Add postponed send loop to QEMU driver

- Measure and report scrub I/O statistics in vitastor-cli status
- Make aggregated statistics in vitastor-cli status much smoother
  (first derive, then sum instead of first summing and then deriving)
- Fix an old rare bug leading to journal corruption
  (try to use scrub if you think you're affected...)
- Do not start EC PGs without at least <data chunks> OSDs in each old set
  (prevents spurious read errors with EC during reconnections/restarts)
- Fix failed assert(!scrub_list_op) on OSD restart with pending scrubs
- Fix future planned scrubs not starting because of incorrect time comparison
- Build packages for Debian 12 (Bookworm)

.gitea/workflows/test.Dockerfile

@@ -10,6 +10,9 @@ RUN set -e -x; \
     ln -s /root/fio-build/fio-*/ ./fio; \
     ln -s /root/qemu-build/qemu-*/ ./qemu; \
     ls /usr/include/linux/raw.h || cp ./debian/raw.h /usr/include/linux/raw.h; \
+    cd mon; \
+    npm install; \
+    cd ..; \
     mkdir build; \
     cd build; \
     cmake .. -DWITH_ASAN=yes -DWITH_QEMU=yes; \

.gitea/workflows/test.yml

@@ -71,7 +71,7 @@ jobs:
     steps:
     - name: Run test
       id: test
-      timeout-minutes: 3
+      timeout-minutes: 10
       run: /root/vitastor/tests/test_add_osd.sh
     - name: Print logs
       if: always() && steps.test.outcome == 'failure'
@@ -190,24 +190,6 @@ jobs:
           echo ""
         done
 
-  test_failure_domain:
-    runs-on: ubuntu-latest
-    needs: build
-    container: ${{env.TEST_IMAGE}}:${{github.sha}}
-    steps:
-    - name: Run test
-      id: test
-      timeout-minutes: 3
-      run: /root/vitastor/tests/test_failure_domain.sh
-    - name: Print logs
-      if: always() && steps.test.outcome == 'failure'
-      run: |
-        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
-          echo "-------- $i --------"
-          cat $i
-          echo ""
-        done
-
   test_interrupted_rebalance:
     runs-on: ubuntu-latest
     needs: build
@@ -280,7 +262,7 @@ jobs:
           echo ""
         done
 
-  test_minsize_1:
+  test_failure_domain:
     runs-on: ubuntu-latest
     needs: build
     container: ${{env.TEST_IMAGE}}:${{github.sha}}
@@ -288,115 +270,7 @@ jobs:
     - name: Run test
       id: test
       timeout-minutes: 3
-      run: /root/vitastor/tests/test_minsize_1.sh
-    - name: Print logs
-      if: always() && steps.test.outcome == 'failure'
-      run: |
-        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
-          echo "-------- $i --------"
-          cat $i
-          echo ""
-        done
-
-  test_move_reappear:
-    runs-on: ubuntu-latest
-    needs: build
-    container: ${{env.TEST_IMAGE}}:${{github.sha}}
-    steps:
-    - name: Run test
-      id: test
-      timeout-minutes: 3
-      run: /root/vitastor/tests/test_move_reappear.sh
-    - name: Print logs
-      if: always() && steps.test.outcome == 'failure'
-      run: |
-        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
-          echo "-------- $i --------"
-          cat $i
-          echo ""
-        done
-
-  test_rebalance_verify:
-    runs-on: ubuntu-latest
-    needs: build
-    container: ${{env.TEST_IMAGE}}:${{github.sha}}
-    steps:
-    - name: Run test
-      id: test
-      timeout-minutes: 3
-      run: /root/vitastor/tests/test_rebalance_verify.sh
-    - name: Print logs
-      if: always() && steps.test.outcome == 'failure'
-      run: |
-        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
-          echo "-------- $i --------"
-          cat $i
-          echo ""
-        done
-
-  test_rebalance_verify_imm:
-    runs-on: ubuntu-latest
-    needs: build
-    container: ${{env.TEST_IMAGE}}:${{github.sha}}
-    steps:
-    - name: Run test
-      id: test
-      timeout-minutes: 3
-      run: IMMEDIATE_COMMIT=1 /root/vitastor/tests/test_rebalance_verify.sh
-    - name: Print logs
-      if: always() && steps.test.outcome == 'failure'
-      run: |
-        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
-          echo "-------- $i --------"
-          cat $i
-          echo ""
-        done
-
-  test_rebalance_verify_ec:
-    runs-on: ubuntu-latest
-    needs: build
-    container: ${{env.TEST_IMAGE}}:${{github.sha}}
-    steps:
-    - name: Run test
-      id: test
-      timeout-minutes: 3
-      run: SCHEME=ec /root/vitastor/tests/test_rebalance_verify.sh
-    - name: Print logs
-      if: always() && steps.test.outcome == 'failure'
-      run: |
-        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
-          echo "-------- $i --------"
-          cat $i
-          echo ""
-        done
-
-  test_rebalance_verify_ec_imm:
-    runs-on: ubuntu-latest
-    needs: build
-    container: ${{env.TEST_IMAGE}}:${{github.sha}}
-    steps:
-    - name: Run test
-      id: test
-      timeout-minutes: 3
-      run: SCHEME=ec IMMEDIATE_COMMIT=1 /root/vitastor/tests/test_rebalance_verify.sh
-    - name: Print logs
-      if: always() && steps.test.outcome == 'failure'
-      run: |
-        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
-          echo "-------- $i --------"
-          cat $i
-          echo ""
-        done
-
-  test_rm:
-    runs-on: ubuntu-latest
-    needs: build
-    container: ${{env.TEST_IMAGE}}:${{github.sha}}
-    steps:
-    - name: Run test
-      id: test
-      timeout-minutes: 3
-      run: /root/vitastor/tests/test_rm.sh
+      run: /root/vitastor/tests/test_failure_domain.sh
     - name: Print logs
       if: always() && steps.test.outcome == 'failure'
       run: |
@@ -442,6 +316,132 @@ jobs:
           echo ""
         done
 
