddServiceAccount(r *http.Request) (context.Context, auth.Credentials

 	return ctx, cred, opts, createReq, targetUser, APIError{}
 }
+
+// setReqInfoPolicyName will set the given policyName as a tag on the context's request info,
+// so that it appears in audit logs.
+func setReqInfoPolicyName(ctx context.Context, policyName string) {
+	reqInfo := logger.GetReqInfo(ctx)
+	reqInfo.SetTags("policyName", policyName)
+}
@@ -28,6 +28,7 @@ import (
 	"net/http"
 	"net/url"
 	"runtime"
+	"slices"
 	"strings"
 	"testing"
 	"time"
@@ -337,16 +338,24 @@ func (s *TestSuiteIAM) TestUserPolicyEscalationBug(c *check) {
  {
   "Effect": "Allow",
   "Action": [
-    "s3:PutObject",
-    "s3:GetObject",
    "s3:ListBucket"
   ],
+   "Resource": [
+    "arn:aws:s3:::%s"
+   ]
+  },
+  {
+   "Effect": "Allow",
+   "Action": [
+    "s3:PutObject",
+    "s3:GetObject"
+   ],
   "Resource": [
    "arn:aws:s3:::%s/*"
   ]
  }
 ]
-}`, bucket))
+}`, bucket, bucket))
 	err = s.adm.AddCannedPolicy(ctx, policy, policyBytes)
 	if err != nil {
 		c.Fatalf("policy add error: %v", err)
@@ -442,7 +451,7 @@ func (s *TestSuiteIAM) TestAddServiceAccountPerms(c *check) {
    "s3:ListBucket"
   ],
   "Resource": [
-    "arn:aws:s3:::testbucket/*"
+    "arn:aws:s3:::testbucket"
   ]
  }
 ]
@@ -559,16 +568,24 @@ func (s *TestSuiteIAM) TestPolicyCreate(c *check) {
  {
   "Effect": "Allow",
   "Action": [
-    "s3:PutObject",
-    "s3:GetObject",
    "s3:ListBucket"
   ],
+   "Resource": [
+    "arn:aws:s3:::%s"
+   ]
+  },
+  {
+   "Effect": "Allow",
+   "Action": [
+    "s3:PutObject",
+    "s3:GetObject"
+   ],
   "Resource": [
    "arn:aws:s3:::%s/*"
   ]
  }
 ]
-}`, bucket))
+}`, bucket, bucket))
 	err = s.adm.AddCannedPolicy(ctx, policy, policyBytes)
 	if err != nil {
 		c.Fatalf("policy add error: %v", err)
@@ -669,16 +686,24 @@ func (s *TestSuiteIAM) TestCannedPolicies(c *check) {
  {
   "Effect": "Allow",
   "Action": [
-    "s3:PutObject",
-    "s3:GetObject",
    "s3:ListBucket"
   ],
+   "Resource": [
+    "arn:aws:s3:::%s"
+   ]
+  },
+  {
+   "Effect": "Allow",
+   "Action": [
+    "s3:PutObject",
+    "s3:GetObject"
+   ],
   "Resource": [
    "arn:aws:s3:::%s/*"
   ]
  }
 ]
-}`, bucket))
+}`, bucket, bucket))

 	// Check that default policies can be overwritten.
 	err = s.adm.AddCannedPolicy(ctx, "readwrite", policyBytes)
@@ -691,6 +716,12 @@ func (s *TestSuiteIAM) TestCannedPolicies(c *check) {
 		c.Fatalf("policy info err: %v", err)
 	}

+	// Check that policy with comma is rejected.
+	err = s.adm.AddCannedPolicy(ctx, "invalid,policy", policyBytes)
+	if err == nil {
+		c.Fatalf("invalid policy created successfully")
+	}
+
 	infoStr := string(info)
 	if !strings.Contains(infoStr, `"s3:PutObject"`) || !strings.Contains(infoStr, ":"+bucket+"/") {
 		c.Fatalf("policy contains unexpected content!")
@@ -714,16 +745,24 @@ func (s *TestSuiteIAM) TestGroupAddRemove(c *check) {
  {
   "Effect": "Allow",
   "Action": [
-    "s3:PutObject",
-    "s3:GetObject",
    "s3:ListBucket"
   ],
+   "Resource": [
+    "arn:aws:s3:::%s"
+   ]
+  },
+  {
+   "Effect": "Allow",
+   "Action": [
+    "s3:PutObject",
+    "s3:GetObject"
+   ],
   "Resource": [
    "arn:aws:s3:::%s/*"
   ]
  }
 ]
-}`, bucket))
+}`, bucket, bucket))
 	err = s.adm.AddCannedPolicy(ctx, policy, policyBytes)
 	if err != nil {
 		c.Fatalf("policy add error: %v", err)
@@ -770,8 +809,9 @@ func (s *TestSuiteIAM) TestGroupAddRemove(c *check) {
 	if err != nil {
 		c.Fatalf("group list err: %v", err)
 	}
-	if !set.CreateStringSet(groups...).Contains(group) {
-		c.Fatalf("created group not present!")
+	expected := []string{group}
+	if !slices.Equal(groups, expected) {
+		c.Fatalf("expected group listing: %v, got: %v", expected, groups)
 	}
 	groupInfo, err := s.adm.GetGroupDescription(ctx, group)
 	if err != nil {
@@ -877,16 +917,24 @@ func (s *TestSuiteIAM) TestServiceAccountOpsByUser(c *check) {
  {
   "Effect": "Allow",
   "Action": [
-    "s3:PutObject",
-    "s3:GetObject",
    "s3:ListBucket"
   ],
+   "Resource": [
+    "arn:aws:s3:::%s"
+   ]
+  },
+  {
+   "Effect": "Allow",
+   "Action": [
+    "s3:PutObject",
+    "s3:GetObject"
+   ],
   "Resource": [
    "arn:aws:s3:::%s/*"
   ]
  }
 ]
-}`, bucket))
+}`, bucket, bucket))
 	err = s.adm.AddCannedPolicy(ctx, policy, policyBytes)
 	if err != nil {
 		c.Fatalf("policy add error: %v", err)
@@ -961,16 +1009,24 @@ func (s *TestSuiteIAM) TestServiceAccountDurationSecondsCondition(c *check) {
  {
   "Effect": "Allow",
   "Action": [
-    "s3:PutObject",
-    "s3:GetObject",
    "s3:ListBucket"
   ],
+   "Resource": [
+    "arn:aws:s3:::%s"
+   ]
+  },
+  {
+   "Effect": "Allow",
+   "Action": [
+    "s3:PutObject",
+    "s3:GetObject"
+   ],
   "Resource": [
    "arn:aws:s3:::%s/*"
   ]
  }
 ]
-}`, bucket))
+}`, bucket, bucket))
 	err = s.adm.AddCannedPolicy(ctx, policy, policyBytes)
 	if err != nil {
 		c.Fatalf("policy add error: %v", err)
@@ -1043,16 +1099,24 @@ func (s *TestSuiteIAM) TestServiceAccountOpsByAdmin(c *check) {
  {
   "Effect": "Allow",
   "Action": [
-    "s3:PutObject",
-    "s3:GetObject",
    "s3:ListBucket"
   ],
+   "Resource": [
+    "arn:aws:s3:::%s"
+   ]
+  },
+  {
+   "Effect": "Allow",
+   "Action": [
+    "s3:PutObject",
+    "s3:GetObject"
+   ],
   "Resource": [
    "arn:aws:s3:::%s/*"
   ]
  }
 ]
-}`, bucket))
+}`, bucket, bucket))
 	err = s.adm.AddCannedPolicy(ctx, policy, policyBytes)
 	if err != nil {
 		c.Fatalf("policy add error: %v", err)
@@ -1600,7 +1664,7 @@ func (c *check) assertSvcAccSessionPolicyUpdate(ctx context.Context, s *TestSuit
    "s3:ListBucket"
   ],
   "Resource": [
-    "arn:aws:s3:::%s/*"
+    "arn:aws:s3:::%s"
   ]
  }
 ]
@@ -93,11 +93,18 @@ func (a adminAPIHandlers) ServerUpdateV2Handler(w http.ResponseWriter, r *http.R
 		return
 	}

-	if globalInplaceUpdateDisabled || currentReleaseTime.IsZero() {
+	if globalInplaceUpdateDisabled {
 		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrMethodNotAllowed), r.URL)
 		return
 	}

+	if currentReleaseTime.IsZero() || currentReleaseTime.Equal(timeSentinel) {
+		apiErr := errorCodes.ToAPIErr(ErrMethodNotAllowed)
+		apiErr.Description = fmt.Sprintf("unable to perform in-place update, release time is unrecognized: %s", currentReleaseTime)
+		writeErrorResponseJSON(ctx, w, apiErr, r.URL)
+		return
+	}
+
 	vars := mux.Vars(r)
 	updateURL := vars["updateURL"]
 	dryRun := r.Form.Get("dry-run") == "true"
@@ -110,6 +117,11 @@ func (a adminAPIHandlers) ServerUpdateV2Handler(w http.ResponseWriter, r *http.R
 		}
 	}

+	local := globalLocalNodeName
+	if local == "" {
+		local = "127.0.0.1"
+	}
+
 	u, err := url.Parse(updateURL)
 	if err != nil {
 		writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
@@ -128,6 +140,39 @@ func (a adminAPIHandlers) ServerUpdateV2Handler(w http.ResponseWriter, r *http.R
 		return
 	}

+	updateStatus := madmin.ServerUpdateStatusV2{
+		DryRun:  dryRun,
+		Results: make([]madmin.ServerPeerUpdateStatus, 0, len(globalNotificationSys.allPeerClients)),
+	}
+	peerResults := make(map[string]madmin.ServerPeerUpdateStatus, len(globalNotificationSys.allPeerClients))
+	failedClients := make(map[int]bool, len(globalNotificationSys.allPeerClients))
+
+	if lrTime.Sub(currentReleaseTime) <= 0 {
+		updateStatus.Results = append(updateStatus.Results, madmin.ServerPeerUpdateStatus{
+			Host:           local,
+			Err:            fmt.Sprintf("server is running the latest version: %s", Version),
+			CurrentVersion: Version,
+		})
+
+		for _, client := range globalNotificationSys.peerClients {
+			updateStatus.Results = append(updateStatus.Results, madmin.ServerPeerUpdateStatus{
+				Host:           client.String(),
+				Err:            fmt.Sprintf("server is running the latest version: %s", Version),
+				CurrentVersion: Version,
+			})
+		}
+
+		// Marshal API response
+		jsonBytes, err := json.Marshal(updateStatus)
+		if err != nil {
+			writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
+			return
+		}
+
+		writeSuccessResponseJSON(w, jsonBytes)
+		return
+	}
+
 	u.Path = path.Dir(u.Path) + SlashSeparator + releaseInfo
 	// Download Binary Once
 	binC, bin, err := downloadBinary(u, mode)
@@ -137,16 +182,6 @@ func (a adminAPIHandlers) ServerUpdateV2Handler(w http.ResponseWriter, r *http.R
 		return
 	}

-	updateStatus := madmin.ServerUpdateStatusV2{DryRun: dryRun}
-	peerResults := make(map[string]madmin.ServerPeerUpdateStatus)
-
-	local := globalLocalNodeName
-	if local == "" {
-		local = "127.0.0.1"
-	}
-
-	failedClients := make(map[int]struct{})
-
 	if globalIsDistErasure {
 		// Push binary to other servers
 		for idx, nerr := range globalNotificationSys.VerifyBinary(ctx, u, sha256Sum, releaseInfo, binC) {
@@ -156,7 +191,7 @@ func (a adminAPIHandlers) ServerUpdateV2Handler(w http.ResponseWriter, r *http.R
 					Err:            nerr.Err.Error(),
 					CurrentVersion: Version,
 				}
-				failedClients[idx] = struct{}{}
+				failedClients[idx] = true
 			} else {
 				peerResults[nerr.Host.String()] = madmin.ServerPeerUpdateStatus{
 					Host:           nerr.Host.String(),
@@ -167,25 +202,17 @@ func (a adminAPIHandlers) ServerUpdateV2Handler(w http.ResponseWriter, r *http.R
 		}
 	}

-	if lrTime.Sub(currentReleaseTime) > 0 {
-		if err = verifyBinary(u, sha256Sum, releaseInfo, mode, bytes.NewReader(bin)); err != nil {
-			peerResults[local] = madmin.ServerPeerUpdateStatus{
-				Host:           local,
-				Err:            err.Error(),
-				CurrentVersion: Version,
-			}
-		} else {
-			peerResults[local] = madmin.ServerPeerUpdateStatus{
-				Host:           local,
-				CurrentVersion: Version,
-				UpdatedVersion: lrTime.Format(MinioReleaseTagTimeLayout),
-			}
+	if err = verifyBinary(u, sha256Sum, releaseInfo, mode, bytes.NewReader(bin)); err != nil {
+		peerResults[local] = madmin.ServerPeerUpdateStatus{
+			Host:           local,
+			Err:            err.Error(),
+			CurrentVersion: Version,
 		}
 	} else {
 		peerResults[local] = madmin.ServerPeerUpdateStatus{
 			Host:           local,
-			Err:            fmt.Sprintf("server is already running the latest version: %s", Version),
 			CurrentVersion: Version,
+			UpdatedVersion: lrTime.Format(MinioReleaseTagTimeLayout),
 		}
 	}

@@ -193,8 +220,7 @@ func (a adminAPIHandlers) ServerUpdateV2Handler(w http.ResponseWriter, r *http.R
 		if globalIsDistErasure {
 			ng := WithNPeers(len(globalNotificationSys.peerClients))
 			for idx, client := range globalNotificationSys.peerClients {
-				_, ok := failedClients[idx]
-				if ok {
+				if failedClients[idx] {
 					continue
 				}
 				client := client
@@ -240,14 +266,14 @@ func (a adminAPIHandlers) ServerUpdateV2Handler(w http.ResponseWriter, r *http.R
 		startTime := time.Now().Add(restartUpdateDelay)
 		ng := WithNPeers(len(globalNotificationSys.peerClients))
 		for idx, client := range globalNotificationSys.peerClients {
-			_, ok := failedClients[idx]
-			if ok {
+			if failedClients[idx] {
 				continue
 			}
 			client := client
 			ng.Go(ctx, func() error {
 				prs, ok := peerResults[client.String()]
-				if ok && prs.CurrentVersion != prs.UpdatedVersion && prs.UpdatedVersion != "" {
+				// We restart only on success, not for any failures.
+				if ok && prs.Err == "" {
 					return client.SignalService(serviceRestart, "", dryRun, &startTime)
 				}
 				return nil
@@ -284,7 +310,9 @@ func (a adminAPIHandlers) ServerUpdateV2Handler(w http.ResponseWriter, r *http.R
 	writeSuccessResponseJSON(w, jsonBytes)

 	if !dryRun {
-		if lrTime.Sub(currentReleaseTime) > 0 {
+		prs, ok := peerResults[local]
+		// We restart only on success, not for any failures.
+		if ok && prs.Err == "" {
 			globalServiceSignalCh <- serviceRestart
 		}
 	}
@@ -848,9 +876,10 @@ 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,
@@ -1035,7 +1064,7 @@ 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)
@@ -1116,7 +1145,11 @@ 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)
@@ -1131,6 +1164,10 @@ 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 {
@@ -2186,7 +2223,7 @@ 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()

@@ -2330,6 +2367,7 @@ 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()
@@ -2370,7 +2408,9 @@ 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}
@@ -2404,18 +2444,24 @@ 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),
@@ -3049,7 +3095,7 @@ 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
@@ -463,6 +463,7 @@ func TestTopLockEntries(t *testing.T) {
 			Owner:      lri.Owner,
 			ID:         lri.UID,
 			Quorum:     lri.Quorum,
+			Timestamp:  time.Unix(0, lri.Timestamp),
 		})
 	}

@@ -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
@@ -792,18 +806,20 @@ func (h *healSequence) healDiskMeta(objAPI ObjectLayer) error {
 	return h.healMinioSysMeta(objAPI, minioConfigPrefix)()
 }

-func (h *healSequence) healItems(objAPI ObjectLayer, bucketsOnly bool) error {
+func (h *healSequence) healItems(objAPI ObjectLayer) error {
 	if h.clientToken == bgHealingUUID {
 		// For background heal do nothing.
 		return nil
 	}

-	if err := h.healDiskMeta(objAPI); err != nil {
-		return err
+	if h.bucket == "" { // heal internal meta only during a site-wide heal
+		if err := h.healDiskMeta(objAPI); err != nil {
+			return err
+		}
 	}

 	// Heal buckets and objects
-	return h.healBuckets(objAPI, bucketsOnly)
+	return h.healBuckets(objAPI)
 }

 // traverseAndHeal - traverses on-disk data and performs healing
@@ -814,8 +830,7 @@ func (h *healSequence) healItems(objAPI ObjectLayer, bucketsOnly bool) error {
 // has to wait until a safe point is reached, such as between scanning
 // two objects.
 func (h *healSequence) traverseAndHeal(objAPI ObjectLayer) {
-	bucketsOnly := false // Heals buckets and objects also.
-	h.traverseAndHealDoneCh <- h.healItems(objAPI, bucketsOnly)
+	h.traverseAndHealDoneCh <- h.healItems(objAPI)
 	xioutil.SafeClose(h.traverseAndHealDoneCh)
 }

@@ -826,6 +841,7 @@ func (h *healSequence) healMinioSysMeta(objAPI ObjectLayer, metaPrefix string) f
 		// NOTE: Healing on meta is run regardless
 		// of any bucket being selected, this is to ensure that
 		// meta are always upto date and correct.
+		h.settings.Recursive = true
 		return objAPI.HealObjects(h.ctx, minioMetaBucket, metaPrefix, h.settings, func(bucket, object, versionID string, scanMode madmin.HealScanMode) error {
 			if h.isQuitting() {
 				return errHealStopSignalled
@@ -842,14 +858,14 @@ func (h *healSequence) healMinioSysMeta(objAPI ObjectLayer, metaPrefix string) f
 }

 // healBuckets - check for all buckets heal or just particular bucket.
-func (h *healSequence) healBuckets(objAPI ObjectLayer, bucketsOnly bool) error {
+func (h *healSequence) healBuckets(objAPI ObjectLayer) error {
 	if h.isQuitting() {
 		return errHealStopSignalled
 	}

 	// 1. If a bucket was specified, heal only the bucket.
 	if h.bucket != "" {
-		return h.healBucket(objAPI, h.bucket, bucketsOnly)
+		return h.healBucket(objAPI, h.bucket, false)
 	}

 	buckets, err := objAPI.ListBuckets(h.ctx, BucketOptions{})
@@ -863,7 +879,7 @@ func (h *healSequence) healBuckets(objAPI ObjectLayer, bucketsOnly bool) error {
 	})

 	for _, bucket := range buckets {
-		if err = h.healBucket(objAPI, bucket.Name, bucketsOnly); err != nil {
+		if err = h.healBucket(objAPI, bucket.Name, false); err != nil {
 			return err
 		}
 	}
@@ -881,16 +897,6 @@ func (h *healSequence) healBucket(objAPI ObjectLayer, bucket string, bucketsOnly
 		return nil
 	}

-	if !h.settings.Recursive {
-		if h.object != "" {
-			if err := h.healObject(bucket, h.object, "", h.settings.ScanMode); err != nil {
-				return err
-			}
-		}
-
-		return nil
-	}
-
 	if err := objAPI.HealObjects(h.ctx, bucket, h.object, h.settings, h.healObject); err != nil {
 		return errFnHealFromAPIErr(h.ctx, err)
 	}
@@ -159,14 +159,14 @@ func registerAdminRouter(router *mux.Router, enableConfigOps bool) {

 		// Info operations
 		adminRouter.Methods(http.MethodGet).Path(adminVersion + "/info").HandlerFunc(adminMiddleware(adminAPI.ServerInfoHandler, traceAllFlag, noObjLayerFlag))
-		adminRouter.Methods(http.MethodGet, http.MethodPost).Path(adminVersion + "/inspect-data").HandlerFunc(adminMiddleware(adminAPI.InspectDataHandler, noGZFlag, traceAllFlag))
+		adminRouter.Methods(http.MethodGet, http.MethodPost).Path(adminVersion + "/inspect-data").HandlerFunc(adminMiddleware(adminAPI.InspectDataHandler, noGZFlag, traceHdrsS3HFlag))

 		// StorageInfo operations
 		adminRouter.Methods(http.MethodGet).Path(adminVersion + "/storageinfo").HandlerFunc(adminMiddleware(adminAPI.StorageInfoHandler, traceAllFlag))
 		// DataUsageInfo operations
 		adminRouter.Methods(http.MethodGet).Path(adminVersion + "/datausageinfo").HandlerFunc(adminMiddleware(adminAPI.DataUsageInfoHandler, traceAllFlag))
 		// Metrics operation
-		adminRouter.Methods(http.MethodGet).Path(adminVersion + "/metrics").HandlerFunc(adminMiddleware(adminAPI.MetricsHandler, traceAllFlag))
+		adminRouter.Methods(http.MethodGet).Path(adminVersion + "/metrics").HandlerFunc(adminMiddleware(adminAPI.MetricsHandler, traceHdrsS3HFlag))

 		if globalIsDistErasure || globalIsErasure {
 			// Heal operations
@@ -193,9 +193,9 @@ func registerAdminRouter(router *mux.Router, enableConfigOps bool) {
 		// Profiling operations - deprecated API
 		adminRouter.Methods(http.MethodPost).Path(adminVersion+"/profiling/start").HandlerFunc(adminMiddleware(adminAPI.StartProfilingHandler, traceAllFlag, noObjLayerFlag)).
 			Queries("profilerType", "{profilerType:.*}")
-		adminRouter.Methods(http.MethodGet).Path(adminVersion + "/profiling/download").HandlerFunc(adminMiddleware(adminAPI.DownloadProfilingHandler, traceAllFlag, noObjLayerFlag))
+		adminRouter.Methods(http.MethodGet).Path(adminVersion + "/profiling/download").HandlerFunc(adminMiddleware(adminAPI.DownloadProfilingHandler, traceHdrsS3HFlag, noObjLayerFlag))
 		// Profiling operations
-		adminRouter.Methods(http.MethodPost).Path(adminVersion + "/profile").HandlerFunc(adminMiddleware(adminAPI.ProfileHandler, traceAllFlag, noObjLayerFlag))
+		adminRouter.Methods(http.MethodPost).Path(adminVersion + "/profile").HandlerFunc(adminMiddleware(adminAPI.ProfileHandler, traceHdrsS3HFlag, noObjLayerFlag))

 		// Config KV operations.
 		if enableConfigOps {
@@ -244,6 +244,9 @@ func registerAdminRouter(router *mux.Router, enableConfigOps bool) {
 		// STS accounts ops
 		adminRouter.Methods(http.MethodGet).Path(adminVersion+"/temporary-account-info").HandlerFunc(adminMiddleware(adminAPI.TemporaryAccountInfo)).Queries("accessKey", "{accessKey:.*}")

+		// Access key (service account/STS) operations
+		adminRouter.Methods(http.MethodGet).Path(adminVersion+"/list-access-keys-bulk").HandlerFunc(adminMiddleware(adminAPI.ListAccessKeysBulk)).Queries("listType", "{listType:.*}")
+
 		// Info policy IAM latest
 		adminRouter.Methods(http.MethodGet).Path(adminVersion+"/info-canned-policy").HandlerFunc(adminMiddleware(adminAPI.InfoCannedPolicy)).Queries("name", "{name:.*}")
 		// List policies latest
@@ -290,6 +293,7 @@ func registerAdminRouter(router *mux.Router, enableConfigOps bool) {

 		// Import IAM info
 		adminRouter.Methods(http.MethodPut).Path(adminVersion + "/import-iam").HandlerFunc(adminMiddleware(adminAPI.ImportIAM, noGZFlag))
+		adminRouter.Methods(http.MethodPut).Path(adminVersion + "/import-iam-v2").HandlerFunc(adminMiddleware(adminAPI.ImportIAMV2, noGZFlag))

 		// IDentity Provider configuration APIs
 		adminRouter.Methods(http.MethodPut).Path(adminVersion + "/idp-config/{type}/{name}").HandlerFunc(adminMiddleware(adminAPI.AddIdentityProviderCfg))
@@ -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"

@@ -213,6 +213,7 @@ const (
 	ErrPolicyAlreadyAttached
 	ErrPolicyNotAttached
 	ErrExcessData
+	ErrPolicyInvalidName
 	// Add new error codes here.