+  test_minsize_1:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: /root/vitastor/tests/test_minsize_1.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
+  test_move_reappear:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: /root/vitastor/tests/test_move_reappear.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
+  test_rm:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: /root/vitastor/tests/test_rm.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
+  test_snapshot_chain:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: /root/vitastor/tests/test_snapshot_chain.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
+  test_snapshot_chain_ec:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: SCHEME=ec /root/vitastor/tests/test_snapshot_chain.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
+  test_snapshot_down:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: /root/vitastor/tests/test_snapshot_down.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
+  test_snapshot_down_ec:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: SCHEME=ec /root/vitastor/tests/test_snapshot_down.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
   test_splitbrain:
     runs-on: ubuntu-latest
     needs: build
@@ -460,6 +460,78 @@ jobs:
           echo ""
         done
 
+  test_rebalance_verify:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 10
+      run: /root/vitastor/tests/test_rebalance_verify.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
+  test_rebalance_verify_imm:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 10
+      run: IMMEDIATE_COMMIT=1 /root/vitastor/tests/test_rebalance_verify.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
+  test_rebalance_verify_ec:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 10
+      run: SCHEME=ec /root/vitastor/tests/test_rebalance_verify.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
+  test_rebalance_verify_ec_imm:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 10
+      run: SCHEME=ec IMMEDIATE_COMMIT=1 /root/vitastor/tests/test_rebalance_verify.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
   test_write:
     runs-on: ubuntu-latest
     needs: build
@@ -550,3 +622,111 @@ jobs:
           echo ""
         done
 
+  test_scrub:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: /root/vitastor/tests/test_scrub.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
+  test_scrub_zero_osd_2:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: ZERO_OSD=2 /root/vitastor/tests/test_scrub.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
+  test_scrub_xor:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: SCHEME=xor /root/vitastor/tests/test_scrub.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
+  test_scrub_pg_size_3:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: PG_SIZE=3 /root/vitastor/tests/test_scrub.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
+  test_scrub_pg_size_6_pg_minsize_4_osd_count_6_ec:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: PG_SIZE=6 PG_MINSIZE=4 OSD_COUNT=6 SCHEME=ec /root/vitastor/tests/test_scrub.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
+  test_scrub_ec:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: SCHEME=ec /root/vitastor/tests/test_scrub.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+

.gitea/workflows/tests-to-yaml.pl

@@ -9,7 +9,8 @@ for my $line (<>)
         chomp $line;
         my $test_name = $1;
         my $timeout = 3;
-        if ($test_name eq 'test_etcd_fail' || $test_name eq 'test_heal' || $test_name eq 'test_interrupted_rebalance')
+        if ($test_name eq 'test_etcd_fail' || $test_name eq 'test_heal' || $test_name eq 'test_add_osd' ||
+            $test_name eq 'test_interrupted_rebalance' || $test_name eq 'test_rebalance_verify')
         {
             $timeout = 10;
         }

CMakeLists.txt

@@ -2,6 +2,6 @@ cmake_minimum_required(VERSION 2.8.12)
 
 project(vitastor)
 