 	// SSE-S3/SSE-KMS related API errors
@@ -561,6 +562,11 @@ var errorCodes = errorCodeMap{
 		Description:    "More data provided than indicated content length",
 		HTTPStatusCode: http.StatusBadRequest,
 	},
+	ErrPolicyInvalidName: {
+		Code:           "PolicyInvalidName",
+		Description:    "Policy name may not contain comma",
+		HTTPStatusCode: http.StatusBadRequest,
+	},
 	ErrPolicyTooLarge: {
 		Code:           "PolicyTooLarge",
 		Description:    "Policy exceeds the maximum allowed document size.",
@@ -623,7 +629,7 @@ var errorCodes = errorCodeMap{
 	},
 	ErrMissingContentMD5: {
 		Code:           "MissingContentMD5",
-		Description:    "Missing required header for this request: Content-Md5.",
+		Description:    "Missing or invalid required header for this request: Content-Md5 or Amz-Content-Checksum",
 		HTTPStatusCode: http.StatusBadRequest,
 	},
 	ErrMissingSecurityHeader: {
@@ -978,7 +984,7 @@ var errorCodes = errorCodeMap{
 	},
 	ErrReplicationNoExistingObjects: {
 		Code:           "XMinioReplicationNoExistingObjects",
-		Description:    "No matching ExistingsObjects rule enabled",
+		Description:    "No matching ExistingObjects rule enabled",
 		HTTPStatusCode: http.StatusBadRequest,
 	},
 	ErrRemoteTargetDenyAddError: {
@@ -1486,7 +1492,7 @@ var errorCodes = errorCodeMap{
 	},
 	ErrTooManyRequests: {
 		Code:           "TooManyRequests",
-		Description:    "Deadline exceeded while waiting in incoming queue, please reduce your request rate",
+		Description:    "Please reduce your request rate",
 		HTTPStatusCode: http.StatusTooManyRequests,
 	},
 	ErrUnsupportedMetadata: {
@@ -2549,11 +2555,11 @@ func toAPIError(ctx context.Context, err error) APIError {
 			if len(e.Errors) >= 1 {
 				apiErr.Code = e.Errors[0].Reason
 			}
-		case azblob.StorageError:
+		case *azcore.ResponseError:
 			apiErr = APIError{
-				Code:           string(e.ServiceCode()),
+				Code:           e.ErrorCode,
 				Description:    e.Error(),
-				HTTPStatusCode: e.Response().StatusCode,
+				HTTPStatusCode: e.StatusCode,
 			}
 			// Add more other SDK related errors here if any in future.
 		default:
@@ -2592,7 +2598,7 @@ 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)
@@ -166,10 +166,11 @@ type Part struct {
 	Size         int64

 	// Checksum values
-	ChecksumCRC32  string `xml:"ChecksumCRC32,omitempty"`
-	ChecksumCRC32C string `xml:"ChecksumCRC32C,omitempty"`
-	ChecksumSHA1   string `xml:"ChecksumSHA1,omitempty"`
-	ChecksumSHA256 string `xml:"ChecksumSHA256,omitempty"`
+	ChecksumCRC32     string `xml:"ChecksumCRC32,omitempty"`
+	ChecksumCRC32C    string `xml:"ChecksumCRC32C,omitempty"`
+	ChecksumSHA1      string `xml:"ChecksumSHA1,omitempty"`
+	ChecksumSHA256    string `xml:"ChecksumSHA256,omitempty"`
+	ChecksumCRC64NVME string `xml:",omitempty"`
 }

 // ListPartsResponse - format for list parts response.
@@ -192,6 +193,8 @@ type ListPartsResponse struct {
 	IsTruncated          bool

 	ChecksumAlgorithm string
+	ChecksumType      string
+
 	// List of parts.
 	Parts []Part `xml:"Part"`
 }
@@ -413,10 +416,11 @@ type CompleteMultipartUploadResponse struct {
 	Key      string
 	ETag     string

-	ChecksumCRC32  string `xml:"ChecksumCRC32,omitempty"`
-	ChecksumCRC32C string `xml:"ChecksumCRC32C,omitempty"`
-	ChecksumSHA1   string `xml:"ChecksumSHA1,omitempty"`
-	ChecksumSHA256 string `xml:"ChecksumSHA256,omitempty"`
+	ChecksumCRC32     string `xml:"ChecksumCRC32,omitempty"`
+	ChecksumCRC32C    string `xml:"ChecksumCRC32C,omitempty"`
+	ChecksumSHA1      string `xml:"ChecksumSHA1,omitempty"`
+	ChecksumSHA256    string `xml:"ChecksumSHA256,omitempty"`
+	ChecksumCRC64NVME string `xml:",omitempty"`
 }

 // DeleteError structure.
@@ -593,8 +597,6 @@ func generateListVersionsResponse(ctx context.Context, bucket, prefix, marker, v
 			for k, v := range cleanReservedKeys(object.UserDefined) {
 				content.UserMetadata.Set(k, v)
 			}
-
-			content.UserMetadata.Set("expires", object.Expires.Format(http.TimeFormat))
 			content.Internal = &ObjectInternalInfo{
 				K: object.DataBlocks,
 				M: object.ParityBlocks,
@@ -729,7 +731,6 @@ func generateListObjectsV2Response(ctx context.Context, bucket, prefix, token, n
 				for k, v := range cleanReservedKeys(object.UserDefined) {
 					content.UserMetadata.Set(k, v)
 				}
-				content.UserMetadata.Set("expires", object.Expires.Format(http.TimeFormat))
 				content.Internal = &ObjectInternalInfo{
 					K: object.DataBlocks,
 					M: object.ParityBlocks,
@@ -796,11 +797,12 @@ func generateCompleteMultipartUploadResponse(bucket, key, location string, oi Ob
 		Bucket:   bucket,
 		Key:      key,
 		// AWS S3 quotes the ETag in XML, make sure we are compatible here.
-		ETag:           "\"" + oi.ETag + "\"",
-		ChecksumSHA1:   cs[hash.ChecksumSHA1.String()],
-		ChecksumSHA256: cs[hash.ChecksumSHA256.String()],
-		ChecksumCRC32:  cs[hash.ChecksumCRC32.String()],
-		ChecksumCRC32C: cs[hash.ChecksumCRC32C.String()],
+		ETag:              "\"" + oi.ETag + "\"",
+		ChecksumSHA1:      cs[hash.ChecksumSHA1.String()],
+		ChecksumSHA256:    cs[hash.ChecksumSHA256.String()],
+		ChecksumCRC32:     cs[hash.ChecksumCRC32.String()],
+		ChecksumCRC32C:    cs[hash.ChecksumCRC32C.String()],
+		ChecksumCRC64NVME: cs[hash.ChecksumCRC64NVME.String()],
 	}
 	return c
 }
@@ -828,6 +830,7 @@ func generateListPartsResponse(partsInfo ListPartsInfo, encodingType string) Lis
 	listPartsResponse.IsTruncated = partsInfo.IsTruncated
 	listPartsResponse.NextPartNumberMarker = partsInfo.NextPartNumberMarker
 	listPartsResponse.ChecksumAlgorithm = partsInfo.ChecksumAlgorithm
+	listPartsResponse.ChecksumType = partsInfo.ChecksumType

 	listPartsResponse.Parts = make([]Part, len(partsInfo.Parts))
 	for index, part := range partsInfo.Parts {
@@ -840,6 +843,7 @@ func generateListPartsResponse(partsInfo ListPartsInfo, encodingType string) Lis
 		newPart.ChecksumCRC32C = part.ChecksumCRC32C
 		newPart.ChecksumSHA1 = part.ChecksumSHA1
 		newPart.ChecksumSHA256 = part.ChecksumSHA256
+		newPart.ChecksumCRC64NVME = part.ChecksumCRC64NVME
 		listPartsResponse.Parts[index] = newPart
 	}
 	return listPartsResponse
@@ -947,19 +951,10 @@ func writeSuccessResponseHeadersOnly(w http.ResponseWriter) {
 // writeErrorResponse writes error headers
 func writeErrorResponse(ctx context.Context, w http.ResponseWriter, err APIError, reqURL *url.URL) {
 	switch err.HTTPStatusCode {
-	case http.StatusServiceUnavailable:
+	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, "60")
-	case http.StatusTooManyRequests:
-		_, deadline := globalAPIConfig.getRequestsPool()
-		if deadline <= 0 {
-			// Set retry-after header to indicate user-agents to retry request after 10 seconds.
-			// https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Retry-After
-			w.Header().Set(xhttp.RetryAfter, "10")
-		} else {
-			w.Header().Set(xhttp.RetryAfter, strconv.Itoa(int(deadline.Seconds())))
-		}
 	}

 	switch err.Code {
@@ -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)).
@@ -222,7 +222,7 @@ func mustGetClaimsFromToken(r *http.Request) map[string]interface{} {
 	return claims
 }

-func getClaimsFromTokenWithSecret(token, secret string) (map[string]interface{}, error) {
+func getClaimsFromTokenWithSecret(token, secret string) (*xjwt.MapClaims, error) {
 	// JWT token for x-amz-security-token is signed with admin
 	// secret key, temporary credentials become invalid if
 	// server admin credentials change. This is done to ensure
@@ -244,7 +244,7 @@ func getClaimsFromTokenWithSecret(token, secret string) (map[string]interface{},

 	// If AuthZPlugin is set, return without any further checks.
 	if newGlobalAuthZPluginFn() != nil {
-		return claims.Map(), nil
+		return claims, nil
 	}

 	// Check if a session policy is set. If so, decode it here.
@@ -263,12 +263,16 @@ func getClaimsFromTokenWithSecret(token, secret string) (map[string]interface{},
 		claims.MapClaims[sessionPolicyNameExtracted] = string(spBytes)
 	}

-	return claims.Map(), nil
+	return claims, nil
 }

 // Fetch claims in the security token returned by the client.
 func getClaimsFromToken(token string) (map[string]interface{}, error) {
-	return getClaimsFromTokenWithSecret(token, globalActiveCred.SecretKey)
+	jwtClaims, err := getClaimsFromTokenWithSecret(token, globalActiveCred.SecretKey)
+	if err != nil {
+		return nil, err
+	}
+	return jwtClaims.Map(), nil
 }

 // Fetch claims in the security token returned by the client and validate the token.
@@ -319,7 +323,7 @@ func checkClaimsFromToken(r *http.Request, cred auth.Credentials) (map[string]in
 		if err != nil {
 			return nil, toAPIErrorCode(r.Context(), err)
 		}
-		return claims, ErrNone
+		return claims.Map(), ErrNone
 	}

 	claims := xjwt.NewMapClaims()
@@ -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
@@ -90,6 +92,9 @@ type healingTracker struct {
 	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
 }
@@ -143,6 +148,26 @@ func initHealingTracker(disk StorageAPI, healID string) *healingTracker {
 	return h
 }

+func (h *healingTracker) resetHealing() {
+	h.mu.Lock()
+	defer h.mu.Unlock()
+
+	h.ItemsHealed = 0
+	h.ItemsFailed = 0
+	h.BytesDone = 0
+	h.BytesFailed = 0
+	h.ResumeItemsHealed = 0
+	h.ResumeItemsFailed = 0
+	h.ResumeBytesDone = 0
+	h.ResumeBytesFailed = 0
+	h.ItemsSkipped = 0
+	h.BytesSkipped = 0
+
+	h.HealedBuckets = nil
+	h.Object = ""
+	h.Bucket = ""
+}
+
 func (h *healingTracker) getLastUpdate() time.Time {
 	h.mu.RLock()
 	defer h.mu.RUnlock()
@@ -198,9 +223,6 @@ func (h *healingTracker) updateProgress(success, skipped bool, bytes uint64) {
 // update will update the tracker on the disk.
 // If the tracker has been deleted an error is returned.
 func (h *healingTracker) update(ctx context.Context) error {
-	if h.disk.Healing() == nil {
-		return fmt.Errorf("healingTracker: drive %q is not marked as healing", h.ID)
-	}
 	h.mu.Lock()
 	if h.ID == "" || h.PoolIndex < 0 || h.SetIndex < 0 || h.DiskIndex < 0 {
 		h.ID, _ = h.disk.GetDiskID()
@@ -262,8 +284,10 @@ 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.
@@ -274,8 +298,10 @@ 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 {
@@ -324,6 +350,7 @@ 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(),
@@ -339,6 +366,7 @@ 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
@@ -353,16 +381,18 @@ 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.DiskInfo(context.Background(), DiskInfoOptions{})
@@ -370,7 +400,7 @@ func getLocalDisksToHeal() (disksToHeal Endpoints) {
 			disksToHeal = append(disksToHeal, disk.Endpoint())
 			continue
 		}
-		if disk.Healing() != nil {
+		if h := disk.Healing(); h != nil && !h.Finished {
 			disksToHeal = append(disksToHeal, disk.Endpoint())
 		}
 	}
@@ -469,16 +499,19 @@ func healFreshDisk(ctx context.Context, z *erasureServerPools, endpoint Endpoint
 	// if objects have failed healing, we attempt a retry to heal the drive upto 3 times before giving up.
 	if tracker.ItemsFailed > 0 && tracker.RetryAttempts < 4 {
 		tracker.RetryAttempts++
-		bugLogIf(ctx, tracker.update(ctx))

 		healingLogEvent(ctx, "Healing of drive '%s' is incomplete, retrying %s time (healed: %d, skipped: %d, failed: %d).", disk,
 			humanize.Ordinal(int(tracker.RetryAttempts)), tracker.ItemsHealed, tracker.ItemsSkipped, tracker.ItemsFailed)
+
+		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-1, tracker.ItemsHealed, tracker.ItemsSkipped, tracker.ItemsFailed)
+			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,
@@ -516,7 +549,8 @@ func healFreshDisk(ctx context.Context, z *erasureServerPools, endpoint Endpoint
 			continue
 		}
 		if t.HealID == tracker.HealID {
-			t.delete(ctx)
+			t.Finished = true
+			t.update(ctx)
 		}
 	}

@@ -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()
@@ -206,6 +218,12 @@ func (z *healingTracker) DecodeMsg(dc *msgp.Reader) (err error) {
 				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 {
@@ -219,9 +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 26
+	// map header, size 29
 	// write "ID"
-	err = en.Append(0xde, 0x0, 0x1a, 0xa2, 0x49, 0x44)
+	err = en.Append(0xde, 0x0, 0x1d, 0xa2, 0x49, 0x44)
 	if err != nil {
 		return
 	}
@@ -400,6 +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 {
@@ -420,6 +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 {
@@ -494,15 +532,25 @@ func (z *healingTracker) EncodeMsg(en *msgp.Writer) (err error) {
 		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 26
+	// map header, size 29
 	// string "ID"
-	o = append(o, 0xde, 0x0, 0x1a, 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)
@@ -555,12 +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)))
@@ -585,6 +639,9 @@ func (z *healingTracker) MarshalMsg(b []byte) (o []byte, err error) {
 	// 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
 }

@@ -714,6 +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 {
@@ -726,6 +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)
@@ -788,6 +857,12 @@ func (z *healingTracker) UnmarshalMsg(bts []byte) (o []byte, err error) {
 				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 {
@@ -802,7 +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])
 	}
@@ -810,6 +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 + 14 + msgp.Uint64Size
+	s += 7 + msgp.StringPrefixSize + len(z.HealID) + 13 + msgp.Uint64Size + 13 + msgp.Uint64Size + 14 + msgp.Uint64Size + 9 + msgp.BoolSize
 	return
 }
@@ -36,6 +36,7 @@ import (
 	"github.com/minio/pkg/v3/env"
 	"github.com/minio/pkg/v3/wildcard"
 	"github.com/minio/pkg/v3/workers"
+	"github.com/minio/pkg/v3/xtime"
 	"gopkg.in/yaml.v3"
 )

@@ -116,7 +117,7 @@ func (p BatchJobExpirePurge) Validate() error {
 // BatchJobExpireFilter holds all the filters currently supported for batch replication
 type BatchJobExpireFilter struct {
 	line, col     int
-	OlderThan     time.Duration       `yaml:"olderThan,omitempty" json:"olderThan"`
+	OlderThan     xtime.Duration      `yaml:"olderThan,omitempty" json:"olderThan"`
 	CreatedBefore *time.Time          `yaml:"createdBefore,omitempty" json:"createdBefore"`
 	Tags          []BatchJobKV        `yaml:"tags,omitempty" json:"tags"`
 	Metadata      []BatchJobKV        `yaml:"metadata,omitempty" json:"metadata"`
@@ -162,7 +163,7 @@ func (ef BatchJobExpireFilter) Matches(obj ObjectInfo, now time.Time) bool {
 	if len(ef.Name) > 0 && !wildcard.Match(ef.Name, obj.Name) {
 		return false
 	}
-	if ef.OlderThan > 0 && now.Sub(obj.ModTime) <= ef.OlderThan {
+	if ef.OlderThan > 0 && now.Sub(obj.ModTime) <= ef.OlderThan.D() {
 		return false
 	}

@@ -280,7 +281,7 @@ 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 +341,24 @@ 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 (
@@ -371,9 +388,12 @@ func (oiCache objInfoCache) Get(toDel ObjectToDelete) (*ObjectInfo, bool) {

 func batchObjsForDelete(ctx context.Context, r *BatchJobExpire, ri *batchJobInfo, job BatchJobRequest, api ObjectLayer, wk *workers.Workers, expireCh <-chan []expireObjInfo) {
 	vc, _ := globalBucketVersioningSys.Get(r.Bucket)
-	retryAttempts := r.Retry.Attempts
+	retryAttempts := job.Expire.Retry.Attempts
+	if retryAttempts <= 0 {
+		retryAttempts = batchExpireJobDefaultRetries
+	}
 	delay := job.Expire.Retry.Delay
-	if delay == 0 {
+	if delay <= 0 {
 		delay = batchExpireJobDefaultRetryDelay
 	}

@@ -534,40 +554,58 @@ 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() {
+		prefixes := r.Prefix.F()
+		if len(prefixes) == 0 {
+			prefixes = []string{""}
+		}
+		for _, prefix := range prefixes {
+			prefixResultCh := make(chan itemOrErr[ObjectInfo], workerSize)
+			err := api.Walk(ctx, r.Bucket, 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 +622,7 @@ 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 +691,10 @@ 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 {
@@ -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 {
@@ -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")
+	}
 }
@@ -276,8 +276,12 @@ func (r *BatchJobReplicateV1) StartFromSource(ctx context.Context, api ObjectLay
 	}
 	globalBatchJobsMetrics.save(job.ID, ri)

+	retryAttempts := job.Replicate.Flags.Retry.Attempts
+	if retryAttempts <= 0 {
+		retryAttempts = batchReplJobDefaultRetries
+	}
 	delay := job.Replicate.Flags.Retry.Delay
-	if delay == 0 {
+	if delay <= 0 {
 		delay = batchReplJobDefaultRetryDelay
 	}
 	rnd := rand.New(rand.NewSource(time.Now().UnixNano()))
@@ -287,22 +291,22 @@ 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
 		}

-		if !r.Flags.Filter.CreatedAfter.IsZero() && r.Flags.Filter.CreatedAfter.Before(oi.ModTime) {
+		if !r.Flags.Filter.CreatedAfter.IsZero() && r.Flags.Filter.CreatedAfter.After(oi.ModTime) {
 			// skip all objects that are created before the specified time.
 			return true
 		}

-		if !r.Flags.Filter.CreatedBefore.IsZero() && r.Flags.Filter.CreatedBefore.After(oi.ModTime) {
+		if !r.Flags.Filter.CreatedBefore.IsZero() && r.Flags.Filter.CreatedBefore.Before(oi.ModTime) {
 			// skip all objects that are created after the specified time.
 			return true
 		}
@@ -377,7 +381,6 @@ func (r *BatchJobReplicateV1) StartFromSource(ctx context.Context, api ObjectLay
 		return err
 	}

-	retryAttempts := ri.RetryAttempts
 	retry := false
 	for attempts := 1; attempts <= retryAttempts; attempts++ {
 		attempts := attempts
@@ -385,12 +388,27 @@ 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() {
+			prefixes := r.Source.Prefix.F()
+			if len(prefixes) == 0 {
+				prefixes = []string{""}
+			}
+			for _, prefix := range prefixes {
+				prefixObjInfoCh := c.ListObjects(ctx, r.Source.Bucket, miniogo.ListObjectsOptions{
+					Prefix:       prefix,
+					WithVersions: minioSrc,
+					Recursive:    true,
+					WithMetadata: true,
+				})
+				for obj := range prefixObjInfoCh {
+					objInfoCh <- obj
+				}
+			}
+			xioutil.SafeClose(objInfoCh)
+		}()
+
 		prevObj := ""
 		skipReplicate := false

@@ -540,7 +558,7 @@ 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{},
@@ -562,6 +580,10 @@ 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,
@@ -725,7 +747,7 @@ const (

 	batchReplJobAPIVersion        = "v1"
 	batchReplJobDefaultRetries    = 3
-	batchReplJobDefaultRetryDelay = 250 * time.Millisecond
+	batchReplJobDefaultRetryDelay = time.Second
 )

 func getJobPath(job BatchJobRequest) string {
@@ -753,22 +775,10 @@ func (ri *batchJobInfo) loadOrInit(ctx context.Context, api ObjectLayer, job Bat
 		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
 	}
@@ -1015,29 +1025,34 @@ func (r *BatchJobReplicateV1) Start(ctx context.Context, api ObjectLayer, job Ba
 	globalBatchJobsMetrics.save(job.ID, ri)
 	lastObject := ri.Object

+	retryAttempts := job.Replicate.Flags.Retry.Attempts
+	if retryAttempts <= 0 {
+		retryAttempts = batchReplJobDefaultRetries
+	}
 	delay := job.Replicate.Flags.Retry.Delay
-	if delay == 0 {
+	if delay <= 0 {
 		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
 		}

-		if !r.Flags.Filter.CreatedAfter.IsZero() && r.Flags.Filter.CreatedAfter.Before(info.ModTime) {
+		if !r.Flags.Filter.CreatedAfter.IsZero() && r.Flags.Filter.CreatedAfter.After(info.ModTime) {
 			// skip all objects that are created before the specified time.
 			return false
 		}

-		if !r.Flags.Filter.CreatedBefore.IsZero() && r.Flags.Filter.CreatedBefore.After(info.ModTime) {
+		if !r.Flags.Filter.CreatedBefore.IsZero() && r.Flags.Filter.CreatedBefore.Before(info.ModTime) {
 			// skip all objects that are created after the specified time.
 			return false
 		}
@@ -1106,7 +1121,6 @@ func (r *BatchJobReplicateV1) Start(ctx context.Context, api ObjectLayer, job Ba

 	c.SetAppInfo("minio-"+batchJobPrefix, r.APIVersion+" "+job.ID)

-	retryAttempts := ri.RetryAttempts
 	retry := false
 	for attempts := 1; attempts <= retryAttempts; attempts++ {
 		attempts := attempts
@@ -1115,7 +1129,7 @@ func (r *BatchJobReplicateV1) Start(ctx context.Context, api ObjectLayer, job Ba
 			slowCh = make(chan itemOrErr[ObjectInfo], 100)
 		)

-		if !*r.Source.Snowball.Disable && r.Source.Type.isMinio() && r.Target.Type.isMinio() {
+		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{
@@ -1125,7 +1139,7 @@ func (r *BatchJobReplicateV1) Start(ctx context.Context, api ObjectLayer, job Ba
 					BucketLookup: lookupStyle(r.Target.Path),
 				})
 				if err != nil {
-					batchLogIf(ctx, err)
+					batchLogOnceIf(ctx, err, job.ID+"miniogo.New")
 					return
 				}

@@ -1135,8 +1149,8 @@ func (r *BatchJobReplicateV1) Start(ctx context.Context, api ObjectLayer, job Ba
 				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)
+						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}
 							}
@@ -1144,7 +1158,7 @@ func (r *BatchJobReplicateV1) Start(ctx context.Context, api ObjectLayer, job Ba
 							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))
+							batchLogOnceIf(ctx, ri.updateAfter(ctx, api, 10*time.Second, job), job.ID+"updateAfter")
 						}
 					}
 				}
@@ -1184,19 +1198,32 @@ func (r *BatchJobReplicateV1) Start(ctx context.Context, api ObjectLayer, job Ba
 		if walkQuorum == "" {
 			walkQuorum = "strict"
 		}
-		ctx, cancel := context.WithCancel(ctx)
+		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() {
+			prefixes := r.Source.Prefix.F()
+			if len(prefixes) == 0 {
+				prefixes = []string{""}
+			}
+			for _, prefix := range prefixes {
+				prefixWalkCh := make(chan itemOrErr[ObjectInfo], 100)
+				if err := api.Walk(ctx, r.Source.Bucket, 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 := ""

@@ -1204,7 +1231,7 @@ 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
@@ -1231,7 +1258,7 @@ 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)
@@ -1239,7 +1266,7 @@ 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)
@@ -1247,20 +1274,23 @@ 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 = ""
@@ -1547,9 +1577,6 @@ func (a adminAPIHandlers) ListBatchJobs(w http.ResponseWriter, r *http.Request)
 	}

 	jobType := r.Form.Get("jobType")
-	if jobType == "" {
-		jobType = string(madmin.BatchJobReplicate)
-	}

 	resultCh := make(chan itemOrErr[ObjectInfo])

@@ -1567,6 +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) {
@@ -1575,7 +1605,7 @@ func (a adminAPIHandlers) ListBatchJobs(w http.ResponseWriter, r *http.Request)
 			continue
 		}

-		if jobType == string(req.Type()) {
+		if jobType == string(req.Type()) || jobType == "" {
 			listResult.Jobs = append(listResult.Jobs, madmin.BatchJobResult{
 				ID:      req.ID,
 				Type:    req.Type(),
@@ -1675,7 +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()

@@ -1883,6 +1913,9 @@ func (j *BatchJobPool) resume(randomWait func() time.Duration) {
 			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
@@ -1953,7 +1986,6 @@ func (j *BatchJobPool) AddWorker() {
 					}
 				}
 			}
-			job.delete(j.ctx, j.objLayer)
 			j.canceler(job.ID, false)
 		case <-j.workerKillCh:
 			return
@@ -2014,7 +2046,9 @@ func (j *BatchJobPool) canceler(jobID string, cancel bool) error {
 			canceler()
 		}
 	}
-	delete(j.jobCancelers, jobID)
+	if cancel {
+		delete(j.jobCancelers, jobID)
+	}
 	return nil
 }

@@ -2140,11 +2174,42 @@ func (m *batchJobMetrics) purgeJobMetrics() {
 			m.RUnlock()
 			for _, jobID := range toDeleteJobMetrics {
 				m.delete(jobID)
+				j := BatchJobRequest{
+					ID: jobID,
+				}
+				j.delete(GlobalContext, newObjectLayerFn())
 			}
 		}
 	}
 }

+// load metrics from disk on startup
+func (m *batchJobMetrics) init(ctx context.Context, objectAPI ObjectLayer) error {
+	resultCh := make(chan itemOrErr[ObjectInfo])
+
+	ctx, cancel := context.WithCancel(ctx)
+	defer cancel()
+
+	if err := objectAPI.Walk(ctx, minioMetaBucket, batchJobReportsPrefix, resultCh, WalkOptions{}); err != nil {
+		return err
+	}
+
+	for result := range resultCh {
+		if result.Err != nil {
+			return result.Err
+		}
+		ri := &batchJobInfo{}
+		if err := ri.loadByPath(ctx, objectAPI, result.Item.Name); err != nil {
+			if !errors.Is(err, errNoSuchJob) {
+				batchLogIf(ctx, err)
+			}
+			continue
+		}
+		m.metrics[ri.JobID] = ri
+	}
+	return nil
+}
+
 func (m *batchJobMetrics) delete(jobID string) {
 	m.Lock()
 	defer m.Unlock()
@@ -2222,3 +2287,30 @@ 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
+}
@@ -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
@@ -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)
@@ -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)
+			}
+		})
+	}
+}
@@ -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"`
@@ -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()
 	}
@@ -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")
+	}
+}
@@ -267,8 +267,12 @@ func (r *BatchJobKeyRotateV1) Start(ctx context.Context, api ObjectLayer, job Ba
 	globalBatchJobsMetrics.save(job.ID, ri)
 	lastObject := ri.Object

+	retryAttempts := job.KeyRotate.Flags.Retry.Attempts
+	if retryAttempts <= 0 {
+		retryAttempts = batchKeyRotateJobDefaultRetries
+	}
 	delay := job.KeyRotate.Flags.Retry.Delay
-	if delay == 0 {
+	if delay <= 0 {
 		delay = batchKeyRotateJobDefaultRetryDelay
 	}

@@ -354,7 +358,6 @@ func (r *BatchJobKeyRotateV1) Start(ctx context.Context, api ObjectLayer, job Ba
 		return err
 	}

-	retryAttempts := ri.RetryAttempts
 	ctx, cancel := context.WithCancel(ctx)

 	results := make(chan itemOrErr[ObjectInfo], 100)
@@ -20,6 +20,7 @@ package cmd
 import (
 	"bytes"
 	"context"
+	"errors"
 	"hash"
 	"io"
 	"sync"
@@ -37,12 +38,22 @@ type streamingBitrotWriter struct {
 	shardSize    int64
 	canClose     *sync.WaitGroup
 	byteBuf      []byte
+	finished     bool
 }

 func (b *streamingBitrotWriter) Write(p []byte) (int, error) {
 	if len(p) == 0 {
 		return 0, nil
 	}
+	if b.finished {
+		return 0, errors.New("bitrot write not allowed")
+	}
+	if int64(len(p)) > b.shardSize {
+		return 0, errors.New("unexpected bitrot buffer size")
+	}
+	if int64(len(p)) < b.shardSize {
+		b.finished = true
+	}
 	b.h.Reset()
 	b.h.Write(p)
 	hashBytes := b.h.Sum(nil)
@@ -141,13 +152,7 @@ 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
@@ -183,11 +183,19 @@ 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)
+	binPath, err := os.Executable()
+	if err != nil {
+		logger.Error("Calculating checksum failed: %s", err)
+		return "00000000000000000000000000000000"
 	}
+	b, err := os.Open(binPath)
+	if err != nil {
+		logger.Error("Calculating checksum failed: %s", err)
+		return "00000000000000000000000000000000"
+	}
+
+	defer b.Close()
+	io.Copy(mw, b)
 	return hex.EncodeToString(mw.Sum(nil))
 }

@@ -28,6 +28,7 @@ import (
 	"errors"
 	"fmt"
 	"io"
+	"mime"
 	"mime/multipart"
 	"net/http"
 	"net/textproto"
@@ -53,6 +54,7 @@ import (
 	"github.com/minio/minio/internal/bucket/replication"
 	"github.com/minio/minio/internal/config/dns"
 	"github.com/minio/minio/internal/crypto"
+	"github.com/minio/minio/internal/etag"
 	"github.com/minio/minio/internal/event"
 	"github.com/minio/minio/internal/handlers"
 	"github.com/minio/minio/internal/hash"
@@ -427,7 +429,7 @@ func (api objectAPIHandlers) DeleteMultipleObjectsHandler(w http.ResponseWriter,

 	// Content-Md5 is required should be set
 	// http://docs.aws.amazon.com/AmazonS3/latest/API/multiobjectdeleteapi.html
-	if _, ok := r.Header[xhttp.ContentMD5]; !ok {
+	if !validateLengthAndChecksum(r) {
 		writeErrorResponse(ctx, w, errorCodes.ToAPIErr(ErrMissingContentMD5), r.URL)
 		return
 	}
@@ -669,8 +671,6 @@ func (api objectAPIHandlers) DeleteMultipleObjectsHandler(w http.ResponseWriter,
 			continue
 		}

-		defer globalCacheConfig.Delete(bucket, dobj.ObjectName)
-
 		if replicateDeletes && (dobj.DeleteMarkerReplicationStatus() == replication.Pending || dobj.VersionPurgeStatus() == Pending) {
 			// copy so we can re-add null ID.
 			dobj := dobj
@@ -889,6 +889,30 @@ func (api objectAPIHandlers) PutBucketHandler(w http.ResponseWriter, r *http.Req
 	})
 }

+// multipartReader is just like https://pkg.go.dev/net/http#Request.MultipartReader but
+// rejects multipart/mixed as its not supported in S3 API.
+func multipartReader(r *http.Request) (*multipart.Reader, error) {
+	v := r.Header.Get("Content-Type")
+	if v == "" {
+		return nil, http.ErrNotMultipart
+	}
+	if r.Body == nil {
+		return nil, errors.New("missing form body")
+	}
+	d, params, err := mime.ParseMediaType(v)
+	if err != nil {
+		return nil, http.ErrNotMultipart
+	}
+	if d != "multipart/form-data" {
+		return nil, http.ErrNotMultipart
+	}
+	boundary, ok := params["boundary"]
+	if !ok {
+		return nil, http.ErrMissingBoundary
+	}
+	return multipart.NewReader(r.Body, boundary), nil
+}
+
 // PostPolicyBucketHandler - POST policy
 // ----------
 // This implementation of the POST operation handles object creation with a specified
@@ -922,9 +946,14 @@ func (api objectAPIHandlers) PostPolicyBucketHandler(w http.ResponseWriter, r *h
 		return
 	}

+	if r.ContentLength <= 0 {
+		writeErrorResponse(ctx, w, errorCodes.ToAPIErr(ErrEmptyRequestBody), r.URL)
+		return
+	}
+
 	// Here the parameter is the size of the form data that should
 	// be loaded in memory, the remaining being put in temporary files.
-	mp, err := r.MultipartReader()
+	mp, err := multipartReader(r)
 	if err != nil {
 		apiErr := errorCodes.ToAPIErr(ErrMalformedPOSTRequest)
 		apiErr.Description = fmt.Sprintf("%s (%v)", apiErr.Description, err)
@@ -936,7 +965,7 @@ func (api objectAPIHandlers) PostPolicyBucketHandler(w http.ResponseWriter, r *h

 	var (
 		reader        io.Reader
-		fileSize      int64 = -1
+		actualSize    int64 = -1
 		fileName      string
 		fanOutEntries = make([]minio.PutObjectFanOutEntry, 0, 100)
 	)
@@ -944,6 +973,7 @@ func (api objectAPIHandlers) PostPolicyBucketHandler(w http.ResponseWriter, r *h
 	maxParts := 1000
 	// Canonicalize the form values into http.Header.
 	formValues := make(http.Header)
+	var headerLen int64
 	for {
 		part, err := mp.NextRawPart()
 		if errors.Is(err, io.EOF) {
@@ -985,7 +1015,7 @@ func (api objectAPIHandlers) PostPolicyBucketHandler(w http.ResponseWriter, r *h
 			return
 		}

-		var b bytes.Buffer
+		headerLen += int64(len(name)) + int64(len(fileName))
 		if name != "file" {
 			if http.CanonicalHeaderKey(name) == http.CanonicalHeaderKey("x-minio-fanout-list") {
 				dec := json.NewDecoder(part)
@@ -996,7 +1026,7 @@ func (api objectAPIHandlers) PostPolicyBucketHandler(w http.ResponseWriter, r *h
 					if err := dec.Decode(&m); err != nil {
 						part.Close()
 						apiErr := errorCodes.ToAPIErr(ErrMalformedPOSTRequest)
-						apiErr.Description = fmt.Sprintf("%s (%v)", apiErr.Description, multipart.ErrMessageTooLarge)
+						apiErr.Description = fmt.Sprintf("%s (%v)", apiErr.Description, err)
 						writeErrorResponse(ctx, w, apiErr, r.URL)
 						return
 					}
@@ -1006,8 +1036,12 @@ func (api objectAPIHandlers) PostPolicyBucketHandler(w http.ResponseWriter, r *h
 				continue
 			}

+			buf := bytebufferpool.Get()
 			// value, store as string in memory
-			n, err := io.CopyN(&b, part, maxMemoryBytes+1)
+			n, err := io.CopyN(buf, part, maxMemoryBytes+1)
+			value := buf.String()
+			buf.Reset()
+			bytebufferpool.Put(buf)
 			part.Close()

 			if err != nil && err != io.EOF {
@@ -1029,7 +1063,8 @@ func (api objectAPIHandlers) PostPolicyBucketHandler(w http.ResponseWriter, r *h
 				writeErrorResponse(ctx, w, apiErr, r.URL)
 				return
 			}
-			formValues[http.CanonicalHeaderKey(name)] = append(formValues[http.CanonicalHeaderKey(name)], b.String())
+			headerLen += n
+			formValues[http.CanonicalHeaderKey(name)] = append(formValues[http.CanonicalHeaderKey(name)], value)
 			continue
 		}

@@ -1038,6 +1073,21 @@ func (api objectAPIHandlers) PostPolicyBucketHandler(w http.ResponseWriter, r *h
 		// The file or text content must be the last field in the form.
 		// You cannot upload more than one file at a time.
 		reader = part
+
+		possibleShardSize := (r.ContentLength - headerLen)
+		if globalStorageClass.ShouldInline(possibleShardSize, false) { // keep versioned false for this check
+			var b bytes.Buffer
+			n, err := io.Copy(&b, reader)
+			if err != nil {
+				apiErr := errorCodes.ToAPIErr(ErrMalformedPOSTRequest)
+				apiErr.Description = fmt.Sprintf("%s (%v)", apiErr.Description, err)
+				writeErrorResponse(ctx, w, apiErr, r.URL)
+				return
+			}
+			reader = &b
+			actualSize = n
+		}
+
 		// we have found the File part of the request we are done processing multipart-form
 		break
 	}
@@ -1139,11 +1189,33 @@ func (api objectAPIHandlers) PostPolicyBucketHandler(w http.ResponseWriter, r *h
 		return
 	}

-	hashReader, err := hash.NewReader(ctx, reader, fileSize, "", "", fileSize)
+	clientETag, err := etag.FromContentMD5(formValues)
+	if err != nil {
+		writeErrorResponse(ctx, w, errorCodes.ToAPIErr(ErrInvalidDigest), r.URL)
+		return
+	}
+
+	var forceMD5 []byte
+	// Optimization: If SSE-KMS and SSE-C did not request Content-Md5. Use uuid as etag. Optionally enable this also
+	// for server that is started with `--no-compat`.
+	kind, _ := crypto.IsRequested(formValues)
+	if !etag.ContentMD5Requested(formValues) && (kind == crypto.SSEC || kind == crypto.S3KMS || !globalServerCtxt.StrictS3Compat) {
+		forceMD5 = mustGetUUIDBytes()
+	}
+
+	hashReader, err := hash.NewReaderWithOpts(ctx, reader, hash.Options{
+		Size:       actualSize,
+		MD5Hex:     clientETag.String(),
+		SHA256Hex:  "",
+		ActualSize: actualSize,
+		DisableMD5: false,
+		ForceMD5:   forceMD5,
+	})
 	if err != nil {
 		writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL)
 		return
 	}
+
 	if checksum != nil && checksum.Valid() {
 		if err = hashReader.AddChecksumNoTrailer(formValues, false); err != nil {
 			writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL)
@@ -1200,9 +1272,9 @@ func (api objectAPIHandlers) PostPolicyBucketHandler(w http.ResponseWriter, r *h
 		writeErrorResponseHeadersOnly(w, toAPIError(ctx, err))
 		return
 	}
+	opts.WantChecksum = checksum

 	fanOutOpts := fanOutOptions{Checksum: checksum}
-
 	if crypto.Requested(formValues) {
 		if crypto.SSECopy.IsRequested(r.Header) {
 			writeErrorResponse(ctx, w, toAPIError(ctx, errInvalidEncryptionParameters), r.URL)
@@ -1247,8 +1319,15 @@ func (api objectAPIHandlers) PostPolicyBucketHandler(w http.ResponseWriter, r *h
 				writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL)
 				return
 			}
+
+			wantSize := int64(-1)
+			if actualSize >= 0 {
+				info := ObjectInfo{Size: actualSize}
+				wantSize = info.EncryptedSize()
+			}
+
 			// do not try to verify encrypted content/
-			hashReader, err = hash.NewReader(ctx, reader, -1, "", "", -1)
+			hashReader, err = hash.NewReader(ctx, reader, wantSize, "", "", actualSize)
 			if err != nil {
 				writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL)
 				return
@@ -1259,6 +1338,7 @@ func (api objectAPIHandlers) PostPolicyBucketHandler(w http.ResponseWriter, r *h
 					return
 				}
 			}
+			opts.EncryptFn = metadataEncrypter(objectEncryptionKey)
 			pReader, err = pReader.WithEncryption(hashReader, &objectEncryptionKey)
 			if err != nil {
 				writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL)
@@ -1328,7 +1408,6 @@ func (api objectAPIHandlers) PostPolicyBucketHandler(w http.ResponseWriter, r *h
 						Key:   objInfo.Name,
 						Error: errs[i].Error(),
 					})
-
 					eventArgsList = append(eventArgsList, eventArgs{
 						EventName:    event.ObjectCreatedPost,
 						BucketName:   objInfo.Bucket,
@@ -1682,7 +1761,7 @@ 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))
@@ -1731,6 +1810,10 @@ func (api objectAPIHandlers) PutBucketObjectLockConfigHandler(w http.ResponseWri
 		return
 	}

+	// Audit log tags.
+	reqInfo := logger.GetReqInfo(ctx)
+	reqInfo.SetTags("retention", config.String())
+
 	configData, err := xml.Marshal(config)
 	if err != nil {
 		writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL)
@@ -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
 }
@@ -19,7 +19,6 @@ package cmd

 import (
 	"encoding/xml"
-	"io"
 	"net/http"
 	"strconv"
 	"time"
@@ -53,7 +52,7 @@ func (api objectAPIHandlers) PutBucketLifecycleHandler(w http.ResponseWriter, r
 	bucket := vars["bucket"]

 	// PutBucketLifecycle always needs a Content-Md5
-	if _, ok := r.Header[xhttp.ContentMD5]; !ok {
+	if !validateLengthAndChecksum(r) {
 		writeErrorResponse(ctx, w, errorCodes.ToAPIErr(ErrMissingContentMD5), r.URL)
 		return
 	}
@@ -70,7 +69,7 @@ func (api objectAPIHandlers) PutBucketLifecycleHandler(w http.ResponseWriter, r
 		return
 	}

-	bucketLifecycle, err := lifecycle.ParseLifecycleConfigWithID(io.LimitReader(r.Body, r.ContentLength))
+	bucketLifecycle, err := lifecycle.ParseLifecycleConfigWithID(r.Body)
 	if err != nil {
 		writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL)
 		return
@@ -71,7 +71,7 @@ func NewLifecycleSys() *LifecycleSys {
 	return &LifecycleSys{}
 }

-func ilmTrace(startTime time.Time, duration time.Duration, oi ObjectInfo, event string) madmin.TraceInfo {
+func ilmTrace(startTime time.Time, duration time.Duration, oi ObjectInfo, event string, metadata map[string]string, err string) madmin.TraceInfo {
 	sz, _ := oi.GetActualSize()
 	return madmin.TraceInfo{
 		TraceType: madmin.TraceILM,
@@ -81,18 +81,22 @@ func ilmTrace(startTime time.Time, duration time.Duration, oi ObjectInfo, event
 		Duration:  duration,
 		Path:      pathJoin(oi.Bucket, oi.Name),
 		Bytes:     sz,
-		Error:     "",
+		Error:     err,
 		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, err error) {
 	startTime := time.Now()
-	return func(event string) {
+	return func(event string, metadata map[string]string, err error) {
 		duration := time.Since(startTime)
 		if globalTrace.NumSubscribers(madmin.TraceILM) > 0 {
-			globalTrace.Publish(ilmTrace(startTime, duration, oi, event))
+			e := ""
+			if err != nil {
+				e = err.Error()
+			}
+			globalTrace.Publish(ilmTrace(startTime, duration, oi, event, metadata, e))
 		}
 	}
 }
@@ -348,7 +352,7 @@ func (es *expiryState) Worker(input <-chan expiryOp) {
 				traceFn := globalLifecycleSys.trace(oi)
 				if !oi.TransitionedObject.FreeVersion {
 					// nothing to be done
-					return
+					continue
 				}

 				ignoreNotFoundErr := func(err error) error {
@@ -362,7 +366,8 @@ func (es *expiryState) Worker(input <-chan expiryOp) {
 				err := deleteObjectFromRemoteTier(es.ctx, oi.TransitionedObject.Name, oi.TransitionedObject.VersionID, oi.TransitionedObject.Tier)
 				if ignoreNotFoundErr(err) != nil {
 					transitionLogIf(es.ctx, err)
-					return
+					traceFn(ILMFreeVersionDelete, nil, err)
+					continue
 				}

 				// Remove this free version
@@ -707,6 +712,11 @@ type auditTierOp struct {
 	Error            string `json:"error,omitempty"`
 }

+func (op auditTierOp) String() string {
+	// flattening the auditTierOp{} for audit
+	return fmt.Sprintf("tier:%s,respNS:%d,tx:%d,err:%s", op.Tier, op.TimeToResponseNS, op.OutputBytes, op.Error)
+}
+
 func auditTierActions(ctx context.Context, tier string, bytes int64) func(err error) {
 	startTime := time.Now()
 	return func(err error) {
@@ -730,7 +740,7 @@ func auditTierActions(ctx context.Context, tier string, bytes int64) func(err er
 			globalTierMetrics.logFailure(tier)
 		}

-		logger.GetReqInfo(ctx).AppendTags("tierStats", op)
+		logger.GetReqInfo(ctx).AppendTags("tierStats", op.String())
 	}
 }

@@ -998,7 +1008,7 @@ 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,
@@ -245,7 +245,7 @@ func parseRequestToken(token string) (subToken string, nodeIndex int) {

 func proxyRequestByToken(ctx context.Context, w http.ResponseWriter, r *http.Request, token string) (string, bool) {
 	subToken, nodeIndex := parseRequestToken(token)
-	if nodeIndex > 0 {
+	if nodeIndex >= 0 {
 		return subToken, proxyRequestByNodeIndex(ctx, w, r, nodeIndex)
 	}
 	return subToken, false
@@ -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) {
@@ -451,13 +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
@@ -480,7 +507,7 @@ func (sys *BucketMetadataSys) Init(ctx context.Context, buckets []string, objAPI
 }

 // concurrently load bucket metadata to speed up loading bucket metadata.
-func (sys *BucketMetadataSys) concurrentLoad(ctx context.Context, buckets []string, 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 {
@@ -521,10 +548,6 @@ func (sys *BucketMetadataSys) concurrentLoad(ctx context.Context, buckets []stri

 	for i, meta := range bucketMetas {
 		if errs[i] != nil {
-			if failedBuckets == nil {
-				failedBuckets = make(map[string]struct{})
-			}
-			failedBuckets[buckets[i]] = struct{}{}
 			continue
 		}
 		globalEventNotifier.set(buckets[i], meta)   // set notification targets
@@ -532,7 +555,7 @@ func (sys *BucketMetadataSys) concurrentLoad(ctx context.Context, buckets []stri
 	}
 }

-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)
@@ -562,7 +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.
@@ -570,14 +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].Name, meta)
-					globalBucketTargetSys.set(buckets[i].Name, meta)
-					delete(failedBuckets, buckets[i].Name)
+				if updated {
+					globalEventNotifier.set(bucket, meta)
+					globalBucketTargetSys.set(bucket, meta)
 				}

 				wait() // wait to proceed to next entry.
@@ -598,13 +626,12 @@ func (sys *BucketMetadataSys) Initialized() bool {
 // Loads bucket metadata for all buckets into BucketMetadataSys.
 func (sys *BucketMetadataSys) init(ctx context.Context, buckets []string) {
 	count := globalEndpoints.ESCount() * 10
-	failedBuckets := make(map[string]struct{})
 	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:]
 	}

@@ -613,7 +640,7 @@ func (sys *BucketMetadataSys) init(ctx context.Context, buckets []string) {
 	sys.Unlock()

 	if globalIsDistErasure {
-		go sys.refreshBucketsMetadataLoop(ctx, failedBuckets)
+		go sys.refreshBucketsMetadataLoop(ctx)
 	}
 }

@@ -630,5 +657,6 @@ func (sys *BucketMetadataSys) Reset() {
 func NewBucketMetadataSys() *BucketMetadataSys {
 	return &BucketMetadataSys{
 		metadataMap: make(map[string]BucketMetadata),
+		group:       &singleflight.Group{},
 	}
 }
@@ -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.
@@ -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
 }
@@ -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}
 		}
@@ -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
@@ -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())
@@ -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)
 }
@@ -50,10 +50,9 @@ 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"
-	"golang.org/x/exp/slices"
 )

 const (
@@ -91,7 +90,10 @@ func getReplicationConfig(ctx context.Context, bucketName string) (rc *replicati

 // validateReplicationDestination returns error if replication destination bucket missing or not configured
 // It also returns true if replication destination is same as this server.
-func validateReplicationDestination(ctx context.Context, bucket string, rCfg *replication.Config, checkRemote bool) (bool, APIError) {
+func validateReplicationDestination(ctx context.Context, bucket string, rCfg *replication.Config, opts *validateReplicationDestinationOptions) (bool, APIError) {
+	if opts == nil {
+		opts = &validateReplicationDestinationOptions{}
+	}
 	var arns []string
 	if rCfg.RoleArn != "" {
 		arns = append(arns, rCfg.RoleArn)
@@ -113,7 +115,7 @@ func validateReplicationDestination(ctx context.Context, bucket string, rCfg *re
 		if clnt == nil {
 			return sameTarget, toAPIError(ctx, BucketRemoteTargetNotFound{Bucket: bucket})
 		}
-		if checkRemote { // validate remote bucket
+		if opts.CheckRemoteBucket { // validate remote bucket
 			found, err := clnt.BucketExists(ctx, arn.Bucket)
 			if err != nil {
 				return sameTarget, errorCodes.ToAPIErrWithErr(ErrRemoteDestinationNotFoundError, err)
@@ -133,23 +135,29 @@ func validateReplicationDestination(ctx context.Context, bucket string, rCfg *re
 				}
 			}
 		}
+		// if checked bucket, then check the ready is unnecessary
+		if !opts.CheckRemoteBucket && opts.CheckReady {
+			endpoint := clnt.EndpointURL().String()
+			if errInt, ok := opts.checkReadyErr.Load(endpoint); !ok {
+				err = checkRemoteEndpoint(ctx, clnt.EndpointURL())
+				opts.checkReadyErr.Store(endpoint, err)
+			} else {
+				if errInt == nil {
+					err = nil
+				} else {
+					err = errInt.(error)
+				}
+			}
+			switch err.(type) {
+			case BucketRemoteIdenticalToSource:
+				return true, errorCodes.ToAPIErrWithErr(ErrBucketRemoteIdenticalToSource, fmt.Errorf("remote target endpoint %s is self referential", clnt.EndpointURL().String()))
+			default:
+			}
+		}
 		// validate replication ARN against target endpoint
-		c := globalBucketTargetSys.GetRemoteTargetClient(bucket, arnStr)
-		if c != nil {
-			if err := checkRemoteEndpoint(ctx, c.EndpointURL()); err != nil {
-				switch err.(type) {
-				case BucketRemoteIdenticalToSource:
-					return true, errorCodes.ToAPIErrWithErr(ErrBucketRemoteIdenticalToSource, fmt.Errorf("remote target endpoint %s is self referential", c.EndpointURL().String()))
-				default:
-				}
-			}
-			if c.EndpointURL().String() == clnt.EndpointURL().String() {
-				selfTarget, _ := isLocalHost(clnt.EndpointURL().Hostname(), clnt.EndpointURL().Port(), globalMinioPort)
-				if !sameTarget {
-					sameTarget = selfTarget
-				}
-				continue
-			}
+		selfTarget, _ := isLocalHost(clnt.EndpointURL().Hostname(), clnt.EndpointURL().Port(), globalMinioPort)
+		if !sameTarget {
+			sameTarget = selfTarget
 		}
 	}

@@ -469,7 +477,7 @@ func replicateDelete(ctx context.Context, dobj DeletedObjectReplicationInfo, obj
 	lk := objectAPI.NewNSLock(bucket, "/[replicate]/"+dobj.ObjectName)
 	lkctx, err := lk.GetLock(ctx, globalOperationTimeout)
 	if err != nil {
-		globalReplicationPool.queueMRFSave(dobj.ToMRFEntry())
+		globalReplicationPool.Get().queueMRFSave(dobj.ToMRFEntry())
 		sendEvent(eventArgs{
 			BucketName: bucket,
 			Object: ObjectInfo{
@@ -539,7 +547,7 @@ func replicateDelete(ctx context.Context, dobj DeletedObjectReplicationInfo, obj
 	// to decrement pending count later.
 	for _, rinfo := range rinfos.Targets {
 		if rinfo.ReplicationStatus != rinfo.PrevReplicationStatus {
-			globalReplicationStats.Update(dobj.Bucket, rinfo, replicationStatus,
+			globalReplicationStats.Load().Update(dobj.Bucket, rinfo, replicationStatus,
 				prevStatus)
 		}
 	}
@@ -547,7 +555,7 @@ func replicateDelete(ctx context.Context, dobj DeletedObjectReplicationInfo, obj
 	eventName := event.ObjectReplicationComplete
 	if replicationStatus == replication.Failed {
 		eventName = event.ObjectReplicationFailed
-		globalReplicationPool.queueMRFSave(dobj.ToMRFEntry())
+		globalReplicationPool.Get().queueMRFSave(dobj.ToMRFEntry())
 	}
 	drs := getReplicationState(rinfos, dobj.ReplicationState, dobj.VersionID)
 	if replicationStatus != prevStatus {
@@ -765,54 +773,44 @@ func (m caseInsensitiveMap) Lookup(key string) (string, bool) {
 	return "", false
 }

-func getCRCMeta(oi ObjectInfo, partNum int, h http.Header) map[string]string {
-	meta := make(map[string]string)
-	cs := oi.decryptChecksums(partNum, h)
-	for k, v := range cs {
-		cksum := hash.NewChecksumString(k, v)
-		if cksum == nil {
-			continue
-		}
-		if cksum.Valid() {
-			meta[cksum.Type.Key()] = v
-		}
-	}
-	return meta
-}
-
 func putReplicationOpts(ctx context.Context, sc string, objInfo ObjectInfo, partNum int) (putOpts minio.PutObjectOptions, err error) {
 	meta := make(map[string]string)
 	isSSEC := crypto.SSEC.IsEncrypted(objInfo.UserDefined)

 	for k, v := range objInfo.UserDefined {
+		_, isValidSSEHeader := validSSEReplicationHeaders[k]
 		// In case of SSE-C objects copy the allowed internal headers as well
-		if !isSSEC || !slices.Contains(maps.Keys(validSSEReplicationHeaders), k) {
+		if !isSSEC || !isValidSSEHeader {
 			if stringsHasPrefixFold(k, ReservedMetadataPrefixLower) {
-				if strings.EqualFold(k, ReservedMetadataPrefixLower+"crc") {
-					for k, v := range getCRCMeta(objInfo, partNum, nil) {
-						meta[k] = v
-					}
-				}
 				continue
 			}
 			if isStandardHeader(k) {
 				continue
 			}
 		}
-		if slices.Contains(maps.Keys(validSSEReplicationHeaders), k) {
+		if isValidSSEHeader {
 			meta[validSSEReplicationHeaders[k]] = v
 		} else {
 			meta[k] = v
 		}
 	}
-
 	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 k, v := range getCRCMeta(objInfo, 0, nil) {
-				meta[k] = v
+			if objInfo.isMultipart() && partNum > 0 {
+				for _, pi := range objInfo.Parts {
+					if pi.Number == partNum {
+						for k, v := range pi.Checksums { // for PutObjectPart
+							meta[k] = v
+						}
+					}
+				}
+			} else {
+				for k, v := range getCRCMeta(objInfo, 0, nil) { // for PutObject/NewMultipartUpload
+					meta[k] = v
+				}
 			}
 		}
 	}
@@ -895,6 +893,14 @@ func putReplicationOpts(ctx context.Context, sc string, objInfo ObjectInfo, part
 	if crypto.S3.IsEncrypted(objInfo.UserDefined) {
 		putOpts.ServerSideEncryption = encrypt.NewSSE()
 	}
+
+	if crypto.S3KMS.IsEncrypted(objInfo.UserDefined) {
+		sseEnc, err := encrypt.NewSSEKMS(objInfo.KMSKeyID(), nil)
+		if err != nil {
+			return putOpts, err
+		}
+		putOpts.ServerSideEncryption = sseEnc
+	}
 	return
 }

@@ -1061,7 +1067,7 @@ func replicateObject(ctx context.Context, ri ReplicateObjectInfo, objectAPI Obje
 			UserAgent:  "Internal: [Replication]",
 			Host:       globalLocalNodeName,
 		})
-		globalReplicationPool.queueMRFSave(ri.ToMRFEntry())
+		globalReplicationPool.Get().queueMRFSave(ri.ToMRFEntry())
 		return
 	}
 	ctx = lkctx.Context()
@@ -1146,7 +1152,7 @@ func replicateObject(ctx context.Context, ri ReplicateObjectInfo, objectAPI Obje
 		for _, rinfo := range rinfos.Targets {
 			if rinfo.ReplicationStatus != rinfo.PrevReplicationStatus {
 				rinfo.OpType = opType // update optype to reflect correct operation.
-				globalReplicationStats.Update(bucket, rinfo, rinfo.ReplicationStatus, rinfo.PrevReplicationStatus)
+				globalReplicationStats.Load().Update(bucket, rinfo, rinfo.ReplicationStatus, rinfo.PrevReplicationStatus)
 			}
 		}
 	}
@@ -1166,7 +1172,7 @@ func replicateObject(ctx context.Context, ri ReplicateObjectInfo, objectAPI Obje
 		ri.EventType = ReplicateMRF
 		ri.ReplicationStatusInternal = rinfos.ReplicationStatusInternal()
 		ri.RetryCount++
-		globalReplicationPool.queueMRFSave(ri.ToMRFEntry())
+		globalReplicationPool.Get().queueMRFSave(ri.ToMRFEntry())
 	}
 }

@@ -1306,7 +1312,7 @@ func (ri ReplicateObjectInfo) replicateObject(ctx context.Context, objectAPI Obj
 		HeaderSize: headerSize,
 	}
 	newCtx := ctx
-	if globalBucketMonitor.IsThrottled(bucket, tgt.ARN) {
+	if globalBucketMonitor.IsThrottled(bucket, tgt.ARN) && objInfo.Size < minLargeObjSize {
 		var cancel context.CancelFunc
 		newCtx, cancel = context.WithTimeout(ctx, throttleDeadline)
 		defer cancel()
@@ -1581,7 +1587,7 @@ applyAction:
 			HeaderSize: headerSize,
 		}
 		newCtx := ctx
-		if globalBucketMonitor.IsThrottled(bucket, tgt.ARN) {
+		if globalBucketMonitor.IsThrottled(bucket, tgt.ARN) && objInfo.Size < minLargeObjSize {
 			var cancel context.CancelFunc
 			newCtx, cancel = context.WithTimeout(ctx, throttleDeadline)
 			defer cancel()
@@ -1648,7 +1654,6 @@ func replicateObjectWithMultipart(ctx context.Context, c *minio.Core, bucket, ob

 	var (
 		hr     *hash.Reader
-		pInfo  minio.ObjectPart
 		isSSEC = crypto.SSEC.IsEncrypted(objInfo.UserDefined)
 	)