-set(VERSION "0.8.9")
+set(VERSION "0.9.2")
 
 add_subdirectory(src)

csi/Makefile

@@ -1,4 +1,4 @@
-VERSION ?= v0.8.9
+VERSION ?= v0.9.2
 
 all: build push
 

csi/deploy/004-csi-nodeplugin.yaml

@@ -49,7 +49,7 @@ spec:
             capabilities:
               add: ["SYS_ADMIN"]
             allowPrivilegeEscalation: true
-          image: vitalif/vitastor-csi:v0.8.9
+          image: vitalif/vitastor-csi:v0.9.2
           args:
             - "--node=$(NODE_ID)"
             - "--endpoint=$(CSI_ENDPOINT)"

csi/deploy/007-csi-provisioner.yaml

@@ -116,7 +116,7 @@ spec:
             privileged: true
             capabilities:
               add: ["SYS_ADMIN"]
-          image: vitalif/vitastor-csi:v0.8.9
+          image: vitalif/vitastor-csi:v0.9.2
           args:
             - "--node=$(NODE_ID)"
             - "--endpoint=$(CSI_ENDPOINT)"

csi/src/config.go

@@ -5,7 +5,7 @@ package vitastor
 
 const (
     vitastorCSIDriverName    = "csi.vitastor.io"
-    vitastorCSIDriverVersion = "0.8.9"
+    vitastorCSIDriverVersion = "0.9.2"
 )
 
 // Config struct fills the parameters of request or user input

debian/build-pve-qemu.sh

@@ -0,0 +1,58 @@
+exit
+
+git clone https://git.yourcmc.ru/vitalif/pve-qemu .
+
+# bookworm
+
+docker run -it -v `pwd`/pve-qemu:/root/pve-qemu --name pve-qemu-bullseye debian:bullseye bash
+
+perl -i -pe 's/Types: deb$/Types: deb deb-src/' /etc/apt/sources.list.d/debian.sources
+echo 'deb [arch=amd64] http://download.proxmox.com/debian/pve bookworm pve-no-subscription' >> /etc/apt/sources.list
+echo 'deb https://vitastor.io/debian bookworm main' >> /etc/apt/sources.list
+echo 'APT::Install-Recommends false;' >> /etc/apt/apt.conf
+echo 'ru_RU UTF-8' >> /etc/locale.gen
+echo 'en_US UTF-8' >> /etc/locale.gen
+apt-get update
+apt-get install wget ca-certificates
+wget https://enterprise.proxmox.com/debian/proxmox-release-bookworm.gpg -O /etc/apt/trusted.gpg.d/proxmox-release-bookworm.gpg
+wget https://vitastor.io/debian/pubkey.gpg -O /etc/apt/trusted.gpg.d/vitastor.gpg
+apt-get update
+apt-get install git devscripts equivs wget mc libjemalloc-dev vitastor-client-dev lintian locales
+mk-build-deps --install ./control
+
+# bullseye
+
+docker run -it -v `pwd`/pve-qemu:/root/pve-qemu --name pve-qemu-bullseye debian:bullseye bash
+
+grep '^deb ' /etc/apt/sources.list | perl -pe 's/^deb /deb-src /' >> /etc/apt/sources.list
+echo 'deb [arch=amd64] http://download.proxmox.com/debian/pve bullseye pve-no-subscription' >> /etc/apt/sources.list
+echo 'deb https://vitastor.io/debian bullseye main' >> /etc/apt/sources.list
+echo 'APT::Install-Recommends false;' >> /etc/apt/apt.conf
+echo 'ru_RU UTF-8' >> /etc/locale.gen
+echo 'en_US UTF-8' >> /etc/locale.gen
+apt-get update
+apt-get install wget
+wget https://enterprise.proxmox.com/debian/proxmox-release-bullseye.gpg -O /etc/apt/trusted.gpg.d/proxmox-release-bullseye.gpg
+wget https://vitastor.io/debian/pubkey.gpg -O /etc/apt/trusted.gpg.d/vitastor.gpg
+apt-get update
+apt-get install git devscripts equivs wget mc libjemalloc-dev vitastor-client-dev lintian locales
+mk-build-deps --install ./control
+
+# buster
+
+docker run -it -v `pwd`/pve-qemu:/root/pve-qemu --name pve-qemu-buster debian:buster bash
+
+grep '^deb ' /etc/apt/sources.list | perl -pe 's/^deb /deb-src /' >> /etc/apt/sources.list
+echo 'deb [arch=amd64] http://download.proxmox.com/debian/pve buster pve-no-subscription' >> /etc/apt/sources.list
+echo 'deb https://vitastor.io/debian buster main' >> /etc/apt/sources.list
+echo 'deb http://deb.debian.org/debian buster-backports main' >> /etc/apt/sources.list
+echo 'APT::Install-Recommends false;' >> /etc/apt/apt.conf
+echo 'ru_RU UTF-8' >> /etc/locale.gen
+echo 'en_US UTF-8' >> /etc/locale.gen
+apt-get update
+apt-get install wget ca-certificates
+wget http://download.proxmox.com/debian/proxmox-ve-release-6.x.gpg -O /etc/apt/trusted.gpg.d/proxmox-ve-release-6.x.gpg
+wget https://vitastor.io/debian/pubkey.gpg -O /etc/apt/trusted.gpg.d/vitastor.gpg
+apt-get update
+apt-get install git devscripts equivs wget mc libjemalloc-dev vitastor-client-dev lintian locales
+mk-build-deps --install ./control