@@ -1666,8 +1671,7 @@ func replicateObjectWithMultipart(ctx context.Context, c *minio.Core, bucket, ob
 		cHeader := http.Header{}
 		cHeader.Add(xhttp.MinIOSourceReplicationRequest, "true")
 		if !isSSEC {
-			crc := getCRCMeta(objInfo, partInfo.Number, nil) // No SSE-C keys here.
-			for k, v := range crc {
+			for k, v := range getCRCMeta(objInfo, partInfo.Number, nil) {
 				cHeader.Add(k, v)
 			}
 		}
@@ -1676,30 +1680,32 @@ func replicateObjectWithMultipart(ctx context.Context, c *minio.Core, bucket, ob
 			CustomHeader: cHeader,
 		}

+		var size int64
 		if isSSEC {
-			objectSize += partInfo.Size
-			pInfo, err = c.PutObjectPart(ctx, bucket, object, uploadID, partInfo.Number, hr, partInfo.Size, popts)
+			size = partInfo.Size
 		} else {
-			objectSize += partInfo.ActualSize
-			pInfo, err = c.PutObjectPart(ctx, bucket, object, uploadID, partInfo.Number, hr, partInfo.ActualSize, popts)
+			size = partInfo.ActualSize
 		}
+		objectSize += size
+		pInfo, err := c.PutObjectPart(ctx, bucket, object, uploadID, partInfo.Number, hr, size, popts)
 		if err != nil {
 			return err
 		}
-		if !isSSEC && pInfo.Size != partInfo.ActualSize {
-			return fmt.Errorf("Part size mismatch: got %d, want %d", pInfo.Size, partInfo.ActualSize)
+		if pInfo.Size != size {
+			return fmt.Errorf("ssec(%t): Part size mismatch: got %d, want %d", isSSEC, pInfo.Size, size)
 		}
 		uploadedParts = append(uploadedParts, minio.CompletePart{
-			PartNumber:     pInfo.PartNumber,
-			ETag:           pInfo.ETag,
-			ChecksumCRC32:  pInfo.ChecksumCRC32,
-			ChecksumCRC32C: pInfo.ChecksumCRC32C,
-			ChecksumSHA1:   pInfo.ChecksumSHA1,
-			ChecksumSHA256: pInfo.ChecksumSHA256,
+			PartNumber:        pInfo.PartNumber,
+			ETag:              pInfo.ETag,
+			ChecksumCRC32:     pInfo.ChecksumCRC32,
+			ChecksumCRC32C:    pInfo.ChecksumCRC32C,
+			ChecksumSHA1:      pInfo.ChecksumSHA1,
+			ChecksumSHA256:    pInfo.ChecksumSHA256,
+			ChecksumCRC64NVME: pInfo.ChecksumCRC64NVME,
 		})
 	}
 	userMeta := map[string]string{
-		validSSEReplicationHeaders[ReservedMetadataPrefix+"Actual-Object-Size"]: objInfo.UserDefined[ReservedMetadataPrefix+"actual-size"],
+		xhttp.MinIOReplicationActualObjectSize: objInfo.UserDefined[ReservedMetadataPrefix+"actual-size"],
 	}
 	if isSSEC && objInfo.UserDefined[ReplicationSsecChecksumHeader] != "" {
 		userMeta[ReplicationSsecChecksumHeader] = objInfo.UserDefined[ReplicationSsecChecksumHeader]
@@ -1708,6 +1714,12 @@ func replicateObjectWithMultipart(ctx context.Context, c *minio.Core, bucket, ob
 	// 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()
+
+	if len(objInfo.Checksum) > 0 {
+		for k, v := range getCRCMeta(objInfo, 0, nil) {
+			userMeta[k] = v
+		}
+	}
 	_, err = c.CompleteMultipartUpload(cctx, bucket, object, uploadID, uploadedParts, minio.PutObjectOptions{
 		UserMetadata: userMeta,
 		Internal: minio.AdvancedPutOptions{
@@ -1795,8 +1807,8 @@ const (
 )

 var (
-	globalReplicationPool  *ReplicationPool
-	globalReplicationStats *ReplicationStats
+	globalReplicationPool  = once.NewSingleton[ReplicationPool]()
+	globalReplicationStats atomic.Pointer[ReplicationStats]
 )

 // ReplicationPool describes replication pool
@@ -1811,6 +1823,7 @@ type ReplicationPool struct {
 	priority    string
 	maxWorkers  int
 	maxLWorkers int
+	stats       *ReplicationStats

 	mu       sync.RWMutex
 	mrfMU    sync.Mutex
@@ -1857,7 +1870,7 @@ 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
@@ -1899,6 +1912,7 @@ 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,
@@ -1926,11 +1940,11 @@ 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")
@@ -1958,9 +1972,9 @@ 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)
 				}
@@ -1968,10 +1982,10 @@ 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)
@@ -1998,9 +2012,9 @@ func (p *ReplicationPool) AddLargeWorker(input <-chan ReplicationWorkerOperation
 				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)
 				}
@@ -2162,9 +2176,9 @@ 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)
@@ -2195,7 +2209,7 @@ 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
@@ -2231,7 +2245,7 @@ func queueReplicateDeletesWrapper(doi DeletedObjectReplicationInfo, existingObje
 			doi.ResetID = v.ResetID
 			doi.TargetArn = k

-			globalReplicationPool.queueReplicaDeleteTask(doi)
+			globalReplicationPool.Get().queueReplicaDeleteTask(doi)
 		}
 	}
 }
@@ -2252,7 +2266,7 @@ 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
@@ -2282,9 +2296,10 @@ type replicationPoolOpts struct {
 }

 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 {
@@ -2490,7 +2505,7 @@ func scheduleReplication(ctx context.Context, oi ObjectInfo, o ObjectLayer, dsc
 	if dsc.Synchronous() {
 		replicateObject(ctx, ri, o)
 	} else {
-		globalReplicationPool.queueReplicaTask(ri)
+		globalReplicationPool.Get().queueReplicaTask(ri)
 	}
 }

@@ -2618,9 +2633,9 @@ 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(""))
 	}
 }

@@ -3050,9 +3065,9 @@ 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 {
@@ -3078,9 +3093,9 @@ 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,
@@ -3088,8 +3103,8 @@ func (s *replicationResyncer) start(ctx context.Context, objAPI ObjectLayer, opt
 		}
 	}
 	brs.TargetsMap[opts.arn] = status
-	globalReplicationPool.resyncer.statusMap[opts.bucket] = brs
-	go globalReplicationPool.resyncer.resyncBucket(GlobalContext, objAPI, false, opts)
+	globalReplicationPool.Get().resyncer.statusMap[opts.bucket] = brs
+	go globalReplicationPool.Get().resyncer.resyncBucket(GlobalContext, objAPI, false, opts)
 	return nil
 }

@@ -3421,7 +3436,7 @@ 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
@@ -3437,12 +3452,12 @@ 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
 }
@@ -3507,8 +3522,8 @@ 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
 	}

@@ -3521,8 +3536,8 @@ 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))
 		}
 	}
 }
@@ -3553,7 +3568,7 @@ func (p *ReplicationPool) persistToDrive(ctx context.Context, v MRFReplicateEntr
 	}

 	globalLocalDrivesMu.RLock()
-	localDrives := cloneDrives(globalLocalDrives)
+	localDrives := cloneDrives(globalLocalDrivesMap)
 	globalLocalDrivesMu.RUnlock()

 	for _, localDrive := range localDrives {
@@ -3571,7 +3586,7 @@ 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
 	}
@@ -3620,7 +3635,7 @@ func (p *ReplicationPool) loadMRF() (mrfRec MRFReplicateEntries, err error) {
 	}

 	globalLocalDrivesMu.RLock()
-	localDrives := cloneDrives(globalLocalDrives)
+	localDrives := cloneDrives(globalLocalDrivesMap)
 	globalLocalDrivesMu.RUnlock()

 	for _, localDrive := range localDrives {
@@ -3741,3 +3756,28 @@ 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
+}
+
+func getCRCMeta(oi ObjectInfo, partNum int, h http.Header) map[string]string {
+	meta := make(map[string]string)
+	cs := oi.decryptChecksums(partNum, h)
+	for k, v := range cs {
+		cksum := hash.NewChecksumString(k, v)
+		if cksum == nil {
+			continue
+		}
+		if cksum.Valid() {
+			meta[cksum.Type.Key()] = v
+		}
+		meta[xhttp.AmzChecksumType] = cksum.Type.ObjType()
+	}
+	return meta
+}
@@ -20,6 +20,7 @@ package cmd
 import (
 	"fmt"
 	"math"
+	"sync/atomic"
 	"time"

 	"github.com/minio/madmin-go/v3"
@@ -107,18 +108,18 @@ func (l ReplicationLastHour) merge(o ReplicationLastHour) (merged ReplicationLas

 // Add  a new duration data
 func (l *ReplicationLastHour) addsize(sz int64) {
-	min := time.Now().Unix() / 60
-	l.forwardTo(min)
-	winIdx := min % 60
-	l.Totals[winIdx].merge(AccElem{Total: min, Size: sz, N: 1})
-	l.LastMin = min
+	minutes := time.Now().Unix() / 60
+	l.forwardTo(minutes)
+	winIdx := minutes % 60
+	l.Totals[winIdx].merge(AccElem{Total: minutes, Size: sz, N: 1})
+	l.LastMin = minutes
 }

 // Merge all recorded counts of last hour into one
 func (l *ReplicationLastHour) getTotal() AccElem {
 	var res AccElem
-	min := time.Now().Unix() / 60
-	l.forwardTo(min)
+	minutes := time.Now().Unix() / 60
+	l.forwardTo(minutes)
 	for _, elem := range l.Totals[:] {
 		res.merge(elem)
 	}
@@ -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 +310,18 @@ type ReplQNodeStats struct {
 func (r *ReplicationStats) getNodeQueueStats(bucket string) (qs ReplQNodeStats) {
 	qs.NodeName = globalLocalNodeName
 	qs.Uptime = UTCNow().Unix() - globalBootTime.Unix()
-	qs.ActiveWorkers = globalReplicationStats.ActiveWorkers()
+	grs := globalReplicationStats.Load()
+	if grs != nil {
+		qs.ActiveWorkers = grs.ActiveWorkers()
+	} else {
+		qs.ActiveWorkers = ActiveWorkerStat{}
+	}
 	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 +407,12 @@ 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()
@@ -91,9 +91,6 @@ func init() {
 	logger.Init(GOPATH, GOROOT)
 	logger.RegisterError(config.FmtError)

-	globalBatchJobsMetrics = batchJobMetrics{metrics: make(map[string]*batchJobInfo)}
-	go globalBatchJobsMetrics.purgeJobMetrics()
-
 	t, _ := minioVersionToReleaseTime(Version)
 	if !t.IsZero() {
 		globalVersionUnix = uint64(t.Unix())
@@ -138,6 +135,9 @@ func minioConfigToConsoleFeatures() {
 	if value := env.Get(config.EnvBrowserRedirectURL, ""); value != "" {
 		os.Setenv("CONSOLE_BROWSER_REDIRECT_URL", value)
 	}
+	if value := env.Get(config.EnvConsoleDebugLogLevel, ""); value != "" {
+		os.Setenv("CONSOLE_DEBUG_LOGLEVEL", value)
+	}
 	// pass the console subpath configuration
 	if globalBrowserRedirectURL != nil {
 		subPath := path.Clean(pathJoin(strings.TrimSpace(globalBrowserRedirectURL.Path), SlashSeparator))
@@ -461,25 +461,27 @@ func handleCommonArgs(ctxt serverCtxt) {
 	certsDir := ctxt.CertsDir
 	certsSet := ctxt.certsDirSet

-	if consoleAddr == "" {
-		p, err := xnet.GetFreePort()
-		if err != nil {
-			logger.FatalIf(err, "Unable to get free port for Console UI on the host")
+	if globalBrowserEnabled {
+		if consoleAddr == "" {
+			p, err := xnet.GetFreePort()
+			if err != nil {
+				logger.FatalIf(err, "Unable to get free port for Console UI on the host")
+			}
+			// hold the port
+			l, err := net.Listen("TCP", fmt.Sprintf(":%s", p.String()))
+			if err == nil {
+				defer l.Close()
+			}
+			consoleAddr = net.JoinHostPort("", p.String())
 		}
-		// hold the port
-		l, err := net.Listen("TCP", fmt.Sprintf(":%s", p.String()))
-		if err == nil {
-			defer l.Close()
+
+		if _, _, err := net.SplitHostPort(consoleAddr); err != nil {
+			logger.FatalIf(err, "Unable to start listening on console port")
 		}
-		consoleAddr = net.JoinHostPort("", p.String())
-	}

-	if _, _, err := net.SplitHostPort(consoleAddr); err != nil {
-		logger.FatalIf(err, "Unable to start listening on console port")
-	}
-
-	if consoleAddr == addr {
-		logger.FatalIf(errors.New("--console-address cannot be same as --address"), "Unable to start the server")
+		if consoleAddr == addr {
+			logger.FatalIf(errors.New("--console-address cannot be same as --address"), "Unable to start the server")
+		}
 	}

 	globalMinioHost, globalMinioPort = mustSplitHostPort(addr)
@@ -492,7 +494,9 @@ func handleCommonArgs(ctxt serverCtxt) {
 		globalDynamicAPIPort = true
 	}

-	globalMinioConsoleHost, globalMinioConsolePort = mustSplitHostPort(consoleAddr)
+	if globalBrowserEnabled {
+		globalMinioConsoleHost, globalMinioConsolePort = mustSplitHostPort(consoleAddr)
+	}

 	if globalMinioPort == globalMinioConsolePort {
 		logger.FatalIf(errors.New("--console-address port cannot be same as --address port"), "Unable to start the server")
@@ -685,16 +689,6 @@ 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) {
@@ -706,12 +700,16 @@ func loadEnvVarsFromFiles() {
 	}
 }

-func serverHandleEnvVars() {
+func serverHandleEarlyEnvVars() {
 	var err error
 	globalBrowserEnabled, err = config.ParseBool(env.Get(config.EnvBrowser, config.EnableOn))
 	if err != nil {
 		logger.Fatal(config.ErrInvalidBrowserValue(err), "Invalid MINIO_BROWSER value in environment variable")
 	}
+}
+
+func serverHandleEnvVars() {
+	var err error
 	if globalBrowserEnabled {
 		if redirectURL := env.Get(config.EnvBrowserRedirectURL, ""); redirectURL != "" {
 			u, err := xnet.ParseHTTPURL(redirectURL)
@@ -768,7 +766,7 @@ 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)
@@ -777,7 +775,7 @@ 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")
 			}
 		}
@@ -881,6 +879,12 @@ 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.
@@ -34,7 +34,6 @@ import (
 	"github.com/minio/minio/internal/config"
 	"github.com/minio/minio/internal/config/api"
 	"github.com/minio/minio/internal/config/batch"
-	"github.com/minio/minio/internal/config/cache"
 	"github.com/minio/minio/internal/config/callhome"
 	"github.com/minio/minio/internal/config/compress"
 	"github.com/minio/minio/internal/config/dns"
@@ -80,7 +79,6 @@ func initHelp() {
 		config.CallhomeSubSys:       callhome.DefaultKVS,
 		config.DriveSubSys:          drive.DefaultKVS,
 		config.ILMSubSys:            ilm.DefaultKVS,
-		config.CacheSubSys:          cache.DefaultKVS,
 		config.BatchSubSys:          batch.DefaultKVS,
 		config.BrowserSubSys:        browser.DefaultKVS,
 	}
@@ -229,12 +227,6 @@ func initHelp() {
 			Key:         config.EtcdSubSys,
 			Description: "persist IAM assets externally to etcd",
 		},
-		config.HelpKV{
-			Key:         config.CacheSubSys,
-			Type:        "string",
-			Description: "enable cache plugin on MinIO for GET/HEAD requests",
-			Optional:    true,
-		},
 		config.HelpKV{
 			Key:         config.BrowserSubSys,
 			Description: "manage Browser HTTP specific features, such as Security headers, etc.",
@@ -291,7 +283,6 @@ func initHelp() {
 		config.SubnetSubSys:         subnet.HelpSubnet,
 		config.CallhomeSubSys:       callhome.HelpCallhome,
 		config.DriveSubSys:          drive.HelpDrive,
-		config.CacheSubSys:          cache.Help,
 		config.BrowserSubSys:        browser.Help,
 		config.ILMSubSys:            ilm.Help,
 	}
@@ -409,10 +400,6 @@ func validateSubSysConfig(ctx context.Context, s config.Config, subSys string, o
 		if _, err := drive.LookupConfig(s[config.DriveSubSys][config.Default]); err != nil {
 			return err
 		}
-	case config.CacheSubSys:
-		if _, err := cache.LookupConfig(s[config.CacheSubSys][config.Default], globalRemoteTargetTransport); err != nil {
-			return err
-		}
 	case config.PolicyOPASubSys:
 		// In case legacy OPA config is being set, we treat it as if the
 		// AuthZPlugin is being set.
@@ -710,13 +697,6 @@ func applyDynamicConfigForSubSys(ctx context.Context, objAPI ObjectLayer, s conf
 				configLogIf(ctx, fmt.Errorf("Unable to update drive config: %v", err))
 			}
 		}
-	case config.CacheSubSys:
-		cacheCfg, err := cache.LookupConfig(s[config.CacheSubSys][config.Default], globalRemoteTargetTransport)
-		if err != nil {
-			configLogIf(ctx, fmt.Errorf("Unable to load cache config: %w", err))
-		} else {
-			globalCacheConfig.Update(cacheCfg)
-		}
 	case config.BrowserSubSys:
 		browserCfg, err := browser.LookupConfig(s[config.BrowserSubSys][config.Default])
 		if err != nil {
@@ -569,7 +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),
 				},
 			})
@@ -1138,7 +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)),
 			},
 		})
@@ -1170,7 +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),
 			},
@@ -1325,10 +1325,13 @@ func applyExpiryOnNonTransitionedObjects(ctx context.Context, objLayer ObjectLay
 	dobj, err = objLayer.DeleteObject(ctx, obj.Bucket, encodeDirObject(obj.Name), opts)
 	if err != nil {
 		if isErrObjectNotFound(err) || isErrVersionNotFound(err) {
+			traceFn(ILMExpiry, nil, nil)
 			return false
 		}
 		// Assume it is still there.
-		ilmLogOnceIf(ctx, fmt.Errorf("DeleteObject(%s, %s): %w", obj.Bucket, obj.Name, err), "non-transition-expiry"+obj.Name)
+		err := fmt.Errorf("DeleteObject(%s, %s): %w", obj.Bucket, obj.Name, err)
+		ilmLogOnceIf(ctx, err, "non-transition-expiry"+obj.Name)
+		traceFn(ILMExpiry, nil, err)
 		return false
 	}
 	if dobj.Name == "" {
@@ -1336,6 +1339,8 @@ 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)

@@ -1547,7 +1552,7 @@ 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, err error)) {
 	var apiName string
 	switch event {
 	case ILMExpiry:
@@ -1565,5 +1570,5 @@ func auditLogLifecycle(ctx context.Context, oi ObjectInfo, event string, tags ma
 		VersionID: oi.VersionID,
 		Tags:      tags,
 	})
-	traceFn(event)
+	traceFn(event, tags, nil)
 }
@@ -39,6 +39,8 @@ import (
 	"github.com/valyala/bytebufferpool"
 )

+//msgp:clearomitted
+
 //go:generate msgp -file $GOFILE -unexported

 // dataUsageHash is the hash type used.
@@ -57,15 +59,14 @@ type versionsHistogram [dataUsageVersionLen]uint64
 type dataUsageEntry struct {
 	Children dataUsageHashMap `msg:"ch"`
 	// These fields do no include any children.
-	Size             int64                `msg:"sz"`
-	Objects          uint64               `msg:"os"`
-	Versions         uint64               `msg:"vs"` // Versions that are not delete markers.
-	DeleteMarkers    uint64               `msg:"dms"`
-	ObjSizes         sizeHistogram        `msg:"szs"`
-	ObjVersions      versionsHistogram    `msg:"vh"`
-	ReplicationStats *replicationAllStats `msg:"rs,omitempty"`
-	AllTierStats     *allTierStats        `msg:"ats,omitempty"`
-	Compacted        bool                 `msg:"c"`
+	Size          int64             `msg:"sz"`
+	Objects       uint64            `msg:"os"`
+	Versions      uint64            `msg:"vs"` // Versions that are not delete markers.
+	DeleteMarkers uint64            `msg:"dms"`
+	ObjSizes      sizeHistogram     `msg:"szs"`
+	ObjVersions   versionsHistogram `msg:"vh"`
+	AllTierStats  *allTierStats     `msg:"ats,omitempty"`
+	Compacted     bool              `msg:"c"`
 }

 // allTierStats is a collection of per-tier stats across all configured remote
@@ -135,93 +136,6 @@ func (ts tierStats) add(u tierStats) tierStats {
 	}
 }

-//msgp:tuple replicationStatsV1
-type replicationStatsV1 struct {
-	PendingSize          uint64
-	ReplicatedSize       uint64
-	FailedSize           uint64
-	ReplicaSize          uint64
-	FailedCount          uint64
-	PendingCount         uint64
-	MissedThresholdSize  uint64
-	AfterThresholdSize   uint64
-	MissedThresholdCount uint64
-	AfterThresholdCount  uint64
-}
-
-func (rsv1 replicationStatsV1) Empty() bool {
-	return rsv1.ReplicatedSize == 0 &&
-		rsv1.FailedSize == 0 &&
-		rsv1.FailedCount == 0
-}
-
-//msgp:tuple replicationStats
-type replicationStats struct {
-	PendingSize          uint64
-	ReplicatedSize       uint64
-	FailedSize           uint64
-	FailedCount          uint64
-	PendingCount         uint64
-	MissedThresholdSize  uint64
-	AfterThresholdSize   uint64
-	MissedThresholdCount uint64
-	AfterThresholdCount  uint64
-	ReplicatedCount      uint64
-}
-
-func (rs replicationStats) Empty() bool {
-	return rs.ReplicatedSize == 0 &&
-		rs.FailedSize == 0 &&
-		rs.FailedCount == 0
-}
-
-type replicationAllStats struct {
-	Targets      map[string]replicationStats `msg:"t,omitempty"`
-	ReplicaSize  uint64                      `msg:"r,omitempty"`
-	ReplicaCount uint64                      `msg:"rc,omitempty"`
-}
-
-//msgp:tuple replicationAllStatsV1
-type replicationAllStatsV1 struct {
-	Targets      map[string]replicationStats
-	ReplicaSize  uint64 `msg:"ReplicaSize,omitempty"`
-	ReplicaCount uint64 `msg:"ReplicaCount,omitempty"`
-}
-
-// empty returns true if the replicationAllStats is empty (contains no entries).
-func (r *replicationAllStats) empty() bool {
-	if r == nil {
-		return true
-	}
-	if r.ReplicaSize != 0 || r.ReplicaCount != 0 {
-		return false
-	}
-	for _, v := range r.Targets {
-		if !v.Empty() {
-			return false
-		}
-	}
-	return true
-}
-
-// clone creates a deep-copy clone.
-func (r *replicationAllStats) clone() *replicationAllStats {
-	if r == nil {
-		return nil
-	}
-
-	// Shallow copy
-	dst := *r
-
-	// Copy individual targets.
-	dst.Targets = make(map[string]replicationStats, len(r.Targets))
-	for k, v := range r.Targets {
-		dst.Targets[k] = v
-	}
-
-	return &dst
-}
-
 //msgp:encode ignore dataUsageEntryV2 dataUsageEntryV3 dataUsageEntryV4 dataUsageEntryV5 dataUsageEntryV6 dataUsageEntryV7
 //msgp:marshal ignore dataUsageEntryV2 dataUsageEntryV3 dataUsageEntryV4 dataUsageEntryV5 dataUsageEntryV6 dataUsageEntryV7

@@ -237,62 +151,54 @@ type dataUsageEntryV2 struct {
 //msgp:tuple dataUsageEntryV3
 type dataUsageEntryV3 struct {
 	// These fields do no include any children.
-	Size                   int64
-	ReplicatedSize         uint64
-	ReplicationPendingSize uint64
-	ReplicationFailedSize  uint64
-	ReplicaSize            uint64
-	Objects                uint64
-	ObjSizes               sizeHistogram
-	Children               dataUsageHashMap
+	Size     int64
+	Objects  uint64
+	ObjSizes sizeHistogram
+	Children dataUsageHashMap
 }

 //msgp:tuple dataUsageEntryV4
 type dataUsageEntryV4 struct {
 	Children dataUsageHashMap
 	// These fields do no include any children.
-	Size             int64
-	Objects          uint64
-	ObjSizes         sizeHistogram
-	ReplicationStats replicationStatsV1
+	Size     int64
+	Objects  uint64
+	ObjSizes sizeHistogram
 }

 //msgp:tuple dataUsageEntryV5
 type dataUsageEntryV5 struct {
 	Children dataUsageHashMap
 	// These fields do no include any children.
-	Size             int64
-	Objects          uint64
-	Versions         uint64 // Versions that are not delete markers.
-	ObjSizes         sizeHistogram
-	ReplicationStats *replicationStatsV1
-	Compacted        bool
+	Size      int64
+	Objects   uint64
+	Versions  uint64 // Versions that are not delete markers.
+	ObjSizes  sizeHistogram
+	Compacted bool
 }

 //msgp:tuple dataUsageEntryV6
 type dataUsageEntryV6 struct {
 	Children dataUsageHashMap
 	// These fields do no include any children.
-	Size             int64
-	Objects          uint64
-	Versions         uint64 // Versions that are not delete markers.
-	ObjSizes         sizeHistogram
-	ReplicationStats *replicationAllStatsV1
-	Compacted        bool
+	Size      int64
+	Objects   uint64
+	Versions  uint64 // Versions that are not delete markers.
+	ObjSizes  sizeHistogram
+	Compacted bool
 }

 type dataUsageEntryV7 struct {
 	Children dataUsageHashMap `msg:"ch"`
 	// These fields do no include any children.
-	Size             int64                `msg:"sz"`
-	Objects          uint64               `msg:"os"`
-	Versions         uint64               `msg:"vs"` // Versions that are not delete markers.
-	DeleteMarkers    uint64               `msg:"dms"`
-	ObjSizes         sizeHistogramV1      `msg:"szs"`
-	ObjVersions      versionsHistogram    `msg:"vh"`
-	ReplicationStats *replicationAllStats `msg:"rs,omitempty"`
-	AllTierStats     *allTierStats        `msg:"ats,omitempty"`
-	Compacted        bool                 `msg:"c"`
+	Size          int64             `msg:"sz"`
+	Objects       uint64            `msg:"os"`
+	Versions      uint64            `msg:"vs"` // Versions that are not delete markers.
+	DeleteMarkers uint64            `msg:"dms"`
+	ObjSizes      sizeHistogramV1   `msg:"szs"`
+	ObjVersions   versionsHistogram `msg:"vh"`
+	AllTierStats  *allTierStats     `msg:"ats,omitempty"`
+	Compacted     bool              `msg:"c"`
 }

 // dataUsageCache contains a cache of data usage entries latest version.
@@ -373,29 +279,6 @@ func (e *dataUsageEntry) addSizes(summary sizeSummary) {
 	e.ObjSizes.add(summary.totalSize)
 	e.ObjVersions.add(summary.versions)