debian/build-vitastor-bookworm.sh

@@ -0,0 +1,7 @@
+#!/bin/bash
+
+cat < vitastor.Dockerfile > ../Dockerfile
+cd ..
+mkdir -p packages
+sudo podman build --build-arg REL=bookworm -v `pwd`/packages:/root/packages -f Dockerfile .
+rm Dockerfile

debian/changelog

@@ -1,10 +1,10 @@
-vitastor (0.8.9-1) unstable; urgency=medium
+vitastor (0.9.2-1) unstable; urgency=medium
 
   * Bugfixes
 
  -- Vitaliy Filippov <vitalif@yourcmc.ru>  Fri, 03 Jun 2022 02:09:44 +0300
 
-vitastor (0.8.9-1) unstable; urgency=medium
+vitastor (0.9.2-1) unstable; urgency=medium
 
   * Implement NFS proxy
   * Add documentation

debian/patched-qemu.Dockerfile

@@ -1,4 +1,4 @@
-# Build patched QEMU for Debian Buster or Bullseye/Sid inside a container
+# Build patched QEMU for Debian inside a container
 # cd ..; podman build --build-arg REL=bullseye -v `pwd`/packages:/root/packages -f debian/patched-qemu.Dockerfile .
 
 ARG REL=
@@ -15,17 +15,19 @@ RUN if [ "$REL" = "buster" -o "$REL" = "bullseye" ]; then \
         echo 'Pin-Priority: 500' >> /etc/apt/preferences; \
     fi; \
     grep '^deb ' /etc/apt/sources.list | perl -pe 's/^deb/deb-src/' >> /etc/apt/sources.list; \
+    perl -i -pe 's/Types: deb$/Types: deb deb-src/' /etc/apt/sources.list.d/debian.sources || true; \
     echo 'APT::Install-Recommends false;' >> /etc/apt/apt.conf; \
     echo 'APT::Install-Suggests false;' >> /etc/apt/apt.conf
 
 RUN apt-get update
-RUN apt-get -y install qemu fio liburing1 liburing-dev libgoogle-perftools-dev devscripts
+RUN apt-get -y install fio liburing-dev libgoogle-perftools-dev devscripts
 RUN apt-get -y build-dep qemu
 # To build a custom version
 #RUN cp /root/packages/qemu-orig/* /root
 RUN apt-get --download-only source qemu
 