-	if e.ReplicationStats == nil {
-		e.ReplicationStats = &replicationAllStats{
-			Targets: make(map[string]replicationStats),
-		}
-	} else if e.ReplicationStats.Targets == nil {
-		e.ReplicationStats.Targets = make(map[string]replicationStats)
-	}
-	e.ReplicationStats.ReplicaSize += uint64(summary.replicaSize)
-	e.ReplicationStats.ReplicaCount += uint64(summary.replicaCount)
-
-	for arn, st := range summary.replTargetStats {
-		tgtStat, ok := e.ReplicationStats.Targets[arn]
-		if !ok {
-			tgtStat = replicationStats{}
-		}
-		tgtStat.PendingSize += uint64(st.pendingSize)
-		tgtStat.FailedSize += uint64(st.failedSize)
-		tgtStat.ReplicatedSize += uint64(st.replicatedSize)
-		tgtStat.ReplicatedCount += uint64(st.replicatedCount)
-		tgtStat.FailedCount += st.failedCount
-		tgtStat.PendingCount += st.pendingCount
-		e.ReplicationStats.Targets[arn] = tgtStat
-	}
 	if len(summary.tiers) != 0 {
 		if e.AllTierStats == nil {
 			e.AllTierStats = newAllTierStats()
@@ -410,26 +293,6 @@ func (e *dataUsageEntry) merge(other dataUsageEntry) {
 	e.Versions += other.Versions
 	e.DeleteMarkers += other.DeleteMarkers
 	e.Size += other.Size
-	if other.ReplicationStats != nil {
-		if e.ReplicationStats == nil {
-			e.ReplicationStats = &replicationAllStats{Targets: make(map[string]replicationStats)}
-		} else if e.ReplicationStats.Targets == nil {
-			e.ReplicationStats.Targets = make(map[string]replicationStats)
-		}
-		e.ReplicationStats.ReplicaSize += other.ReplicationStats.ReplicaSize
-		e.ReplicationStats.ReplicaCount += other.ReplicationStats.ReplicaCount
-		for arn, stat := range other.ReplicationStats.Targets {
-			st := e.ReplicationStats.Targets[arn]
-			e.ReplicationStats.Targets[arn] = replicationStats{
-				PendingSize:     stat.PendingSize + st.PendingSize,
-				FailedSize:      stat.FailedSize + st.FailedSize,
-				ReplicatedSize:  stat.ReplicatedSize + st.ReplicatedSize,
-				PendingCount:    stat.PendingCount + st.PendingCount,
-				FailedCount:     stat.FailedCount + st.FailedCount,
-				ReplicatedCount: stat.ReplicatedCount + st.ReplicatedCount,
-			}
-		}
-	}

 	for i, v := range other.ObjSizes[:] {
 		e.ObjSizes[i] += v
@@ -490,10 +353,7 @@ func (e dataUsageEntry) clone() dataUsageEntry {
 		}
 		e.Children = ch
 	}
-	if e.ReplicationStats != nil {
-		// Clone ReplicationStats
-		e.ReplicationStats = e.ReplicationStats.clone()
-	}
+
 	if e.AllTierStats != nil {
 		e.AllTierStats = e.AllTierStats.clone()
 	}
@@ -931,22 +791,6 @@ func (d *dataUsageCache) bucketsUsageInfo(buckets []BucketInfo) map[string]Bucke
 			ObjectSizesHistogram:    flat.ObjSizes.toMap(),
 			ObjectVersionsHistogram: flat.ObjVersions.toMap(),
 		}
-		if flat.ReplicationStats != nil {
-			bui.ReplicaSize = flat.ReplicationStats.ReplicaSize
-			bui.ReplicaCount = flat.ReplicationStats.ReplicaCount
-
-			bui.ReplicationInfo = make(map[string]BucketTargetUsageInfo, len(flat.ReplicationStats.Targets))
-			for arn, stat := range flat.ReplicationStats.Targets {
-				bui.ReplicationInfo[arn] = BucketTargetUsageInfo{
-					ReplicationPendingSize:  stat.PendingSize,
-					ReplicatedSize:          stat.ReplicatedSize,
-					ReplicationFailedSize:   stat.FailedSize,
-					ReplicationPendingCount: stat.PendingCount,
-					ReplicationFailedCount:  stat.FailedCount,
-					ReplicatedCount:         stat.ReplicatedCount,
-				}
-			}
-		}
 		dst[bucket.Name] = bui
 	}
 	return dst
@@ -959,9 +803,6 @@ func (d *dataUsageCache) sizeRecursive(path string) *dataUsageEntry {
 		return root
 	}
 	flat := d.flatten(*root)
-	if flat.ReplicationStats.empty() {
-		flat.ReplicationStats = nil
-	}
 	return &flat
 }

@@ -1238,20 +1079,6 @@ func (d *dataUsageCache) deserialize(r io.Reader) error {
 				ObjSizes: v.ObjSizes,
 				Children: v.Children,
 			}
-			if v.ReplicatedSize > 0 || v.ReplicaSize > 0 || v.ReplicationFailedSize > 0 || v.ReplicationPendingSize > 0 {
-				cfg, _ := getReplicationConfig(GlobalContext, d.Info.Name)
-				if cfg != nil && cfg.RoleArn != "" {
-					due.ReplicationStats = &replicationAllStats{
-						Targets: make(map[string]replicationStats),
-					}
-					due.ReplicationStats.ReplicaSize = v.ReplicaSize
-					due.ReplicationStats.Targets[cfg.RoleArn] = replicationStats{
-						ReplicatedSize: v.ReplicatedSize,
-						FailedSize:     v.ReplicationFailedSize,
-						PendingSize:    v.ReplicationPendingSize,
-					}
-				}
-			}
 			due.Compacted = len(due.Children) == 0 && k != d.Info.Name

 			d.Cache[k] = due
@@ -1277,36 +1104,10 @@ func (d *dataUsageCache) deserialize(r io.Reader) error {
 				ObjSizes: v.ObjSizes,
 				Children: v.Children,
 			}
-			empty := replicationStatsV1{}
-
-			if v.ReplicationStats != empty {
-				cfg, _ := getReplicationConfig(GlobalContext, d.Info.Name)
-				if cfg != nil && cfg.RoleArn != "" {
-					due.ReplicationStats = &replicationAllStats{
-						Targets: make(map[string]replicationStats),
-					}
-					due.ReplicationStats.Targets[cfg.RoleArn] = replicationStats{
-						ReplicatedSize: v.ReplicationStats.ReplicatedSize,
-						FailedSize:     v.ReplicationStats.FailedSize,
-						FailedCount:    v.ReplicationStats.FailedCount,
-						PendingSize:    v.ReplicationStats.PendingSize,
-						PendingCount:   v.ReplicationStats.PendingCount,
-					}
-					due.ReplicationStats.ReplicaSize = v.ReplicationStats.ReplicaSize
-				}
-			}
 			due.Compacted = len(due.Children) == 0 && k != d.Info.Name

 			d.Cache[k] = due
 		}
-
-		// Populate compacted value and remove unneeded replica stats.
-		for k, e := range d.Cache {
-			if e.ReplicationStats != nil && len(e.ReplicationStats.Targets) == 0 {
-				e.ReplicationStats = nil
-			}
-			d.Cache[k] = e
-		}
 		return nil
 	case dataUsageCacheVerV5:
 		// Zstd compressed.
@@ -1328,36 +1129,10 @@ func (d *dataUsageCache) deserialize(r io.Reader) error {
 				ObjSizes: v.ObjSizes,
 				Children: v.Children,
 			}
-			if v.ReplicationStats != nil && !v.ReplicationStats.Empty() {
-				cfg, _ := getReplicationConfig(GlobalContext, d.Info.Name)
-				if cfg != nil && cfg.RoleArn != "" {
-					due.ReplicationStats = &replicationAllStats{
-						Targets: make(map[string]replicationStats),
-					}
-					d.Info.replication = replicationConfig{Config: cfg}
-
-					due.ReplicationStats.Targets[cfg.RoleArn] = replicationStats{
-						ReplicatedSize: v.ReplicationStats.ReplicatedSize,
-						FailedSize:     v.ReplicationStats.FailedSize,
-						FailedCount:    v.ReplicationStats.FailedCount,
-						PendingSize:    v.ReplicationStats.PendingSize,
-						PendingCount:   v.ReplicationStats.PendingCount,
-					}
-					due.ReplicationStats.ReplicaSize = v.ReplicationStats.ReplicaSize
-				}
-			}
 			due.Compacted = len(due.Children) == 0 && k != d.Info.Name

 			d.Cache[k] = due
 		}
-
-		// Populate compacted value and remove unneeded replica stats.
-		for k, e := range d.Cache {
-			if e.ReplicationStats != nil && len(e.ReplicationStats.Targets) == 0 {
-				e.ReplicationStats = nil
-			}
-			d.Cache[k] = e
-		}
 		return nil
 	case dataUsageCacheVerV6:
 		// Zstd compressed.
@@ -1373,22 +1148,13 @@ func (d *dataUsageCache) deserialize(r io.Reader) error {
 		d.Info = dold.Info
 		d.Cache = make(map[string]dataUsageEntry, len(dold.Cache))
 		for k, v := range dold.Cache {
-			var replicationStats *replicationAllStats
-			if v.ReplicationStats != nil {
-				replicationStats = &replicationAllStats{
-					Targets:      v.ReplicationStats.Targets,
-					ReplicaSize:  v.ReplicationStats.ReplicaSize,
-					ReplicaCount: v.ReplicationStats.ReplicaCount,
-				}
-			}
 			due := dataUsageEntry{
-				Children:         v.Children,
-				Size:             v.Size,
-				Objects:          v.Objects,
-				Versions:         v.Versions,
-				ObjSizes:         v.ObjSizes,
-				ReplicationStats: replicationStats,
-				Compacted:        v.Compacted,
+				Children:  v.Children,
+				Size:      v.Size,
+				Objects:   v.Objects,
+				Versions:  v.Versions,
+				ObjSizes:  v.ObjSizes,
+				Compacted: v.Compacted,
 			}
 			d.Cache[k] = due
 		}
@@ -1410,13 +1176,12 @@ func (d *dataUsageCache) deserialize(r io.Reader) error {
 			var szHist sizeHistogram
 			szHist.mergeV1(v.ObjSizes)
 			d.Cache[k] = dataUsageEntry{
-				Children:         v.Children,
-				Size:             v.Size,
-				Objects:          v.Objects,
-				Versions:         v.Versions,
-				ObjSizes:         szHist,
-				ReplicationStats: v.ReplicationStats,
-				Compacted:        v.Compacted,
+				Children:  v.Children,
+				Size:      v.Size,
+				Objects:   v.Objects,
+				Versions:  v.Versions,
+				ObjSizes:  szHist,
+				Compacted: v.Compacted,
 			}
 		}

@@ -519,458 +519,6 @@ func BenchmarkDecodedataUsageEntry(b *testing.B) {
 	}
 }

-func TestMarshalUnmarshalreplicationAllStats(t *testing.T) {
-	v := replicationAllStats{}
-	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 BenchmarkMarshalMsgreplicationAllStats(b *testing.B) {
-	v := replicationAllStats{}
-	b.ReportAllocs()
-	b.ResetTimer()
-	for i := 0; i < b.N; i++ {
-		v.MarshalMsg(nil)
-	}
-}
-
-func BenchmarkAppendMsgreplicationAllStats(b *testing.B) {
-	v := replicationAllStats{}
-	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 BenchmarkUnmarshalreplicationAllStats(b *testing.B) {
-	v := replicationAllStats{}
-	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 TestEncodeDecodereplicationAllStats(t *testing.T) {
-	v := replicationAllStats{}
-	var buf bytes.Buffer
-	msgp.Encode(&buf, &v)
-
-	m := v.Msgsize()
-	if buf.Len() > m {
-		t.Log("WARNING: TestEncodeDecodereplicationAllStats Msgsize() is inaccurate")
-	}
-
-	vn := replicationAllStats{}
-	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 BenchmarkEncodereplicationAllStats(b *testing.B) {
-	v := replicationAllStats{}
-	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 BenchmarkDecodereplicationAllStats(b *testing.B) {
-	v := replicationAllStats{}
-	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 TestMarshalUnmarshalreplicationAllStatsV1(t *testing.T) {
-	v := replicationAllStatsV1{}
-	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 BenchmarkMarshalMsgreplicationAllStatsV1(b *testing.B) {
-	v := replicationAllStatsV1{}
-	b.ReportAllocs()
-	b.ResetTimer()
-	for i := 0; i < b.N; i++ {
-		v.MarshalMsg(nil)
-	}
-}
-
-func BenchmarkAppendMsgreplicationAllStatsV1(b *testing.B) {
-	v := replicationAllStatsV1{}
-	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 BenchmarkUnmarshalreplicationAllStatsV1(b *testing.B) {
-	v := replicationAllStatsV1{}
-	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 TestEncodeDecodereplicationAllStatsV1(t *testing.T) {
-	v := replicationAllStatsV1{}
-	var buf bytes.Buffer
-	msgp.Encode(&buf, &v)
-
-	m := v.Msgsize()
-	if buf.Len() > m {
-		t.Log("WARNING: TestEncodeDecodereplicationAllStatsV1 Msgsize() is inaccurate")
-	}
-
-	vn := replicationAllStatsV1{}
-	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 BenchmarkEncodereplicationAllStatsV1(b *testing.B) {
-	v := replicationAllStatsV1{}
-	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 BenchmarkDecodereplicationAllStatsV1(b *testing.B) {
-	v := replicationAllStatsV1{}
-	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 TestMarshalUnmarshalreplicationStats(t *testing.T) {
-	v := replicationStats{}
-	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 BenchmarkMarshalMsgreplicationStats(b *testing.B) {
-	v := replicationStats{}
-	b.ReportAllocs()
-	b.ResetTimer()
-	for i := 0; i < b.N; i++ {
-		v.MarshalMsg(nil)
-	}
-}
-
-func BenchmarkAppendMsgreplicationStats(b *testing.B) {
-	v := replicationStats{}
-	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 BenchmarkUnmarshalreplicationStats(b *testing.B) {
-	v := replicationStats{}
-	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 TestEncodeDecodereplicationStats(t *testing.T) {
-	v := replicationStats{}
-	var buf bytes.Buffer
-	msgp.Encode(&buf, &v)
-
-	m := v.Msgsize()
-	if buf.Len() > m {
-		t.Log("WARNING: TestEncodeDecodereplicationStats Msgsize() is inaccurate")
-	}
-
-	vn := replicationStats{}
-	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 BenchmarkEncodereplicationStats(b *testing.B) {
-	v := replicationStats{}
-	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 BenchmarkDecodereplicationStats(b *testing.B) {
-	v := replicationStats{}
-	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 TestMarshalUnmarshalreplicationStatsV1(t *testing.T) {
-	v := replicationStatsV1{}
-	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 BenchmarkMarshalMsgreplicationStatsV1(b *testing.B) {
-	v := replicationStatsV1{}
-	b.ReportAllocs()
-	b.ResetTimer()
-	for i := 0; i < b.N; i++ {
-		v.MarshalMsg(nil)
-	}
-}
-
-func BenchmarkAppendMsgreplicationStatsV1(b *testing.B) {
-	v := replicationStatsV1{}
-	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 BenchmarkUnmarshalreplicationStatsV1(b *testing.B) {
-	v := replicationStatsV1{}
-	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 TestEncodeDecodereplicationStatsV1(t *testing.T) {
-	v := replicationStatsV1{}
-	var buf bytes.Buffer
-	msgp.Encode(&buf, &v)
-
-	m := v.Msgsize()
-	if buf.Len() > m {
-		t.Log("WARNING: TestEncodeDecodereplicationStatsV1 Msgsize() is inaccurate")
-	}
-
-	vn := replicationStatsV1{}
-	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 BenchmarkEncodereplicationStatsV1(b *testing.B) {
-	v := replicationStatsV1{}
-	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 BenchmarkDecodereplicationStatsV1(b *testing.B) {
-	v := replicationStatsV1{}
-	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 TestMarshalUnmarshalsizeHistogram(t *testing.T) {
 	v := sizeHistogram{}
 	bts, err := v.MarshalMsg(nil)
@@ -587,17 +587,6 @@ func TestDataUsageCacheSerialize(t *testing.T) {
 		t.Fatal(err)
 	}
 	e := want.find("abucket/dir2")
-	e.ReplicationStats = &replicationAllStats{
-		Targets: map[string]replicationStats{
-			"arn": {
-				PendingSize:    1,
-				ReplicatedSize: 2,
-				FailedSize:     3,
-				FailedCount:    5,
-				PendingCount:   6,
-			},
-		},
-	}
 	want.replace("abucket/dir2", "", *e)
 	var buf bytes.Buffer
 	err = want.serializeTo(&buf)
@@ -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)
+}
@@ -117,12 +117,12 @@ func (dt *dynamicTimeout) logEntry(duration time.Duration) {
 // adjust changes the value of the dynamic timeout based on the
 // previous results
 func (dt *dynamicTimeout) adjust(entries [dynamicTimeoutLogSize]time.Duration) {
-	failures, max := 0, time.Duration(0)
+	failures, maxDur := 0, time.Duration(0)
 	for _, dur := range entries[:] {
 		if dur == maxDuration {
 			failures++
-		} else if dur > max {
-			max = dur
+		} else if dur > maxDur {
+			maxDur = dur
 		}
 	}

@@ -144,12 +144,12 @@ func (dt *dynamicTimeout) adjust(entries [dynamicTimeoutLogSize]time.Duration) {
 	} else if failPct < dynamicTimeoutDecreaseThresholdPct {
 		// We are hitting the timeout relatively few times,
 		// so decrease the timeout towards 25 % of maximum time spent.
-		max = max * 125 / 100
+		maxDur = maxDur * 125 / 100

 		timeout := atomic.LoadInt64(&dt.timeout)
-		if max < time.Duration(timeout) {
+		if maxDur < time.Duration(timeout) {
 			// Move 50% toward the max.
-			timeout = (int64(max) + timeout) / 2
+			timeout = (int64(maxDur) + timeout) / 2
 		}
 		if timeout < dt.minimum {
 			timeout = dt.minimum
@@ -109,7 +109,7 @@ func kmsKeyIDFromMetadata(metadata map[string]string) string {
 // be AWS S3 compliant.
 //
 // DecryptETags uses a KMS bulk decryption API, if available, which
-// is more efficient than decrypting ETags sequentually.
+// is more efficient than decrypting ETags sequentially.
 func DecryptETags(ctx context.Context, k *kms.KMS, objects []ObjectInfo) error {
 	const BatchSize = 250 // We process the objects in batches - 250 is a reasonable default.
 	var (
@@ -601,28 +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
 }

@@ -638,48 +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)

@@ -743,8 +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 {
@@ -757,7 +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)
 	}
@@ -1103,7 +1089,7 @@ func (o *ObjectInfo) metadataDecrypter(h http.Header) objectMetaDecryptFn {
 	}
 }

-// decryptChecksums will attempt to decode checksums and return it/them if set.
+// decryptPartsChecksums will attempt to decode checksums and return it/them if set.
 // if part > 0, and we have the checksum for the part that will be returned.
 func (o *ObjectInfo) decryptPartsChecksums(h http.Header) {
 	data := o.Checksum
@@ -1113,7 +1099,9 @@ func (o *ObjectInfo) decryptPartsChecksums(h http.Header) {
 	if _, encrypted := crypto.IsEncrypted(o.UserDefined); encrypted {
 		decrypted, err := o.metadataDecrypter(h)("object-checksum", data)
 		if err != nil {
-			encLogIf(GlobalContext, err)
+			if !errors.Is(err, crypto.ErrSecretKeyMismatch) {
+				encLogIf(GlobalContext, err)
+			}
 			return
 		}
 		data = decrypted
@@ -1172,6 +1160,14 @@ func (o *ObjectInfo) decryptChecksums(part int, h http.Header) map[string]string
 	if len(data) == 0 {
 		return nil
 	}
+	if part > 0 && !crypto.SSEC.IsEncrypted(o.UserDefined) {
+		// already decrypted in ToObjectInfo for multipart objects
+		for _, pi := range o.Parts {
+			if pi.Number == part {
+				return pi.Checksums
+			}
+		}
+	}
 	if _, encrypted := crypto.IsEncrypted(o.UserDefined); encrypted {
 		decrypted, err := o.metadataDecrypter(h)("object-checksum", data)
 		if err != nil {
@@ -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)
 		}
@@ -26,6 +26,7 @@ import (
 	"path/filepath"
 	"reflect"
 	"runtime"
+	"slices"
 	"sort"
 	"strconv"
 	"strings"
@@ -38,7 +39,6 @@ import (
 	"github.com/minio/minio/internal/mountinfo"
 	"github.com/minio/pkg/v3/env"
 	xnet "github.com/minio/pkg/v3/net"
-	"golang.org/x/exp/slices"
 )

 // EndpointType - enum for endpoint type.
@@ -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
-}
@@ -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]
@@ -277,23 +277,21 @@ 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
+func countPartNotSuccess(partErrs []int) (c int) {
+	for _, pe := range partErrs {
+		if pe != checkPartSuccess {
+			c++
+		}
+	}
+	return
 }

-// disksWithAllParts - This function needs to be called with
-// []StorageAPI returned by listOnlineDisks. Returns,
-//
-// - disks which have all parts specified in the latest xl.meta.
-//
-//   - slice of errors about the state of data files on disk - can have
-//     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,
+// checkObjectWithAllParts sets partsMetadata and onlineDisks when xl.meta is inexistant/corrupted or outdated
+// it also checks if the status of each part (corrupted, missing, ok) in each drive
+func checkObjectWithAllParts(ctx context.Context, onlineDisks []StorageAPI, partsMetadata []FileInfo,
+	errs []error, latestMeta FileInfo, filterByETag bool, bucket, object string,
 	scanMode madmin.HealScanMode,
-) (availableDisks []StorageAPI, dataErrsByDisk map[int][]int, dataErrsByPart map[int][]int) {
-	availableDisks = make([]StorageAPI, len(onlineDisks))
-
+) (dataErrsByDisk map[int][]int, dataErrsByPart map[int][]int) {
 	dataErrsByDisk = make(map[int][]int, len(onlineDisks))
 	for i := range onlineDisks {
 		dataErrsByDisk[i] = make([]int, len(latestMeta.Parts))
@@ -334,20 +332,28 @@ func disksWithAllParts(ctx context.Context, onlineDisks []StorageAPI, partsMetad

 	metaErrs := make([]error, len(errs))

-	for i, onlineDisk := range onlineDisks {
+	for i := range onlineDisks {
 		if errs[i] != nil {
 			metaErrs[i] = errs[i]
 			continue
 		}
-		if onlineDisk == OfflineDisk {
+		if onlineDisks[i] == OfflineDisk {
 			metaErrs[i] = errDiskNotFound
 			continue
 		}

 		meta := partsMetadata[i]
-		if !meta.ModTime.Equal(latestMeta.ModTime) || meta.DataDir != latestMeta.DataDir {
+		corrupted := false
+		if filterByETag {
+			corrupted = meta.Metadata["etag"] != latestMeta.Metadata["etag"]
+		} else {
+			corrupted = !meta.ModTime.Equal(latestMeta.ModTime) || meta.DataDir != latestMeta.DataDir
+		}
+
+		if corrupted {
 			metaErrs[i] = errFileCorrupt
 			partsMetadata[i] = FileInfo{}
+			onlineDisks[i] = nil
 			continue
 		}

@@ -355,6 +361,7 @@ func disksWithAllParts(ctx context.Context, onlineDisks []StorageAPI, partsMetad
 			if !meta.IsValid() {
 				partsMetadata[i] = FileInfo{}
 				metaErrs[i] = errFileCorrupt
+				onlineDisks[i] = nil
 				continue
 			}

@@ -365,6 +372,7 @@ func disksWithAllParts(ctx context.Context, onlineDisks []StorageAPI, partsMetad
 					// might have the right erasure distribution.
 					partsMetadata[i] = FileInfo{}
 					metaErrs[i] = errFileCorrupt
+					onlineDisks[i] = nil
 					continue
 				}
 			}
@@ -385,8 +393,8 @@ func disksWithAllParts(ctx context.Context, onlineDisks []StorageAPI, partsMetad
 		if metaErrs[i] != nil {
 			continue
 		}
-		meta := partsMetadata[i]

+		meta := partsMetadata[i]
 		if meta.Deleted || meta.IsRemote() {
 			continue
 		}
@@ -408,7 +416,6 @@ func disksWithAllParts(ctx context.Context, onlineDisks []StorageAPI, partsMetad
 			verifyResp *CheckPartsResp
 		)

-		meta.DataDir = latestMeta.DataDir
 		switch scanMode {
 		case madmin.HealDeepScan:
 			// disk has a valid xl.meta but may not have all the
@@ -434,16 +441,5 @@ func disksWithAllParts(ctx context.Context, onlineDisks []StorageAPI, partsMetad
 			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 {
-			// upon errors just make that disk's fileinfo invalid
-			partsMetadata[i] = FileInfo{}
-		}
-	}
-
 	return
 }
@@ -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)
@@ -308,11 +313,11 @@ func TestListOnlineDisks(t *testing.T) {
 				t.Fatalf("Expected modTime to be equal to %v but was found to be %v",
 					test.expectedTime, modTime)
 			}
-			availableDisks, _, _ := disksWithAllParts(ctx, onlineDisks, partsMetadata,
-				test.errs, fi, bucket, object, madmin.HealDeepScan)
+			_, _ = checkObjectWithAllParts(ctx, onlineDisks, partsMetadata,
+				test.errs, fi, false, bucket, object, madmin.HealDeepScan)

 			if test._tamperBackend != noTamper {
-				if tamperedIndex != -1 && availableDisks[tamperedIndex] != nil {
+				if tamperedIndex != -1 && onlineDisks[tamperedIndex] != nil {
 					t.Fatalf("Drive (%v) with part.1 missing is not a drive with available data",
 						erasureDisks[tamperedIndex])
 				}
@@ -490,11 +495,11 @@ func TestListOnlineDisksSmallObjects(t *testing.T) {
 					test.expectedTime, modTime)
 			}

-			availableDisks, _, _ := disksWithAllParts(ctx, onlineDisks, partsMetadata,
-				test.errs, fi, bucket, object, madmin.HealDeepScan)
+			_, _ = checkObjectWithAllParts(ctx, onlineDisks, partsMetadata,
+				test.errs, fi, false, bucket, object, madmin.HealDeepScan)

 			if test._tamperBackend != noTamper {
-				if tamperedIndex != -1 && availableDisks[tamperedIndex] != nil {
+				if tamperedIndex != -1 && onlineDisks[tamperedIndex] != nil {
 					t.Fatalf("Drive (%v) with part.1 missing is not a drive with available data",
 						erasureDisks[tamperedIndex])
 				}
@@ -540,6 +545,7 @@ func TestDisksWithAllParts(t *testing.T) {

 	// Test 1: Test that all disks are returned without any failures with
 	// unmodified meta data
+	erasureDisks = s.getDisks()
 	partsMetadata, errs := readAllFileInfo(ctx, erasureDisks, "", bucket, object, "", false, true)
 	if err != nil {
 		t.Fatalf("Failed to read xl meta data %v", err)
@@ -552,14 +558,10 @@ func TestDisksWithAllParts(t *testing.T) {

 	erasureDisks, _, _ = listOnlineDisks(erasureDisks, partsMetadata, errs, readQuorum)

-	filteredDisks, _, dataErrsPerDisk := disksWithAllParts(ctx, erasureDisks, partsMetadata,
-		errs, fi, bucket, object, madmin.HealDeepScan)
+	dataErrsPerDisk, _ := checkObjectWithAllParts(ctx, erasureDisks, partsMetadata,
+		errs, fi, false, bucket, object, madmin.HealDeepScan)

-	if len(filteredDisks) != len(erasureDisks) {
-		t.Errorf("Unexpected number of drives: %d", len(filteredDisks))
-	}
-
-	for diskIndex, disk := range filteredDisks {
+	for diskIndex, disk := range erasureDisks {
 		if partNeedsHealing(dataErrsPerDisk[diskIndex]) {
 			t.Errorf("Unexpected error: %v", dataErrsPerDisk[diskIndex])
 		}
@@ -570,17 +572,15 @@ func TestDisksWithAllParts(t *testing.T) {
 	}

 	// Test 2: Not synchronized modtime
+	erasureDisks = s.getDisks()
 	partsMetadataBackup := partsMetadata[0]
 	partsMetadata[0].ModTime = partsMetadata[0].ModTime.Add(-1 * time.Hour)

 	errs = make([]error, len(erasureDisks))
-	filteredDisks, _, _ = disksWithAllParts(ctx, erasureDisks, partsMetadata,
-		errs, fi, bucket, object, madmin.HealDeepScan)
+	_, _ = checkObjectWithAllParts(ctx, erasureDisks, partsMetadata,
+		errs, fi, false, bucket, object, madmin.HealDeepScan)

-	if len(filteredDisks) != len(erasureDisks) {
-		t.Errorf("Unexpected number of drives: %d", len(filteredDisks))
-	}
-	for diskIndex, disk := range filteredDisks {
+	for diskIndex, disk := range erasureDisks {
 		if diskIndex == 0 && disk != nil {
 			t.Errorf("Drive not filtered as expected, drive: %d", diskIndex)
 		}
@@ -591,17 +591,15 @@ func TestDisksWithAllParts(t *testing.T) {
 	partsMetadata[0] = partsMetadataBackup // Revert before going to the next test

 	// Test 3: Not synchronized DataDir
+	erasureDisks = s.getDisks()
 	partsMetadataBackup = partsMetadata[1]
 	partsMetadata[1].DataDir = "foo-random"

 	errs = make([]error, len(erasureDisks))
-	filteredDisks, _, _ = disksWithAllParts(ctx, erasureDisks, partsMetadata,
-		errs, fi, bucket, object, madmin.HealDeepScan)
+	_, _ = checkObjectWithAllParts(ctx, erasureDisks, partsMetadata,
+		errs, fi, false, bucket, object, madmin.HealDeepScan)

-	if len(filteredDisks) != len(erasureDisks) {
-		t.Errorf("Unexpected number of drives: %d", len(filteredDisks))
-	}
-	for diskIndex, disk := range filteredDisks {
+	for diskIndex, disk := range erasureDisks {
 		if diskIndex == 1 && disk != nil {
 			t.Errorf("Drive not filtered as expected, drive: %d", diskIndex)
 		}
@@ -612,6 +610,7 @@ func TestDisksWithAllParts(t *testing.T) {
 	partsMetadata[1] = partsMetadataBackup // Revert before going to the next test

 	// Test 4: key = disk index, value = part name with hash mismatch
+	erasureDisks = s.getDisks()
 	diskFailures := make(map[int]string)
 	diskFailures[0] = "part.1"
 	diskFailures[3] = "part.1"
@@ -632,29 +631,18 @@ func TestDisksWithAllParts(t *testing.T) {
 	}

 	errs = make([]error, len(erasureDisks))
-	filteredDisks, dataErrsPerDisk, _ = disksWithAllParts(ctx, erasureDisks, partsMetadata,
-		errs, fi, bucket, object, madmin.HealDeepScan)
+	dataErrsPerDisk, _ = checkObjectWithAllParts(ctx, erasureDisks, partsMetadata,
+		errs, fi, false, bucket, object, madmin.HealDeepScan)

-	if len(filteredDisks) != len(erasureDisks) {
-		t.Errorf("Unexpected number of drives: %d", len(filteredDisks))
-	}
-
-	for diskIndex, disk := range filteredDisks {
+	for diskIndex := range erasureDisks {
 		if _, ok := diskFailures[diskIndex]; ok {
-			if disk != nil {
-				t.Errorf("Drive not filtered as expected, drive: %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 partNeedsHealing(dataErrsPerDisk[diskIndex]) {
 				t.Errorf("Disk not expected to be healed, driveIndex: %d", diskIndex)
 			}
-
 		}
 	}
 }
@@ -32,7 +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"
+	"github.com/puzpuzpuz/xsync/v3"
 )

 //go:generate stringer -type=healingMetric -trimprefix=healingMetric $GOFILE
@@ -45,7 +45,8 @@ const (
 	healingMetricCheckAbandonedParts
 )

-func (er erasureObjects) listAndHeal(ctx context.Context, bucket, prefix string, scanMode madmin.HealScanMode, healEntry func(string, metaCacheEntry, madmin.HealScanMode) error) error {
+// List a prefix or a single object versions and heal
+func (er erasureObjects) listAndHeal(ctx context.Context, bucket, prefix string, recursive bool, scanMode madmin.HealScanMode, healEntry func(string, metaCacheEntry, madmin.HealScanMode) error) error {
 	ctx, cancel := context.WithCancel(ctx)
 	defer cancel()

@@ -78,11 +79,14 @@ func (er erasureObjects) listAndHeal(ctx context.Context, bucket, prefix string,
 		bucket:         bucket,
 		path:           path,
 		filterPrefix:   filterPrefix,
-		recursive:      true,
+		recursive:      recursive,
 		forwardTo:      "",
 		minDisks:       1,
 		reportNotFound: false,
 		agreed: func(entry metaCacheEntry) {
+			if !recursive && prefix != entry.name {
+				return
+			}
 			if err := healEntry(bucket, entry, scanMode); err != nil {
 				cancel()
 			}
@@ -94,7 +98,9 @@ func (er erasureObjects) listAndHeal(ctx context.Context, bucket, prefix string,
 				// proceed to heal nonetheless.
 				entry, _ = entries.firstFound()
 			}
-
+			if !recursive && prefix != entry.name {
+				return
+			}
 			if err := healEntry(bucket, *entry, scanMode); err != nil {
 				cancel()
 				return
@@ -112,9 +118,8 @@ func (er erasureObjects) listAndHeal(ctx context.Context, bucket, prefix string,

 // listAllBuckets lists all buckets from all disks. It also
 // returns the occurrence of each buckets in all disks
-func listAllBuckets(ctx context.Context, storageDisks []StorageAPI, healBuckets map[string]VolInfo, readQuorum int) error {
+func listAllBuckets(ctx context.Context, storageDisks []StorageAPI, healBuckets *xsync.MapOf[string, VolInfo], readQuorum int) error {
 	g := errgroup.WithNErrs(len(storageDisks))
-	var mu sync.Mutex
 	for index := range storageDisks {
 		index := index
 		g.Go(func() error {
@@ -126,6 +131,7 @@ func listAllBuckets(ctx context.Context, storageDisks []StorageAPI, healBuckets
 			if err != nil {
 				return err
 			}
+
 			for _, volInfo := range volsInfo {
 				// StorageAPI can send volume names which are
 				// incompatible with buckets - these are
@@ -133,53 +139,75 @@ func listAllBuckets(ctx context.Context, storageDisks []StorageAPI, healBuckets
 				if isReservedOrInvalidBucket(volInfo.Name, false) {
 					continue
 				}
-				mu.Lock()
-				if _, ok := healBuckets[volInfo.Name]; !ok {
-					healBuckets[volInfo.Name] = volInfo
-				}
-				mu.Unlock()
+
+				healBuckets.Compute(volInfo.Name, func(oldValue VolInfo, loaded bool) (newValue VolInfo, del bool) {
+					if loaded {
+						newValue = oldValue
+						newValue.count = oldValue.count + 1
+						return newValue, false
+					}
+					return VolInfo{
+						Name:    volInfo.Name,
+						Created: volInfo.Created,
+						count:   1,
+					}, false
+				})
 			}
+
 			return nil
 		}, index)
 	}
-	return reduceReadQuorumErrs(ctx, g.Wait(), bucketMetadataOpIgnoredErrs, readQuorum)
+
+	if err := reduceReadQuorumErrs(ctx, g.Wait(), bucketMetadataOpIgnoredErrs, readQuorum); err != nil {
+		return err
+	}
+
+	healBuckets.Range(func(volName string, volInfo VolInfo) bool {
+		if volInfo.count < readQuorum {
+			healBuckets.Delete(volName)
+		}
+		return true
+	})
+
+	return nil
 }

-var errLegacyXLMeta = errors.New("legacy XL meta")
-
-var errOutdatedXLMeta = errors.New("outdated XL meta")
-
-var errPartMissingOrCorrupt = errors.New("part missing or corrupt")
+var (
+	errLegacyXLMeta   = errors.New("legacy XL meta")
+	errOutdatedXLMeta = errors.New("outdated XL meta")
+	errPartCorrupt    = errors.New("part corrupt")
+	errPartMissing    = errors.New("part missing")
+)

 // 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 error, partsErrs []int, meta FileInfo, latestMeta FileInfo) (bool, error) {
+func shouldHealObjectOnDisk(erErr error, partsErrs []int, meta FileInfo, latestMeta FileInfo) (bool, bool, error) {
 	if errors.Is(erErr, errFileNotFound) || errors.Is(erErr, errFileVersionNotFound) || errors.Is(erErr, errFileCorrupt) {
-		return true, erErr
+		return true, true, erErr
 	}
 	if erErr == nil {
 		if meta.XLV1 {
 			// Legacy means heal always
 			// always check first.
-			return true, errLegacyXLMeta
+			return true, true, errLegacyXLMeta
 		}
 		if !latestMeta.Equals(meta) {
-			return true, errOutdatedXLMeta
+			return true, true, errOutdatedXLMeta
 		}
 		if !meta.Deleted && !meta.IsRemote() {
 			// If xl.meta was read fine but there may be problem with the part.N files.
 			for _, partErr := range partsErrs {
-				if slices.Contains([]int{
-					checkPartFileNotFound,
-					checkPartFileCorrupt,
-				}, partErr) {
-					return true, errPartMissingOrCorrupt
+				if partErr == checkPartFileNotFound {
+					return true, false, errPartMissing
+				}
+				if partErr == checkPartFileCorrupt {
+					return true, false, errPartCorrupt
 				}
 			}
 		}
-		return false, nil
+		return false, false, nil
 	}
-	return false, erErr
+	return false, false, erErr
 }

 const (
@@ -226,26 +254,49 @@ 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 > 0 && b == a {
+		opts.Error = fmt.Sprintf("unable to heal %d corrupted blocks on drives", b)
+	}
+
+	b, a = result.GetMissingCounts()
+	if b > 0 && 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)
 }

+func objectErrToDriveState(reason error) string {
+	switch {
+	case reason == nil:
+		return madmin.DriveStateOk
+	case IsErr(reason, errDiskNotFound):
+		return madmin.DriveStateOffline
+	case IsErr(reason, errFileNotFound, errFileVersionNotFound, errVolumeNotFound, errPartMissing, errOutdatedXLMeta, errLegacyXLMeta):
+		return madmin.DriveStateMissing
+	case IsErr(reason, errFileCorrupt, errPartCorrupt):
+		return madmin.DriveStateCorrupt
+	default:
+		return fmt.Sprintf("%s (%s)", madmin.DriveStateUnknown, reason.Error())
+	}
+}
+
 // Heals an object by re-writing corrupt/missing erasure blocks.
 func (er *erasureObjects) healObject(ctx context.Context, bucket string, object string, versionID string, opts madmin.HealOpts) (result madmin.HealResultItem, err error) {
 	dryRun := opts.DryRun
@@ -320,26 +371,20 @@ func (er *erasureObjects) healObject(ctx context.Context, bucket string, object

 	// List of disks having latest version of the object xl.meta
 	// (by modtime).
-	onlineDisks, modTime, etag := listOnlineDisks(storageDisks, partsMetadata, errs, readQuorum)
+	onlineDisks, quorumModTime, quorumETag := listOnlineDisks(storageDisks, partsMetadata, errs, readQuorum)

 	// Latest FileInfo for reference. If a valid metadata is not
 	// present, it is as good as object not found.
-	latestMeta, err := pickValidFileInfo(ctx, partsMetadata, modTime, etag, readQuorum)
+	latestMeta, err := pickValidFileInfo(ctx, partsMetadata, quorumModTime, quorumETag, readQuorum)
 	if err != nil {
 		return result, err
 	}

-	// List of disks having all parts as per latest metadata.
-	// NOTE: do not pass in latestDisks to diskWithAllParts since
-	// the diskWithAllParts needs to reach the drive to ensure
-	// validity of the metadata content, we should make sure that
-	// we pass in disks as is for it to be verified. Once verified
-	// the disksWithAllParts() returns the actual disks that can be
-	// 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, dataErrsByDisk, dataErrsByPart := disksWithAllParts(ctx, onlineDisks, partsMetadata,
-		errs, latestMeta, bucket, object, scanMode)
+	// No modtime quorum
+	filterDisksByETag := quorumETag != ""
+
+	dataErrsByDisk, dataErrsByPart := checkObjectWithAllParts(ctx, onlineDisks, partsMetadata,
+		errs, latestMeta, filterDisksByETag, bucket, object, scanMode)

 	var erasure Erasure
 	if !latestMeta.Deleted && !latestMeta.IsRemote() {
@@ -360,26 +405,18 @@ func (er *erasureObjects) healObject(ctx context.Context, bucket string, object
 	// data state and a list of outdated disks on which data needs
 	// to be healed.
 	outDatedDisks := make([]StorageAPI, len(storageDisks))
-	disksToHealCount := 0
-	for i := range availableDisks {
-		yes, reason := shouldHealObjectOnDisk(errs[i], dataErrsByDisk[i], partsMetadata[i], latestMeta)
+	disksToHealCount, xlMetaToHealCount := 0, 0
+	for i := range onlineDisks {
+		yes, isMeta, reason := shouldHealObjectOnDisk(errs[i], dataErrsByDisk[i], partsMetadata[i], latestMeta)
 		if yes {
 			outDatedDisks[i] = storageDisks[i]
 			disksToHealCount++
+			if isMeta {
+				xlMetaToHealCount++
+			}
 		}

-		driveState := ""
-		switch {
-		case reason == nil:
-			driveState = madmin.DriveStateOk
-		case IsErr(reason, errDiskNotFound):
-			driveState = madmin.DriveStateOffline
-		case IsErr(reason, errFileNotFound, errFileVersionNotFound, errVolumeNotFound, errPartMissingOrCorrupt, errOutdatedXLMeta, errLegacyXLMeta):
-			driveState = madmin.DriveStateMissing
-		default:
-			// all remaining cases imply corrupt data/metadata
-			driveState = madmin.DriveStateCorrupt
-		}
+		driveState := objectErrToDriveState(reason)

 		result.Before.Drives = append(result.Before.Drives, madmin.HealDriveInfo{
 			UUID:     "",
@@ -393,16 +430,6 @@ func (er *erasureObjects) healObject(ctx context.Context, bucket string, object
 		})
 	}

-	if isAllNotFound(errs) {
-		// File is fully gone, fileInfo is empty.
-		err := errFileNotFound
-		if versionID != "" {
-			err = errFileVersionNotFound
-		}
-		return er.defaultHealResult(FileInfo{}, storageDisks, storageEndpoints, errs,
-			bucket, object, versionID), err
-	}
-
 	if disksToHealCount == 0 {
 		// Nothing to heal!
 		return result, nil
@@ -414,7 +441,24 @@ func (er *erasureObjects) healObject(ctx context.Context, bucket string, object
 		return result, nil
 	}

-	if !latestMeta.XLV1 && !latestMeta.Deleted && disksToHealCount > latestMeta.Erasure.ParityBlocks {
+	cannotHeal := !latestMeta.XLV1 && !latestMeta.Deleted && xlMetaToHealCount > latestMeta.Erasure.ParityBlocks
+	if cannotHeal && quorumETag != "" {
+		// This is an object that is supposed to be removed by the dangling code
+		// but we noticed that ETag is the same for all objects, let's give it a shot
+		cannotHeal = false
+	}
+
+	if !latestMeta.Deleted && !latestMeta.IsRemote() {
+		// check if there is a part that lost its quorum
+		for _, partErrs := range dataErrsByPart {
+			if countPartNotSuccess(partErrs) > latestMeta.Erasure.ParityBlocks {
+				cannotHeal = true
+				break
+			}
+		}
+	}
+
+	if cannotHeal {
 		// 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, dataErrsByPart, ObjectOptions{
@@ -452,16 +496,15 @@ func (er *erasureObjects) healObject(ctx context.Context, bucket string, object
 	tmpID := mustGetUUID()
 	migrateDataDir := mustGetUUID()

-	// Reorder so that we have data disks first and parity disks next.
-	if !latestMeta.Deleted && len(latestMeta.Erasure.Distribution) != len(availableDisks) {
-		err := fmt.Errorf("unexpected file distribution (%v) from available disks (%v), looks like backend disks have been manually modified refusing to heal %s/%s(%s)",
-			latestMeta.Erasure.Distribution, availableDisks, bucket, object, versionID)
-		healingLogOnceIf(ctx, err, "heal-object-available-disks")
+	if !latestMeta.Deleted && len(latestMeta.Erasure.Distribution) != len(onlineDisks) {
+		err := fmt.Errorf("unexpected file distribution (%v) from online disks (%v), looks like backend disks have been manually modified refusing to heal %s/%s(%s)",
+			latestMeta.Erasure.Distribution, onlineDisks, bucket, object, versionID)
+		healingLogOnceIf(ctx, err, "heal-object-online-disks")
 		return er.defaultHealResult(latestMeta, storageDisks, storageEndpoints, errs,
 			bucket, object, versionID), err
 	}

-	latestDisks := shuffleDisks(availableDisks, latestMeta.Erasure.Distribution)
+	latestDisks := shuffleDisks(onlineDisks, latestMeta.Erasure.Distribution)

 	if !latestMeta.Deleted && len(latestMeta.Erasure.Distribution) != len(outDatedDisks) {
 		err := fmt.Errorf("unexpected file distribution (%v) from outdated disks (%v), looks like backend disks have been manually modified refusing to heal %s/%s(%s)",
@@ -532,6 +575,10 @@ func (er *erasureObjects) healObject(ctx context.Context, bucket string, object
 				if disk == OfflineDisk {
 					continue
 				}
+				thisPartErrs := shuffleCheckParts(dataErrsByPart[partIndex], latestMeta.Erasure.Distribution)
+				if thisPartErrs[i] != checkPartSuccess {
+					continue
+				}
 				checksumInfo := copyPartsMetadata[i].Erasure.GetChecksumInfo(partNumber)
 				partPath := pathJoin(object, srcDataDir, fmt.Sprintf("part.%d", partNumber))
 				readers[i] = newBitrotReader(disk, copyPartsMetadata[i].Data, bucket, partPath, tillOffset, checksumAlgo,
@@ -799,13 +846,7 @@ func (er *erasureObjects) defaultHealResult(lfi FileInfo, storageDisks []Storage
 			})
 			continue
 		}
-		driveState := madmin.DriveStateCorrupt
-		switch errs[index] {
-		case errFileNotFound, errVolumeNotFound:
-			driveState = madmin.DriveStateMissing
-		case nil:
-			driveState = madmin.DriveStateOk
-		}
+		driveState := objectErrToDriveState(errs[index])
 		result.Before.Drives = append(result.Before.Drives, madmin.HealDriveInfo{
 			UUID:     "",
 			Endpoint: storageEndpoints[index].String(),
@@ -733,7 +733,7 @@ func TestHealingDanglingObject(t *testing.T) {
 		t.Fatalf("Expected versions 1, got %d", fileInfoPreHeal.NumVersions)
 	}

-	if err = objLayer.HealObjects(ctx, bucket, "", madmin.HealOpts{Remove: true},
+	if err = objLayer.HealObjects(ctx, bucket, "", madmin.HealOpts{Recursive: true, Remove: true},
 		func(bucket, object, vid string, scanMode madmin.HealScanMode) error {
 			_, err := objLayer.HealObject(ctx, bucket, object, vid, madmin.HealOpts{ScanMode: scanMode, Remove: true})
 			return err
@@ -780,7 +780,7 @@ func TestHealingDanglingObject(t *testing.T) {
 		t.Fatalf("Expected versions 1, got %d", fileInfoPreHeal.NumVersions)
 	}

-	if err = objLayer.HealObjects(ctx, bucket, "", madmin.HealOpts{Remove: true},
+	if err = objLayer.HealObjects(ctx, bucket, "", madmin.HealOpts{Recursive: true, Remove: true},
 		func(bucket, object, vid string, scanMode madmin.HealScanMode) error {
 			_, err := objLayer.HealObject(ctx, bucket, object, vid, madmin.HealOpts{ScanMode: scanMode, Remove: true})
 			return err
@@ -829,7 +829,7 @@ func TestHealingDanglingObject(t *testing.T) {
 		t.Fatalf("Expected versions 3, got %d", fileInfoPreHeal.NumVersions)
 	}

-	if err = objLayer.HealObjects(ctx, bucket, "", madmin.HealOpts{Remove: true},
+	if err = objLayer.HealObjects(ctx, bucket, "", madmin.HealOpts{Recursive: true, Remove: true},
 		func(bucket, object, vid string, scanMode madmin.HealScanMode) error {
 			_, err := objLayer.HealObject(ctx, bucket, object, vid, madmin.HealOpts{ScanMode: scanMode, Remove: true})
 			return err
@@ -52,15 +52,15 @@ func reduceCommonVersions(diskVersions [][]byte, writeQuorum int) (versions []by
 	}

 	var commonVersions uint64
-	max := 0
+	maxCnt := 0
 	for versions, count := range diskVersionsCount {
-		if max < count {
-			max = count
+		if maxCnt < count {
+			maxCnt = count
 			commonVersions = versions
 		}
 	}

-	if max >= writeQuorum {
+	if maxCnt >= writeQuorum {
 		for _, versions := range diskVersions {
 			if binary.BigEndian.Uint64(versions) == commonVersions {
 				return versions
@@ -80,15 +80,15 @@ func reduceCommonDataDir(dataDirs []string, writeQuorum int) (dataDir string) {
 		dataDirsCount[ddir]++
 	}

-	max := 0
+	maxCnt := 0
 	for ddir, count := range dataDirsCount {
-		if max < count {
-			max = count
+		if maxCnt < count {
+			maxCnt = count
 			dataDir = ddir
 		}
 	}

-	if max >= writeQuorum {
+	if maxCnt >= writeQuorum {
 		return dataDir
 	}

@@ -115,20 +115,20 @@ func reduceErrs(errs []error, ignoredErrs []error) (maxCount int, maxErr error)
 		errorCounts[err]++
 	}

-	max := 0
+	maxCnt := 0
 	for err, count := range errorCounts {
 		switch {
-		case max < count:
-			max = count
+		case maxCnt < count:
+			maxCnt = count
 			maxErr = err

 		// Prefer `nil` over other error values with the same
 		// number of occurrences.
-		case max == count && err == nil:
+		case maxCnt == count && err == nil:
 			maxErr = err
 		}
 	}
-	return max, maxErr
+	return maxCnt, maxErr
 }

 // reduceQuorumErrs behaves like reduceErrs by only for returning
@@ -295,34 +295,34 @@ func shuffleDisksAndPartsMetadata(disks []StorageAPI, partsMetadata []FileInfo,
 	return shuffledDisks, shuffledPartsMetadata
 }

-// Return shuffled partsMetadata depending on distribution.
-func shufflePartsMetadata(partsMetadata []FileInfo, distribution []int) (shuffledPartsMetadata []FileInfo) {
+func shuffleWithDist[T any](input []T, distribution []int) []T {
 	if distribution == nil {
-		return partsMetadata
+		return input
 	}
-	shuffledPartsMetadata = make([]FileInfo, len(partsMetadata))
-	// Shuffle slice xl metadata for expected distribution.
-	for index := range partsMetadata {
+	shuffled := make([]T, len(input))
+	for index := range input {
 		blockIndex := distribution[index]
-		shuffledPartsMetadata[blockIndex-1] = partsMetadata[index]
+		shuffled[blockIndex-1] = input[index]
 	}
-	return shuffledPartsMetadata
+	return shuffled
+}
+
+// Return shuffled partsMetadata depending on distribution.
+func shufflePartsMetadata(partsMetadata []FileInfo, distribution []int) []FileInfo {
+	return shuffleWithDist[FileInfo](partsMetadata, distribution)
+}
+
+// shuffleCheckParts - shuffle CheckParts slice depending on the
+// erasure distribution.
+func shuffleCheckParts(parts []int, distribution []int) []int {
+	return shuffleWithDist[int](parts, distribution)
 }

 // shuffleDisks - shuffle input disks slice depending on the
 // erasure distribution. Return shuffled slice of disks with
 // their expected distribution.
-func shuffleDisks(disks []StorageAPI, distribution []int) (shuffledDisks []StorageAPI) {
-	if distribution == nil {
-		return disks
-	}
-	shuffledDisks = make([]StorageAPI, len(disks))
-	// Shuffle disks for expected distribution.
-	for index := range disks {
-		blockIndex := distribution[index]
-		shuffledDisks[blockIndex-1] = disks[index]
-	}
-	return shuffledDisks
+func shuffleDisks(disks []StorageAPI, distribution []int) []StorageAPI {
+	return shuffleWithDist[StorageAPI](disks, distribution)
 }

 // evalDisks - returns a new slice of disks where nil is set if
@@ -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
@@ -214,6 +216,7 @@ func (fi FileInfo) ReplicationInfoEquals(ofi FileInfo) bool {
 }

 // objectPartIndex - returns the index of matching object part number.
+// Returns -1 if the part cannot be found.
 func objectPartIndex(parts []ObjectPartInfo, partNumber int) int {
 	for i, part := range parts {
 		if partNumber == part.Number {
@@ -223,6 +226,17 @@ func objectPartIndex(parts []ObjectPartInfo, partNumber int) int {
 	return -1
 }

+// objectPartIndexNums returns the index of the specified part number.
+// Returns -1 if the part cannot be found.
+func objectPartIndexNums(parts []int, partNumber int) int {
+	for i, part := range parts {
+		if part != 0 && partNumber == part {
+			return i
+		}
+	}
+	return -1
+}
+
 // AddObjectPart - add a new object part in order.
 func (fi *FileInfo) AddObjectPart(partNumber int, partETag string, partSize, actualSize int64, modTime time.Time, idx []byte, checksums map[string]string) {
 	partInfo := ObjectPartInfo{
@@ -287,13 +301,14 @@ 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 +403,7 @@ 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 +427,37 @@ 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 +498,7 @@ 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 +509,15 @@ 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.
+			// However, we never go below EcM, in case of a EcM=EcN setup.
+			parities[index] = max(totalShards-(totalShards/2+1), metadata.Erasure.ParityBlocks)
 		} else {
 			parities[index] = metadata.Erasure.ParityBlocks
 		}
@@ -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)
+			}
+		})
+	}
+}
@@ -32,7 +32,6 @@ import (
 	"sync"
 	"time"

-	"github.com/dustin/go-humanize"
 	"github.com/klauspost/readahead"
 	"github.com/minio/madmin-go/v3"
 	"github.com/minio/minio-go/v7/pkg/tags"
@@ -49,6 +48,7 @@ import (
 	"github.com/minio/minio/internal/logger"
 	"github.com/minio/pkg/v3/mimedb"
 	"github.com/minio/pkg/v3/sync/errgroup"
+	"github.com/minio/sio"
 )

 // list all errors which can be ignored in object operations.
@@ -70,7 +70,7 @@ func countOnlineDisks(onlineDisks []StorageAPI) (online int) {
 // update metadata.
 func (er erasureObjects) CopyObject(ctx context.Context, srcBucket, srcObject, dstBucket, dstObject string, srcInfo ObjectInfo, srcOpts, dstOpts ObjectOptions) (oi ObjectInfo, err error) {
 	if !dstOpts.NoAuditLog {
-		auditObjectErasureSet(ctx, dstObject, &er)
+		auditObjectErasureSet(ctx, "CopyObject", dstObject, &er)
 	}

 	// This call shouldn't be used for anything other than metadata updates or adding self referential versions.
@@ -97,12 +97,12 @@ 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)
 	if err != nil {
-		if errors.Is(err, errErasureReadQuorum) && !strings.HasPrefix(srcBucket, minioMetaBucket) {
+		if shouldCheckForDangling(err, errs, srcBucket) {
 			_, derr := er.deleteIfDangling(context.Background(), srcBucket, srcObject, metaArr, errs, nil, srcOpts)
 			if derr == nil {
 				if srcOpts.VersionID != "" {
@@ -199,7 +199,7 @@ func (er erasureObjects) CopyObject(ctx context.Context, srcBucket, srcObject, d
 // Read(Closer). When err != nil, the returned reader is always nil.
 func (er erasureObjects) GetObjectNInfo(ctx context.Context, bucket, object string, rs *HTTPRangeSpec, h http.Header, opts ObjectOptions) (gr *GetObjectReader, err error) {
 	if !opts.NoAuditLog {
-		auditObjectErasureSet(ctx, object, &er)
+		auditObjectErasureSet(ctx, "GetObject", object, &er)
 	}

 	var unlockOnDefer bool
@@ -283,7 +283,7 @@ func (er erasureObjects) GetObjectNInfo(ctx context.Context, bucket, object stri
 	}

 	if unlockOnDefer {
-		unlockOnDefer = fi.InlineData()
+		unlockOnDefer = fi.InlineData() || len(fi.Data) > 0
 	}

 	pr, pw := xioutil.WaitPipe()
@@ -395,24 +395,16 @@ func (er erasureObjects) getObjectWithFileInfo(ctx context.Context, bucket, obje
 			// that we have some parts or data blocks missing or corrupted
 			// - attempt a heal to successfully heal them for future calls.
 			if written == partLength {
-				var scan madmin.HealScanMode
-				switch {
-				case errors.Is(err, errFileNotFound):
-					scan = madmin.HealNormalScan
-				case errors.Is(err, errFileCorrupt):
-					scan = madmin.HealDeepScan
-				}
-				switch scan {
-				case madmin.HealNormalScan, madmin.HealDeepScan:
+				if errors.Is(err, errFileNotFound) || errors.Is(err, errFileCorrupt) {
 					healOnce.Do(func() {
-						globalMRFState.addPartialOp(partialOperation{
-							bucket:    bucket,
-							object:    object,
-							versionID: fi.VersionID,
-							queued:    time.Now(),
-							setIndex:  er.setIndex,
-							poolIndex: er.poolIndex,
-							scanMode:  scan,
+						globalMRFState.addPartialOp(PartialOperation{
+							Bucket:     bucket,
+							Object:     object,
+							VersionID:  fi.VersionID,
+							Queued:     time.Now(),
+							SetIndex:   er.setIndex,
+							PoolIndex:  er.poolIndex,
+							BitrotScan: errors.Is(err, errFileCorrupt),
 						})
 					})
 					// Healing is triggered and we have written
@@ -439,7 +431,7 @@ func (er erasureObjects) getObjectWithFileInfo(ctx context.Context, bucket, obje
 // GetObjectInfo - reads object metadata and replies back ObjectInfo.
 func (er erasureObjects) GetObjectInfo(ctx context.Context, bucket, object string, opts ObjectOptions) (info ObjectInfo, err error) {
 	if !opts.NoAuditLog {
-		auditObjectErasureSet(ctx, object, &er)
+		auditObjectErasureSet(ctx, "GetObjectInfo", object, &er)
 	}

 	if !opts.NoLock {
@@ -456,7 +448,7 @@ func (er erasureObjects) GetObjectInfo(ctx context.Context, bucket, object strin
 	return er.getObjectInfo(ctx, bucket, object, opts)
 }

-func auditDanglingObjectDeletion(ctx context.Context, bucket, object, versionID string, tags map[string]interface{}) {
+func auditDanglingObjectDeletion(ctx context.Context, bucket, object, versionID string, tags map[string]string) {
 	if len(logger.AuditTargets()) == 0 {
 		return
 	}
@@ -472,13 +464,16 @@ func auditDanglingObjectDeletion(ctx context.Context, bucket, object, versionID
 	auditLogInternal(ctx, opts)
 }

-func joinErrs(errs []error) []string {
-	s := make([]string, len(errs))
+func joinErrs(errs []error) string {
+	var s string
 	for i := range s {
+		if s != "" {
+			s += ","
+		}
 		if errs[i] == nil {
-			s[i] = "<nil>"
+			s += "<nil>"
 		} else {
-			s[i] = errs[i].Error()
+			s += errs[i].Error()
 		}
 	}
 	return s
@@ -491,20 +486,18 @@ func (er erasureObjects) deleteIfDangling(ctx context.Context, bucket, object st
 		// can be deleted safely, in such a scenario return ReadQuorum error.
 		return FileInfo{}, errErasureReadQuorum
 	}
-	tags := make(map[string]interface{}, 4)
-	tags["set"] = er.setIndex
-	tags["pool"] = er.poolIndex
+	tags := make(map[string]string, 16)
+	tags["set"] = strconv.Itoa(er.setIndex)
+	tags["pool"] = strconv.Itoa(er.poolIndex)
 	tags["merrs"] = joinErrs(errs)
-	tags["derrs"] = dataErrsByPart
+	tags["derrs"] = fmt.Sprintf("%v", dataErrsByPart)
 	if m.IsValid() {
-		tags["size"] = m.Size
-		tags["mtime"] = m.ModTime.Format(http.TimeFormat)
-		tags["data"] = m.Erasure.DataBlocks
-		tags["parity"] = m.Erasure.ParityBlocks
+		tags["sz"] = strconv.FormatInt(m.Size, 10)
+		tags["mt"] = m.ModTime.Format(iso8601Format)
+		tags["d:p"] = fmt.Sprintf("%d:%d", m.Erasure.DataBlocks, m.Erasure.ParityBlocks)
 	} else {
-		tags["invalid-meta"] = true
-		tags["data"] = er.setDriveCount - er.defaultParityCount
-		tags["parity"] = er.defaultParityCount
+		tags["invalid"] = "1"
+		tags["d:p"] = fmt.Sprintf("%d:%d", er.setDriveCount-er.defaultParityCount, er.defaultParityCount)
 	}

 	// count the number of offline disks
@@ -527,7 +520,7 @@ func (er erasureObjects) deleteIfDangling(ctx context.Context, bucket, object st
 		}
 	}
 	if offline > 0 {
-		tags["offline"] = offline
+		tags["offline"] = strconv.Itoa(offline)
 	}

 	_, file, line, cok := runtime.Caller(1)
@@ -556,30 +549,23 @@ 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) {
+func fileInfoFromRaw(ri RawFileInfo, bucket, object string, readData, inclFreeVers bool) (FileInfo, error) {
 	return getFileInfo(ri.Buf, bucket, object, "", fileInfoOpts{
 		Data:             readData,
 		InclFreeVersions: inclFreeVers,
-		AllParts:         allParts,
 	})
 }

@@ -604,7 +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))
@@ -641,7 +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 {
@@ -670,7 +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
 		}
@@ -704,20 +690,16 @@ func shouldCheckForDangling(err error, errs []error, bucket string) bool {
 	// Check if we have a read quorum issue
 	case errors.Is(err, errErasureReadQuorum):
 		return true
-	// Check if the object is inexistent in most disks but not all of them
-	case errors.Is(err, errFileNotFound) || errors.Is(err, errFileVersionNotFound):
-		for i := range errs {
-			if errs[i] == nil {
-				return true
-			}
-		}
+	// Check if the object is non-existent on most disks but not all of them
+	case (errors.Is(err, errFileNotFound) || errors.Is(err, errFileVersionNotFound)) && (countErrs(errs, nil) > 0):
+		return true
 	}
 	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) {
@@ -774,7 +756,7 @@ 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)
 					}
 				}

@@ -820,13 +802,13 @@ 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,
 			})
 		}

@@ -908,14 +890,16 @@ 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
 		}
 	}
@@ -933,6 +917,16 @@ func (er erasureObjects) getObjectFileInfo(ctx context.Context, bucket, object s
 				}
 			}
 		}
+		// when we have insufficient read quorum and inconsistent metadata return
+		// file not found, since we can't possibly have a way to recover this object
+		// anyway.
+		if v, ok := err.(InsufficientReadQuorum); ok && v.Type == RQInconsistentMeta {
+			if opts.VersionID != "" {
+				err = errFileVersionNotFound
+			} else {
+				err = errFileNotFound
+			}
+		}
 		return fi, nil, nil, toObjectErr(err, bucket, object)
 	}

@@ -1250,7 +1244,7 @@ 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
@@ -1341,16 +1335,12 @@ func (er erasureObjects) putObject(ctx context.Context, bucket string, object st
 	}

 	fi.DataDir = mustGetUUID()
-	fi.Checksum = opts.WantChecksum.AppendTo(nil, nil)
-	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 {
+	if ckSum := userDefined[ReplicationSsecChecksumHeader]; ckSum != "" {
+		if v, err := base64.StdEncoding.DecodeString(ckSum); err == nil {
 			fi.Checksum = v
 		}
+		delete(userDefined, ReplicationSsecChecksumHeader)
 	}
-	delete(userDefined, ReplicationSsecChecksumHeader)
 	uniqueID := mustGetUUID()
 	tempObj := uniqueID

@@ -1390,30 +1380,13 @@ func (er erasureObjects) putObject(ctx context.Context, bucket string, object st

 	defer er.deleteAll(context.Background(), minioMetaTmpBucket, tempObj)

-	shardFileSize := erasure.ShardFileSize(data.Size())
-	inlineBlock := globalStorageClass.InlineBlock()
-	if inlineBlock <= 0 {
-		inlineBlock = 128 * humanize.KiByte
+	var inlineBuffers []*bytes.Buffer
+	if globalStorageClass.ShouldInline(erasure.ShardFileSize(data.ActualSize()), opts.Versioned) {
+		inlineBuffers = make([]*bytes.Buffer, len(onlineDisks))
 	}

+	shardFileSize := erasure.ShardFileSize(data.Size())
 	writers := make([]io.Writer, len(onlineDisks))
-	var inlineBuffers []*bytes.Buffer
-	if shardFileSize >= 0 {
-		if !opts.Versioned && shardFileSize < inlineBlock {
-			inlineBuffers = make([]*bytes.Buffer, len(onlineDisks))
-		} else if shardFileSize < inlineBlock/8 {
-			inlineBuffers = make([]*bytes.Buffer, len(onlineDisks))
-		}
-	} else {
-		// If compressed, use actual size to determine.
-		if sz := erasure.ShardFileSize(data.ActualSize()); sz > 0 {
-			if !opts.Versioned && sz < inlineBlock {
-				inlineBuffers = make([]*bytes.Buffer, len(onlineDisks))
-			} else if sz < inlineBlock/8 {
-				inlineBuffers = make([]*bytes.Buffer, len(onlineDisks))
-			}
-		}
-	}
 	for i, disk := range onlineDisks {
 		if disk == nil {
 			continue
@@ -1465,21 +1438,37 @@ func (er erasureObjects) putObject(ctx context.Context, bucket string, object st
 	if opts.IndexCB != nil {
 		compIndex = opts.IndexCB()
 	}
-	if !opts.NoLock {
-		lk := er.NewNSLock(bucket, object)
-		lkctx, err := lk.GetLock(ctx, globalOperationTimeout)
-		if err != nil {
-			return ObjectInfo{}, err
-		}
-		ctx = lkctx.Context()
-		defer lk.Unlock(lkctx)
-	}

 	modTime := opts.MTime
 	if opts.MTime.IsZero() {
 		modTime = UTCNow()
 	}

+	kind, encrypted := crypto.IsEncrypted(userDefined)
+	actualSize := data.ActualSize()
+	if actualSize < 0 {
+		compressed := fi.IsCompressed()
+		switch {
+		case compressed:
+			// ... nothing changes for compressed stream.
+			// if actualSize is -1 we have no known way to
+			// determine what is the actualSize.
+		case encrypted:
+			decSize, err := sio.DecryptedSize(uint64(n))
+			if err == nil {
+				actualSize = int64(decSize)
+			}
+		default:
+			actualSize = n
+		}
+	}
+	if fi.Checksum == nil {
+		// Trailing headers checksums should now be filled.
+		fi.Checksum = opts.WantChecksum.AppendTo(nil, nil)
+		if opts.EncryptFn != nil {
+			fi.Checksum = opts.EncryptFn("object-checksum", fi.Checksum)
+		}
+	}
 	for i, w := range writers {
 		if w == nil {
 			onlineDisks[i] = nil
@@ -1491,12 +1480,12 @@ func (er erasureObjects) putObject(ctx context.Context, bucket string, object st
 			partsMetadata[i].Data = nil
 		}
 		// No need to add checksum to part. We already have it on the object.
-		partsMetadata[i].AddObjectPart(1, "", n, data.ActualSize(), modTime, compIndex, nil)
+		partsMetadata[i].AddObjectPart(1, "", n, actualSize, modTime, compIndex, nil)
 		partsMetadata[i].Versioned = opts.Versioned || opts.VersionSuspended
+		partsMetadata[i].Checksum = fi.Checksum
 	}

 	userDefined["etag"] = r.MD5CurrentHexString()
-	kind, _ := crypto.IsEncrypted(userDefined)
 	if opts.PreserveETag != "" {
 		if !opts.ReplicationRequest {
 			userDefined["etag"] = opts.PreserveETag
@@ -1532,6 +1521,16 @@ func (er erasureObjects) putObject(ctx context.Context, bucket string, object st
 		}
 	}

+	if !opts.NoLock {
+		lk := er.NewNSLock(bucket, object)
+		lkctx, err := lk.GetLock(ctx, globalOperationTimeout)
+		if err != nil {
+			return ObjectInfo{}, err
+		}
+		ctx = lkctx.Context()
+		defer lk.Unlock(lkctx)
+	}
+
 	// Rename the successfully written temporary object to final location.
 	onlineDisks, versions, oldDataDir, err := renameData(ctx, onlineDisks, minioMetaTmpBucket, tempObj, partsMetadata, bucket, object, writeQuorum)
 	if err != nil {
@@ -1576,13 +1575,13 @@ 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,
 			})
 		}
 	}
@@ -1625,7 +1624,7 @@ 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)
 		}
 	}

@@ -1845,7 +1844,7 @@ 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
@@ -2111,11 +2110,11 @@ 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(),
 	})
 }

@@ -2140,12 +2139,12 @@ 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)
 	if err != nil {
-		if errors.Is(err, errErasureReadQuorum) && !strings.HasPrefix(bucket, minioMetaBucket) {
+		if shouldCheckForDangling(err, errs, bucket) {
 			_, derr := er.deleteIfDangling(context.Background(), bucket, object, metaArr, errs, nil, opts)
 			if derr == nil {
 				if opts.VersionID != "" {
@@ -2199,14 +2198,16 @@ func (er erasureObjects) PutObjectMetadata(ctx context.Context, bucket, object s

 // PutObjectTags - replace or add tags to an existing object
 func (er erasureObjects) PutObjectTags(ctx context.Context, bucket, object string, tags string, opts ObjectOptions) (ObjectInfo, error) {
-	// Lock the object before updating tags.
-	lk := er.NewNSLock(bucket, object)
-	lkctx, err := lk.GetLock(ctx, globalOperationTimeout)
-	if err != nil {
-		return ObjectInfo{}, err
+	if !opts.NoLock {
+		// Lock the object before updating tags.
+		lk := er.NewNSLock(bucket, object)
+		lkctx, err := lk.GetLock(ctx, globalOperationTimeout)
+		if err != nil {
+			return ObjectInfo{}, err
+		}
+		ctx = lkctx.Context()
+		defer lk.Unlock(lkctx)
 	}
-	ctx = lkctx.Context()
-	defer lk.Unlock(lkctx)

 	disks := er.getDisks()

@@ -2217,12 +2218,12 @@ 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)
 	if err != nil {
-		if errors.Is(err, errErasureReadQuorum) && !strings.HasPrefix(bucket, minioMetaBucket) {
+		if shouldCheckForDangling(err, errs, bucket) {
 			_, derr := er.deleteIfDangling(context.Background(), bucket, object, metaArr, errs, nil, opts)
 			if derr == nil {
 				if opts.VersionID != "" {
@@ -2317,14 +2318,16 @@ func (er erasureObjects) TransitionObject(ctx context.Context, bucket, object st
 		return err
 	}

-	// Acquire write lock before starting to transition the object.
-	lk := er.NewNSLock(bucket, object)
-	lkctx, err := lk.GetLock(ctx, globalDeleteOperationTimeout)
-	if err != nil {
-		return err
+	if !opts.NoLock {
+		// Acquire write lock before starting to transition the object.
+		lk := er.NewNSLock(bucket, object)
+		lkctx, err := lk.GetLock(ctx, globalDeleteOperationTimeout)
+		if err != nil {
+			return err
+		}
+		ctx = lkctx.Context()
+		defer lk.Unlock(lkctx)
 	}
-	ctx = lkctx.Context()
-	defer lk.Unlock(lkctx)

 	fi, metaArr, onlineDisks, err := er.getObjectFileInfo(ctx, bucket, object, opts, true)
 	if err != nil {
@@ -2362,6 +2365,7 @@ func (er erasureObjects) TransitionObject(ctx context.Context, bucket, object st

 	destObj, err := genTransitionObjName(bucket)
 	if err != nil {
+		traceFn(ILMTransition, nil, err)
 		return err
 	}

@@ -2372,9 +2376,14 @@ func (er erasureObjects) TransitionObject(ctx context.Context, bucket, object st
 	}()

 	var rv remoteVersionID
-	rv, err = tgtClient.Put(ctx, destObj, pr, fi.Size)
+	rv, err = tgtClient.PutWithMeta(ctx, destObj, pr, fi.Size, map[string]string{
+		"name": object, // preserve the original name of the object on the remote tier object metadata.
+		// this is just for future reverse lookup() purposes (applies only for new objects)
+		// does not apply retro-actively on already transitioned objects.
+	})
 	pr.CloseWithError(err)
 	if err != nil {
+		traceFn(ILMTransition, nil, err)
 		return err
 	}
 	fi.TransitionStatus = lifecycle.TransitionComplete
@@ -1,4 +1,4 @@
-// Copyright (c) 2015-2023 MinIO, Inc.
+// Copyright (c) 2015-2024 MinIO, Inc.
 //
 // This file is part of MinIO Object Storage stack
 //
@@ -490,14 +490,11 @@ const (
 // in 'pool.bin', this is eventually used for decommissioning the pool.
 func (z *erasureServerPools) Init(ctx context.Context) error {
 	// Load rebalance metadata if present
-	err := z.loadRebalanceMeta(ctx)
-	if err != nil {
-		return fmt.Errorf("failed to load rebalance data: %w", err)
+	if err := z.loadRebalanceMeta(ctx); err == nil {
+		// Start rebalance routine if we can reload rebalance metadata.
+		z.StartRebalance()
 	}

-	// Start rebalance routine
-	z.StartRebalance()
-
 	meta := poolMeta{}
 	if err := meta.load(ctx, z.serverPools[0], z.serverPools); err != nil {
 		return err
@@ -608,7 +605,7 @@ 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 +620,11 @@ 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 +659,7 @@ 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 +681,7 @@ 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,
@@ -844,7 +845,7 @@ func (z *erasureServerPools) decommissionPool(ctx context.Context, idx int, pool
 				// to decommission, just skip it, this also includes
 				// any other versions that have already expired.
 				remainingVersions := len(fivs.Versions) - expired
-				if version.Deleted && remainingVersions == 1 {
+				if version.Deleted && remainingVersions == 1 && rcfg == nil {
 					decommissioned++
 					stopFn(version.Size, errors.New("DELETE marked object with no other non-current versions will be skipped"))
 					continue
@@ -855,6 +856,7 @@ func (z *erasureServerPools) decommissionPool(ctx context.Context, idx int, pool
 					versionID = nullVersionID
 				}

+				var failure, ignore bool
 				if version.Deleted {
 					_, err := z.DeleteObject(ctx,
 						bi.Name,
@@ -867,14 +869,19 @@ func (z *erasureServerPools) decommissionPool(ctx context.Context, idx int, pool
 							VersionID:          versionID,
 							MTime:              version.ModTime,
 							DeleteReplication:  version.ReplicationState,
+							SrcPoolIdx:         idx,
+							DataMovement:       true,
 							DeleteMarker:       true, // make sure we create a delete marker
 							SkipDecommissioned: true, // make sure we skip the decommissioned pool
 							NoAuditLog:         true,
 						})
-					var failure bool
 					if err != nil {
-						if isErrObjectNotFound(err) || isErrVersionNotFound(err) {
-							err = nil
+						// This can happen when rebalance stop races with ongoing rebalance workers.
+						// These rebalance failures can be ignored.
+						if isErrObjectNotFound(err) || isErrVersionNotFound(err) || isDataMovementOverWriteErr(err) {
+							ignore = true
+							stopFn(0, nil)
+							continue
 						}
 					}
 					stopFn(version.Size, err)
@@ -893,22 +900,26 @@ func (z *erasureServerPools) decommissionPool(ctx context.Context, idx int, pool
 					continue
 				}

-				var failure, ignore bool
 				// gr.Close() is ensured by decommissionObject().
 				for try := 0; try < 3; try++ {
 					if version.IsRemote() {
 						if err := z.DecomTieredObject(ctx, bi.Name, version.Name, version, ObjectOptions{
-							VersionID:   versionID,
-							MTime:       version.ModTime,
-							UserDefined: version.Metadata,
+							VersionID:    versionID,
+							MTime:        version.ModTime,
+							UserDefined:  version.Metadata,
+							SrcPoolIdx:   idx,
+							DataMovement: true,
 						}); err != nil {
-							stopFn(version.Size, err)
-							failure = true
-							decomLogIf(ctx, err)
-							continue
+							if isErrObjectNotFound(err) || isErrVersionNotFound(err) || isDataMovementOverWriteErr(err) {
+								ignore = true
+								stopFn(0, nil)
+							}
+						}
+						if !ignore {
+							stopFn(version.Size, err)
+							failure = err != nil
+							decomLogIf(ctx, err)
 						}
-						stopFn(version.Size, nil)
-						failure = false
 						break
 					}
 					gr, err := set.GetObjectNInfo(ctx,
@@ -925,7 +936,7 @@ func (z *erasureServerPools) decommissionPool(ctx context.Context, idx int, pool
 					if isErrObjectNotFound(err) || isErrVersionNotFound(err) {
 						// object deleted by the application, nothing to do here we move on.
 						ignore = true
-						stopFn(version.Size, nil)
+						stopFn(0, nil)
 						break
 					}
 					if err != nil && !ignore {
@@ -943,7 +954,12 @@ func (z *erasureServerPools) decommissionPool(ctx context.Context, idx int, pool
 						stopFn(version.Size, err)
 						continue
 					}
-					if err = z.decommissionObject(ctx, bi.Name, gr); err != nil {
+					if err = z.decommissionObject(ctx, idx, bi.Name, gr); err != nil {
+						if isErrObjectNotFound(err) || isErrVersionNotFound(err) || isDataMovementOverWriteErr(err) {
+							ignore = true
+							stopFn(0, nil)
+							break
+						}
 						stopFn(version.Size, err)
 						failure = true
 						decomLogIf(ctx, err)
@@ -110,19 +110,14 @@ var errRebalanceNotStarted = errors.New("rebalance not started")

 func (z *erasureServerPools) loadRebalanceMeta(ctx context.Context) error {
 	r := &rebalanceMeta{}
-	err := r.load(ctx, z.serverPools[0])
-	if err != nil {
-		if errors.Is(err, errConfigNotFound) {
-			return nil
-		}
-		return err
+	if err := r.load(ctx, z.serverPools[0]); err == nil {
+		z.rebalMu.Lock()
+		z.rebalMeta = r
+		z.updateRebalanceStats(ctx)
+		z.rebalMu.Unlock()
+	} else if !errors.Is(err, errConfigNotFound) {
+		rebalanceLogIf(ctx, fmt.Errorf("failed to load rebalance metadata, continue to restart rebalance as needed: %w", err))
 	}
-
-	z.rebalMu.Lock()
-	z.rebalMeta = r
-	z.updateRebalanceStats(ctx)
-	z.rebalMu.Unlock()
-
 	return nil
 }

@@ -626,14 +621,18 @@ 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
 				}

@@ -643,6 +642,7 @@ 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
 				}

@@ -651,6 +651,7 @@ func (z *erasureServerPools) rebalanceBucket(ctx context.Context, bucket string,
 					versionID = nullVersionID
 				}

+				var failure, ignore bool
 				if version.Deleted {
 					_, err := z.DeleteObject(ctx,
 						bucket,
@@ -660,16 +661,26 @@ 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++
@@ -678,10 +689,8 @@ 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),
@@ -706,7 +715,14 @@ func (z *erasureServerPools) rebalanceBucket(ctx context.Context, bucket string,
 						continue
 					}

-					if err = z.rebalanceObject(ctx, bucket, gr); err != nil {
+					if err = z.rebalanceObject(ctx, poolIdx, bucket, gr); err != nil {
+						// This can happen when rebalance stop races with ongoing rebalance workers.
+						// These rebalance failures can be ignored.
+						if isErrObjectNotFound(err) || isErrVersionNotFound(err) || isDataMovementOverWriteErr(err) {
+							ignore = true
+							stopFn(0, nil)
+							break
+						}
 						failure = true
 						rebalanceLogIf(ctx, err)
 						stopFn(version.Size, err)
@@ -822,7 +838,7 @@ 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() {
@@ -837,9 +853,11 @@ 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)
@@ -891,6 +909,7 @@ func (z *erasureServerPools) rebalanceObject(ctx context.Context, bucket string,
 		oi.Name,
 		NewPutObjReader(hr),
 		ObjectOptions{
+			SrcPoolIdx:   poolIdx,
 			DataMovement: true,
 			VersionID:    oi.VersionID,
 			MTime:        oi.ModTime,
@@ -981,6 +1000,8 @@ const (
 	rebalanceMetricSaveMetadata
 )

+var errDataMovementSrcDstPoolSame = errors.New("source and destination pool are the same")
+
 func rebalanceTrace(r rebalanceMetric, poolIdx int, startTime time.Time, duration time.Duration, err error, path string, sz int64) madmin.TraceInfo {
 	var errStr string
 	if err != nil {
@@ -45,6 +45,7 @@ import (
 	"github.com/minio/minio/internal/logger"
 	"github.com/minio/pkg/v3/sync/errgroup"
 	"github.com/minio/pkg/v3/wildcard"
+	"github.com/puzpuzpuz/xsync/v3"
 )

 type erasureServerPools struct {
@@ -63,6 +64,8 @@ type erasureServerPools struct {
 	decommissionCancelers []context.CancelFunc

 	s3Peer *S3PeerSys
+
+	mpCache *xsync.MapOf[string, MultipartInfo]
 }

 func (z *erasureServerPools) SinglePool() bool {
@@ -168,7 +171,7 @@ func newErasureServerPools(ctx context.Context, endpointServerPools EndpointServ

 	if !globalIsDistErasure {
 		globalLocalDrivesMu.Lock()
-		globalLocalDrives = localDrives
+		globalLocalDrivesMap = make(map[string]StorageAPI, len(localDrives))
 		for _, drive := range localDrives {
 			globalLocalDrivesMap[drive.Endpoint().String()] = drive
 		}
@@ -187,19 +190,12 @@ 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)

@@ -223,20 +219,44 @@ func newErasureServerPools(ctx context.Context, endpointServerPools EndpointServ
 		break
 	}

+	// initialize the incomplete uploads cache
+	z.mpCache = xsync.NewMapOf[string, MultipartInfo]()
+
+	go z.cleanupStaleMPCache(ctx)
+
 	return z, nil
 }

-func (z *erasureServerPools) NewNSLock(bucket string, objects ...string) RWLocker {
-	poolID := hashKey(z.distributionAlgo, "", len(z.serverPools), z.deploymentID)
-	if len(objects) >= 1 {
-		poolID = hashKey(z.distributionAlgo, objects[0], len(z.serverPools), z.deploymentID)
+func (z *erasureServerPools) cleanupStaleMPCache(ctx context.Context) {
+	timer := time.NewTimer(globalAPIConfig.getStaleUploadsCleanupInterval())
+	defer timer.Stop()
+
+	for {
+		select {
+		case <-ctx.Done():
+			return
+		case <-timer.C:
+			z.mpCache.Range(func(id string, info MultipartInfo) bool {
+				if time.Since(info.Initiated) >= globalAPIConfig.getStaleUploadsExpiry() {
+					z.mpCache.Delete(id)
+					// No need to notify to peers, each node will delete its own cache.
+				}
+				return true
+			})
+
+			// Reset for the next interval
+			timer.Reset(globalAPIConfig.getStaleUploadsCleanupInterval())
+		}
 	}
-	return z.serverPools[poolID].NewNSLock(bucket, objects...)
+}
+
+func (z *erasureServerPools) NewNSLock(bucket string, objects ...string) RWLocker {
+	return z.serverPools[0].NewNSLock(bucket, objects...)
 }

 // GetDisksID will return disks by their ID.
 func (z *erasureServerPools) GetDisksID(ids ...string) []StorageAPI {
-	idMap := make(map[string]struct{})
+	idMap := make(map[string]struct{}, len(ids))
 	for _, id := range ids {
 		idMap[id] = struct{}{}
 	}
@@ -337,7 +357,7 @@ func (p serverPoolsAvailableSpace) TotalAvailable() uint64 {

 // FilterMaxUsed will filter out any pools that has used percent bigger than max,
 // unless all have that, in which case all are preserved.
-func (p serverPoolsAvailableSpace) FilterMaxUsed(max int) {
+func (p serverPoolsAvailableSpace) FilterMaxUsed(maxUsed int) {
 	// We aren't modifying p, only entries in it, so we don't need to receive a pointer.
 	if len(p) <= 1 {
 		// Nothing to do.
@@ -345,7 +365,7 @@ func (p serverPoolsAvailableSpace) FilterMaxUsed(max int) {
 	}
 	var ok bool
 	for _, z := range p {
-		if z.Available > 0 && z.MaxUsedPct < max {
+		if z.Available > 0 && z.MaxUsedPct < maxUsed {
 			ok = true
 			break
 		}
@@ -358,7 +378,7 @@ func (p serverPoolsAvailableSpace) FilterMaxUsed(max int) {

 	// Remove entries that are above.
 	for i, z := range p {
-		if z.Available > 0 && z.MaxUsedPct < max {
+		if z.Available > 0 && z.MaxUsedPct < maxUsed {
 			continue
 		}
 		p[i].Available = 0
@@ -536,7 +556,7 @@ func (z *erasureServerPools) getPoolInfoExistingWithOpts(ctx context.Context, bu
 			return pinfo, z.poolsWithObject(poolObjInfos, opts), nil
 		}
 		defPool = pinfo
-		if !isErrObjectNotFound(pinfo.Err) {
+		if !isErrObjectNotFound(pinfo.Err) && !isErrVersionNotFound(pinfo.Err) {
 			return pinfo, noReadQuorumPools, pinfo.Err
 		}

@@ -1071,23 +1091,20 @@ func (z *erasureServerPools) PutObject(ctx context.Context, bucket string, objec
 		return z.serverPools[0].PutObject(ctx, bucket, object, data, opts)
 	}

-	if !opts.NoLock {
-		ns := z.NewNSLock(bucket, object)
-		lkctx, err := ns.GetLock(ctx, globalOperationTimeout)
-		if err != nil {
-			return ObjectInfo{}, err
-		}
-		ctx = lkctx.Context()
-		defer ns.Unlock(lkctx)
-		opts.NoLock = true
-	}
-
-	idx, err := z.getPoolIdxNoLock(ctx, bucket, object, data.Size())
+	idx, err := z.getPoolIdx(ctx, bucket, object, data.Size())
 	if err != nil {
 		return ObjectInfo{}, err
 	}

-	// Overwrite the object at the right pool
+	if opts.DataMovement && idx == opts.SrcPoolIdx {
+		return ObjectInfo{}, DataMovementOverwriteErr{
+			Bucket:    bucket,
+			Object:    object,
+			VersionID: opts.VersionID,
+			Err:       errDataMovementSrcDstPoolSame,
+		}
+	}
+
 	return z.serverPools[idx].PutObject(ctx, bucket, object, data, opts)
 }

@@ -1139,13 +1156,44 @@ 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,
+		}
+	}
+
+	if opts.DataMovement {
+		objInfo, err = z.serverPools[pinfo.Index].DeleteObject(ctx, bucket, object, opts)
+		objInfo.Name = decodeDirObject(object)
+		return objInfo, err
+	}
+
 	// 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)
 	}
-	objInfo, err = z.serverPools[pinfo.Index].DeleteObject(ctx, bucket, object, opts)
+
+	for _, pool := range z.serverPools {
+		objInfo, err := pool.DeleteObject(ctx, bucket, object, opts)
+		if err != nil && !isErrObjectNotFound(err) && !isErrVersionNotFound(err) {
+			objInfo.Name = decodeDirObject(object)
+			return objInfo, err
+		}
+		if err == nil {
+			objInfo.Name = decodeDirObject(object)
+			return objInfo, nil
+		}
+	}
+
 	objInfo.Name = decodeDirObject(object)
-	return objInfo, err
+	if opts.VersionID != "" {
+		return objInfo, VersionNotFound{Bucket: bucket, Object: object, VersionID: opts.VersionID}
+	}
+	return objInfo, ObjectNotFound{Bucket: bucket, Object: object}
 }

 func (z *erasureServerPools) deleteObjectFromAllPools(ctx context.Context, bucket string, object string, opts ObjectOptions, poolIndices []poolErrs) (objInfo ObjectInfo, err error) {
@@ -1264,19 +1312,20 @@ func (z *erasureServerPools) DeleteObjects(ctx context.Context, bucket string, o
 			go func(idx int, pool *erasureSets) {
 				defer wg.Done()
 				objs := poolObjIdxMap[idx]
-				if len(objs) > 0 {
-					orgIndexes := origIndexMap[idx]
-					deletedObjects, errs := pool.DeleteObjects(ctx, bucket, objs, opts)
-					mu.Lock()
-					for i, derr := range errs {
-						if derr != nil {
-							derrs[orgIndexes[i]] = derr
-						}
-						deletedObjects[i].ObjectName = decodeDirObject(deletedObjects[i].ObjectName)
-						dobjects[orgIndexes[i]] = deletedObjects[i]
-					}
-					mu.Unlock()
+				if len(objs) == 0 {
+					return
 				}
+				orgIndexes := origIndexMap[idx]
+				deletedObjects, errs := pool.DeleteObjects(ctx, bucket, objs, opts)
+				mu.Lock()
+				for i, derr := range errs {
+					if derr != nil {
+						derrs[orgIndexes[i]] = derr
+					}
+					deletedObjects[i].ObjectName = decodeDirObject(deletedObjects[i].ObjectName)
+					dobjects[orgIndexes[i]] = deletedObjects[i]
+				}
+				mu.Unlock()
 			}(idx, pool)
 		}
 		wg.Wait()
@@ -1526,14 +1575,12 @@ 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 != "" {
@@ -1697,15 +1744,32 @@ func (z *erasureServerPools) ListMultipartUploads(ctx context.Context, bucket, p
 		return ListMultipartsInfo{}, err
 	}

-	if z.SinglePool() {
-		return z.serverPools[0].ListMultipartUploads(ctx, bucket, prefix, keyMarker, uploadIDMarker, delimiter, maxUploads)
-	}
-
 	poolResult := ListMultipartsInfo{}
 	poolResult.MaxUploads = maxUploads
 	poolResult.KeyMarker = keyMarker
 	poolResult.Prefix = prefix
 	poolResult.Delimiter = delimiter
+
+	// if no prefix provided, return the list from cache
+	if prefix == "" {
+		if _, err := z.GetBucketInfo(ctx, bucket, BucketOptions{}); err != nil {
+			return ListMultipartsInfo{}, toObjectErr(err, bucket)
+		}
+
+		z.mpCache.Range(func(_ string, mp MultipartInfo) bool {
+			poolResult.Uploads = append(poolResult.Uploads, mp)
+			return true
+		})
+		sort.Slice(poolResult.Uploads, func(i int, j int) bool {
+			return poolResult.Uploads[i].Initiated.Before(poolResult.Uploads[j].Initiated)
+		})
+		return poolResult, nil
+	}
+
+	if z.SinglePool() {
+		return z.serverPools[0].ListMultipartUploads(ctx, bucket, prefix, keyMarker, uploadIDMarker, delimiter, maxUploads)
+	}
+
 	for idx, pool := range z.serverPools {
 		if z.IsSuspended(idx) {
 			continue
@@ -1717,15 +1781,27 @@ func (z *erasureServerPools) ListMultipartUploads(ctx context.Context, bucket, p
 		}
 		poolResult.Uploads = append(poolResult.Uploads, result.Uploads...)
 	}
+
 	return poolResult, nil
 }

 // Initiate a new multipart upload on a hashedSet based on object name.
-func (z *erasureServerPools) NewMultipartUpload(ctx context.Context, bucket, object string, opts ObjectOptions) (*NewMultipartUploadResult, error) {
+func (z *erasureServerPools) NewMultipartUpload(ctx context.Context, bucket, object string, opts ObjectOptions) (mp *NewMultipartUploadResult, err error) {
 	if err := checkNewMultipartArgs(ctx, bucket, object); err != nil {
 		return nil, err
 	}

+	defer func() {
+		if err == nil && mp != nil {
+			z.mpCache.Store(mp.UploadID, MultipartInfo{
+				Bucket:    bucket,
+				Object:    object,
+				UploadID:  mp.UploadID,
+				Initiated: time.Now(),
+			})
+		}
+	}()
+
 	if z.SinglePool() {
 		return z.serverPools[0].NewMultipartUpload(ctx, bucket, object, opts)
 	}
@@ -1754,6 +1830,15 @@ 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)
 }

@@ -1805,9 +1890,18 @@ func (z *erasureServerPools) GetMultipartInfo(ctx context.Context, bucket, objec
 		return MultipartInfo{}, err
 	}

+	uploadIDLock := z.NewNSLock(bucket, pathJoin(object, uploadID))
+	lkctx, err := uploadIDLock.GetRLock(ctx, globalOperationTimeout)
+	if err != nil {
+		return MultipartInfo{}, err
+	}
+	ctx = lkctx.Context()
+	defer uploadIDLock.RUnlock(lkctx)
+
 	if z.SinglePool() {
 		return z.serverPools[0].GetMultipartInfo(ctx, bucket, object, uploadID, opts)
 	}
+
 	for idx, pool := range z.serverPools {
 		if z.IsSuspended(idx) {
 			continue
@@ -1823,6 +1917,7 @@ func (z *erasureServerPools) GetMultipartInfo(ctx context.Context, bucket, objec
 		// any other unhandled error return right here.
 		return MultipartInfo{}, err
 	}
+
 	return MultipartInfo{}, InvalidUploadID{
 		Bucket:   bucket,
 		Object:   object,
@@ -1836,9 +1931,18 @@ func (z *erasureServerPools) ListObjectParts(ctx context.Context, bucket, object
 		return ListPartsInfo{}, err
 	}

+	uploadIDLock := z.NewNSLock(bucket, pathJoin(object, uploadID))
+	lkctx, err := uploadIDLock.GetRLock(ctx, globalOperationTimeout)
+	if err != nil {
+		return ListPartsInfo{}, err
+	}
+	ctx = lkctx.Context()
+	defer uploadIDLock.RUnlock(lkctx)
+
 	if z.SinglePool() {
 		return z.serverPools[0].ListObjectParts(ctx, bucket, object, uploadID, partNumberMarker, maxParts, opts)
 	}
+
 	for idx, pool := range z.serverPools {
 		if z.IsSuspended(idx) {
 			continue
@@ -1852,6 +1956,7 @@ func (z *erasureServerPools) ListObjectParts(ctx context.Context, bucket, object
 		}
 		return ListPartsInfo{}, err
 	}
+
 	return ListPartsInfo{}, InvalidUploadID{
 		Bucket:   bucket,
 		Object:   object,
@@ -1860,11 +1965,26 @@ func (z *erasureServerPools) ListObjectParts(ctx context.Context, bucket, object
 }

 // Aborts an in-progress multipart operation on hashedSet based on the object name.
-func (z *erasureServerPools) AbortMultipartUpload(ctx context.Context, bucket, object, uploadID string, opts ObjectOptions) error {
+func (z *erasureServerPools) AbortMultipartUpload(ctx context.Context, bucket, object, uploadID string, opts ObjectOptions) (err error) {
 	if err := checkAbortMultipartArgs(ctx, bucket, object, uploadID); err != nil {
 		return err
 	}

+	defer func() {
+		if err == nil {
+			z.mpCache.Delete(uploadID)
+			globalNotificationSys.DeleteUploadID(ctx, uploadID)
+		}
+	}()
+
+	lk := z.NewNSLock(bucket, pathJoin(object, uploadID))
+	lkctx, err := lk.GetLock(ctx, globalOperationTimeout)
+	if err != nil {
+		return err
+	}
+	ctx = lkctx.Context()
+	defer lk.Unlock(lkctx)
+
 	if z.SinglePool() {
 		return z.serverPools[0].AbortMultipartUpload(ctx, bucket, object, uploadID, opts)
 	}
@@ -1896,6 +2016,23 @@ func (z *erasureServerPools) CompleteMultipartUpload(ctx context.Context, bucket
 		return objInfo, err
 	}

+	defer func() {
+		if err == nil {
+			z.mpCache.Delete(uploadID)
+			globalNotificationSys.DeleteUploadID(ctx, uploadID)
+		}
+	}()
+
+	// Hold write locks to verify uploaded parts, also disallows any
+	// parallel PutObjectPart() requests.
+	uploadIDLock := z.NewNSLock(bucket, pathJoin(object, uploadID))
+	wlkctx, err := uploadIDLock.GetLock(ctx, globalOperationTimeout)
+	if err != nil {
+		return objInfo, err
+	}
+	ctx = wlkctx.Context()
+	defer uploadIDLock.Unlock(wlkctx)
+
 	if z.SinglePool() {
 		return z.serverPools[0].CompleteMultipartUpload(ctx, bucket, object, uploadID, uploadedParts, opts)
 	}
@@ -1938,6 +2075,12 @@ func (z *erasureServerPools) GetBucketInfo(ctx context.Context, bucket string, o
 	return bucketInfo, nil
 }

+// ClearUploadID deletes given uploadID from cache
+func (z *erasureServerPools) ClearUploadID(uploadID string) error {
+	z.mpCache.Delete(uploadID)
+	return nil
+}
+
 // DeleteBucket - deletes a bucket on all serverPools simultaneously,
 // even if one of the serverPools fail to delete buckets, we proceed to
 // undo a successful operation.
@@ -1962,6 +2105,34 @@ 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
@@ -2104,7 +2275,7 @@ 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
 			}
@@ -2180,6 +2351,7 @@ 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,
@@ -2296,6 +2468,7 @@ func (z *erasureServerPools) Walk(ctx context.Context, bucket, prefix string, re
 // HealObjectFn closure function heals the object.
 type HealObjectFn func(bucket, object, versionID string, scanMode madmin.HealScanMode) error

+// List a prefix or a single object versions and heal
 func (z *erasureServerPools) HealObjects(ctx context.Context, bucket, prefix string, opts madmin.HealOpts, healObjectFn HealObjectFn) error {
 	healEntry := func(bucket string, entry metaCacheEntry, scanMode madmin.HealScanMode) error {
 		if entry.isDir() {
@@ -2359,7 +2532,7 @@ func (z *erasureServerPools) HealObjects(ctx context.Context, bucket, prefix str
 			go func(idx int, set *erasureObjects) {
 				defer wg.Done()

-				errs[idx] = set.listAndHeal(ctx, bucket, prefix, opts.ScanMode, healEntry)
+				errs[idx] = set.listAndHeal(ctx, bucket, prefix, opts.Recursive, opts.ScanMode, healEntry)
 			}(idx, set)
 		}
 		wg.Wait()
@@ -2411,15 +2584,22 @@ func (z *erasureServerPools) HealObject(ctx context.Context, bucket, object, ver
 		}
 	}

+	hr := madmin.HealResultItem{
+		Type:      madmin.HealItemObject,
+		Bucket:    bucket,
+		Object:    object,
+		VersionID: versionID,
+	}
+
 	// At this stage, all errors are 'not found'
 	if versionID != "" {
-		return madmin.HealResultItem{}, VersionNotFound{
+		return hr, VersionNotFound{
 			Bucket:    bucket,
 			Object:    object,
 			VersionID: versionID,
 		}
 	}
-	return madmin.HealResultItem{}, ObjectNotFound{
+	return hr, ObjectNotFound{
 		Bucket: bucket,
 		Object: object,
 	}
@@ -2640,7 +2820,19 @@ func (z *erasureServerPools) PutObjectMetadata(ctx context.Context, bucket, obje
 		return z.serverPools[0].PutObjectMetadata(ctx, bucket, object, opts)
 	}

+	if !opts.NoLock {
+		// Lock the object before updating metadata.
+		lk := z.NewNSLock(bucket, object)
+		lkctx, err := lk.GetLock(ctx, globalOperationTimeout)
+		if err != nil {
+			return ObjectInfo{}, err
+		}
+		ctx = lkctx.Context()
+		defer lk.Unlock(lkctx)
+	}
+
 	opts.MetadataChg = true
+	opts.NoLock = true
 	// We don't know the size here set 1GiB at least.
 	idx, err := z.getPoolIdxExistingWithOpts(ctx, bucket, object, opts)
 	if err != nil {
@@ -2657,7 +2849,19 @@ func (z *erasureServerPools) PutObjectTags(ctx context.Context, bucket, object s
 		return z.serverPools[0].PutObjectTags(ctx, bucket, object, tags, opts)
 	}

+	if !opts.NoLock {
+		// Lock the object before updating tags.
+		lk := z.NewNSLock(bucket, object)
+		lkctx, err := lk.GetLock(ctx, globalOperationTimeout)
+		if err != nil {
+			return ObjectInfo{}, err
+		}
+		ctx = lkctx.Context()
+		defer lk.Unlock(lkctx)
+	}
+
 	opts.MetadataChg = true
+	opts.NoLock = true

 	// We don't know the size here set 1GiB at least.
 	idx, err := z.getPoolIdxExistingWithOpts(ctx, bucket, object, opts)
@@ -2675,8 +2879,19 @@ func (z *erasureServerPools) DeleteObjectTags(ctx context.Context, bucket, objec
 		return z.serverPools[0].DeleteObjectTags(ctx, bucket, object, opts)
 	}

-	opts.MetadataChg = true
+	if !opts.NoLock {
+		// Lock the object before deleting tags.
+		lk := z.NewNSLock(bucket, object)
+		lkctx, err := lk.GetLock(ctx, globalOperationTimeout)
+		if err != nil {
+			return ObjectInfo{}, err
+		}
+		ctx = lkctx.Context()
+		defer lk.Unlock(lkctx)
+	}

+	opts.MetadataChg = true
+	opts.NoLock = true
 	idx, err := z.getPoolIdxExistingWithOpts(ctx, bucket, object, opts)
 	if err != nil {
 		return ObjectInfo{}, err
@@ -2707,8 +2922,20 @@ func (z *erasureServerPools) TransitionObject(ctx context.Context, bucket, objec
 		return z.serverPools[0].TransitionObject(ctx, bucket, object, opts)
 	}

+	if !opts.NoLock {
+		// Acquire write lock before starting to transition the object.
+		lk := z.NewNSLock(bucket, object)
+		lkctx, err := lk.GetLock(ctx, globalDeleteOperationTimeout)
+		if err != nil {
+			return err
+		}
+		ctx = lkctx.Context()
+		defer lk.Unlock(lkctx)
+	}
+
 	// Avoid transitioning an object from a pool being decommissioned.
 	opts.SkipDecommissioned = true
+	opts.NoLock = true
 	idx, err := z.getPoolIdxExistingWithOpts(ctx, bucket, object, opts)
 	if err != nil {
 		return err
@@ -2724,8 +2951,20 @@ func (z *erasureServerPools) RestoreTransitionedObject(ctx context.Context, buck
 		return z.serverPools[0].RestoreTransitionedObject(ctx, bucket, object, opts)
 	}

+	if !opts.NoLock {
+		// Acquire write lock before restoring transitioned object
+		lk := z.NewNSLock(bucket, object)
+		lkctx, err := lk.GetLock(ctx, globalDeleteOperationTimeout)
+		if err != nil {
+			return err
+		}
+		ctx = lkctx.Context()
+		defer lk.Unlock(lkctx)
+	}
+
 	// Avoid restoring object from a pool being decommissioned.
 	opts.SkipDecommissioned = true
+	opts.NoLock = true
 	idx, err := z.getPoolIdxExistingWithOpts(ctx, bucket, object, opts)
 	if err != nil {
 		return err
@@ -2782,5 +3021,14 @@ 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)
 }
@@ -39,6 +39,7 @@ import (
 	"github.com/minio/minio/internal/logger"
 	"github.com/minio/pkg/v3/console"
 	"github.com/minio/pkg/v3/sync/errgroup"
+	"github.com/puzpuzpuz/xsync/v3"
 )

 // setsDsyncLockers is encapsulated type for Close()
@@ -86,6 +87,8 @@ type erasureSets struct {
 	lastConnectDisksOpTime time.Time
 }

+var staleUploadsCleanupIntervalChangedCh = make(chan struct{})
+
 func (s *erasureSets) getDiskMap() map[Endpoint]StorageAPI {
 	diskMap := make(map[Endpoint]StorageAPI)

@@ -230,8 +233,11 @@ func (s *erasureSets) connectDisks(log bool) {
 				}
 				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)
@@ -262,13 +268,7 @@ func (s *erasureSets) connectDisks(log bool) {
 				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()
@@ -531,46 +531,47 @@ func (s *erasureSets) cleanupStaleUploads(ctx context.Context) {
 					if set == nil {
 						return
 					}
-					set.cleanupStaleUploads(ctx, globalAPIConfig.getStaleUploadsExpiry())
+					set.cleanupStaleUploads(ctx)
 				}(set)
 			}
 			wg.Wait()
-
-			// Reset for the next interval
-			timer.Reset(globalAPIConfig.getStaleUploadsCleanupInterval())
+		case <-staleUploadsCleanupIntervalChangedCh:
 		}
+
+		// Reset for the next interval
+		timer.Reset(globalAPIConfig.getStaleUploadsCleanupInterval())
 	}
 }

 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.
 func (s *erasureSets) NewNSLock(bucket string, objects ...string) RWLocker {
-	if len(objects) == 1 {
-		return s.getHashedSet(objects[0]).NewNSLock(bucket, objects...)
-	}
-	return s.getHashedSet("").NewNSLock(bucket, objects...)
+	return s.sets[0].NewNSLock(bucket, objects...)
 }

 // SetDriveCount returns the current drives per set.
@@ -701,9 +702,8 @@ func (s *erasureSets) getHashedSet(input string) (set *erasureObjects) {
 }

 // listDeletedBuckets lists deleted buckets from all disks.
-func listDeletedBuckets(ctx context.Context, storageDisks []StorageAPI, delBuckets map[string]VolInfo, readQuorum int) error {
+func listDeletedBuckets(ctx context.Context, storageDisks []StorageAPI, delBuckets *xsync.MapOf[string, VolInfo], readQuorum int) error {
 	g := errgroup.WithNErrs(len(storageDisks))
-	var mu sync.Mutex
 	for index := range storageDisks {
 		index := index
 		g.Go(func() error {
@@ -722,11 +722,7 @@ func listDeletedBuckets(ctx context.Context, storageDisks []StorageAPI, delBucke
 				vi, err := storageDisks[index].StatVol(ctx, pathJoin(minioMetaBucket, bucketMetaPrefix, deletedBucketsPrefix, volName))
 				if err == nil {
 					vi.Name = strings.TrimSuffix(volName, SlashSeparator)
-					mu.Lock()
-					if _, ok := delBuckets[volName]; !ok {
-						delBuckets[volName] = vi
-					}
-					mu.Unlock()
+					delBuckets.Store(volName, vi)
 				}
 			}
 			return nil
@@ -1135,13 +1131,7 @@ 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()
 				}
 			}
@@ -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):
@@ -198,11 +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)),
@@ -548,9 +551,6 @@ func (er erasureObjects) nsScanner(ctx context.Context, buckets []BucketInfo, wa
 				var root dataUsageEntry
 				if r := cache.root(); r != nil {
 					root = cache.flatten(*r)
-					if root.ReplicationStats.empty() {
-						root.ReplicationStats = nil
-					}
 				}
 				select {
 				case <-ctx.Done():
@@ -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)
+			}
+		}
+	}
+}
@@ -24,6 +24,8 @@ import (
 	"errors"
 	"fmt"
 	"io"
+	"net"
+	"net/http"
 	"os"
 	"path"
 	"strings"
@@ -114,7 +116,7 @@ func ftpTrace(s *ftp.Context, startTime time.Time, source, objPath string, err e
 		TraceType: madmin.TraceFTP,
 		Time:      startTime,
 		NodeName:  globalLocalNodeName,
-		FuncName:  fmt.Sprintf(s.Cmd),
+		FuncName:  s.Cmd,
 		Duration:  time.Since(startTime),
 		Path:      objPath,
 		Error:     errStr,
@@ -286,6 +288,10 @@ func (driver *ftpDriver) CheckPasswd(c *ftp.Context, username, password string)
 }

 func (driver *ftpDriver) getMinIOClient(ctx *ftp.Context) (*minio.Client, error) {
+	tr := http.RoundTripper(globalRemoteFTPClientTransport)
+	if host, _, err := net.SplitHostPort(ctx.Sess.RemoteAddr().String()); err == nil {
+		tr = forwardForTransport{tr: tr, fwd: host}
+	}
 	ui, ok := globalIAMSys.GetUser(context.Background(), ctx.Sess.LoginUser())
 	if !ok && !globalIAMSys.LDAPConfig.Enabled() {
 		return nil, errNoSuchUser
@@ -363,7 +369,7 @@ func (driver *ftpDriver) getMinIOClient(ctx *ftp.Context) (*minio.Client, error)
 		return minio.New(driver.endpoint, &minio.Options{
 			Creds:     mcreds,
 			Secure:    globalIsTLS,
-			Transport: globalRemoteFTPClientTransport,
+			Transport: tr,
 		})
 	}

@@ -377,7 +383,7 @@ func (driver *ftpDriver) getMinIOClient(ctx *ftp.Context) (*minio.Client, error)
 	return minio.New(driver.endpoint, &minio.Options{
 		Creds:     credentials.NewStaticV4(ui.Credentials.AccessKey, ui.Credentials.SecretKey, ""),
 		Secure:    globalIsTLS,
-		Transport: globalRemoteFTPClientTransport,
+		Transport: tr,
 	})
 }

@@ -33,8 +33,6 @@ import (
 	"github.com/minio/minio-go/v7/pkg/set"
 	"github.com/minio/minio/internal/grid"
 	xnet "github.com/minio/pkg/v3/net"
-	"golang.org/x/exp/maps"
-	"golang.org/x/exp/slices"

 	"github.com/minio/minio/internal/amztime"
 	"github.com/minio/minio/internal/config/dns"
@@ -75,7 +73,7 @@ const (
 // and must not set by clients
 func containsReservedMetadata(header http.Header) bool {
 	for key := range header {
-		if slices.Contains(maps.Keys(validSSEReplicationHeaders), key) {
+		if _, ok := validSSEReplicationHeaders[key]; ok {
 			return false
 		}
 		if stringsHasPrefixFold(key, ReservedMetadataPrefix) {
@@ -247,11 +245,14 @@ 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 +470,7 @@ 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
 		}
@@ -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 {
--- a/Show More
+++ b/Show More