-ADD patches/qemu-5.0-vitastor.patch patches/qemu-5.1-vitastor.patch patches/qemu-6.1-vitastor.patch src/qemu_driver.c /root/vitastor/patches/
+ADD patches /root/vitastor/patches
+ADD src/qemu_driver.c /root/vitastor/src/qemu_driver.c
 RUN set -e; \
     apt-get install -y wget; \
     wget -q -O /etc/apt/trusted.gpg.d/vitastor.gpg https://vitastor.io/debian/pubkey.gpg; \
@@ -37,23 +39,14 @@ RUN set -e; \
     rm -rf /root/packages/qemu-$REL/*; \
     cd /root/packages/qemu-$REL; \
     dpkg-source -x /root/qemu*.dsc; \
-    if ls -d /root/packages/qemu-$REL/qemu-5.0*; then \
-        D=$(ls -d /root/packages/qemu-$REL/qemu-5.0*); \
-        cp /root/vitastor/patches/qemu-5.0-vitastor.patch $D/debian/patches; \
-        echo qemu-5.0-vitastor.patch >> $D/debian/patches/series; \
-    elif ls /root/packages/qemu-$REL/qemu-6.1*; then \
-        D=$(ls -d /root/packages/qemu-$REL/qemu-6.1*); \
-        cp /root/vitastor/patches/qemu-6.1-vitastor.patch $D/debian/patches; \
-        echo qemu-6.1-vitastor.patch >> $D/debian/patches/series; \
-    else \
-        cp /root/vitastor/patches/qemu-5.1-vitastor.patch /root/packages/qemu-$REL/qemu-*/debian/patches; \
-        P=`ls -d /root/packages/qemu-$REL/qemu-*/debian/patches`; \
-        echo qemu-5.1-vitastor.patch >> $P/series; \
-    fi; \
+    QEMU_VER=$(ls -d qemu*/ | perl -pe 's!^.*(\d+\.\d+).*!$1!'); \
+    D=$(ls -d qemu*/); \
+    cp /root/vitastor/patches/qemu-$QEMU_VER-vitastor.patch ./qemu-*/debian/patches; \
+    echo qemu-$QEMU_VER-vitastor.patch >> $D/debian/patches/series; \
     cd /root/packages/qemu-$REL/qemu-*/; \
     quilt push -a; \
     quilt add block/vitastor.c; \
-    cp /root/vitastor/patches/qemu_driver.c block/vitastor.c; \
+    cp /root/vitastor/src/qemu_driver.c block/vitastor.c; \
     quilt refresh; \
     V=$(head -n1 debian/changelog | perl -pe 's/^.*\((.*?)(~bpo[\d\+]*)?\).*$/$1/')+vitastor1; \
     DEBEMAIL="Vitaliy Filippov <vitalif@yourcmc.ru>" dch -D $REL -v $V 'Plug Vitastor block driver'; \

debian/vitastor.Dockerfile

@@ -1,4 +1,4 @@
-# Build Vitastor packages for Debian Buster or Bullseye/Sid inside a container
+# Build Vitastor packages for Debian inside a container
 # cd ..; podman build --build-arg REL=bullseye -v `pwd`/packages:/root/packages -f debian/vitastor.Dockerfile .
 
 ARG REL=
@@ -15,11 +15,12 @@ RUN if [ "$REL" = "buster" -o "$REL" = "bullseye" ]; then \
         echo 'Pin-Priority: 500' >> /etc/apt/preferences; \
     fi; \
     grep '^deb ' /etc/apt/sources.list | perl -pe 's/^deb/deb-src/' >> /etc/apt/sources.list; \
+    perl -i -pe 's/Types: deb$/Types: deb deb-src/' /etc/apt/sources.list.d/debian.sources || true; \
     echo 'APT::Install-Recommends false;' >> /etc/apt/apt.conf; \
     echo 'APT::Install-Suggests false;' >> /etc/apt/apt.conf
 
 RUN apt-get update
-RUN apt-get -y install fio liburing1 liburing-dev libgoogle-perftools-dev devscripts
+RUN apt-get -y install fio liburing-dev libgoogle-perftools-dev devscripts
 RUN apt-get -y build-dep fio
 RUN apt-get --download-only source fio
 RUN apt-get update && apt-get -y install libjerasure-dev cmake libibverbs-dev libisal-dev
@@ -34,8 +35,8 @@ RUN set -e -x; \
     mkdir -p /root/packages/vitastor-$REL; \
     rm -rf /root/packages/vitastor-$REL/*; \
     cd /root/packages/vitastor-$REL; \
-    cp -r /root/vitastor vitastor-0.8.9; \
-    cd vitastor-0.8.9; \
+    cp -r /root/vitastor vitastor-0.9.2; \
+    cd vitastor-0.9.2; \
     ln -s /root/fio-build/fio-*/ ./fio; \
     FIO=$(head -n1 fio/debian/changelog | perl -pe 's/^.*\((.*?)\).*$/$1/'); \
     ls /usr/include/linux/raw.h || cp ./debian/raw.h /usr/include/linux/raw.h; \
@@ -48,8 +49,8 @@ RUN set -e -x; \
     rm -rf a b; \
     echo "dep:fio=$FIO" > debian/fio_version; \
     cd /root/packages/vitastor-$REL; \
-    tar --sort=name --mtime='2020-01-01' --owner=0 --group=0 --exclude=debian -cJf vitastor_0.8.9.orig.tar.xz vitastor-0.8.9; \
-    cd vitastor-0.8.9; \
+    tar --sort=name --mtime='2020-01-01' --owner=0 --group=0 --exclude=debian -cJf vitastor_0.9.2.orig.tar.xz vitastor-0.9.2; \
+    cd vitastor-0.9.2; \
     V=$(head -n1 debian/changelog | perl -pe 's/^.*\((.*?)\).*$/$1/'); \
     DEBFULLNAME="Vitaliy Filippov <vitalif@yourcmc.ru>" dch -D $REL -v "$V""$REL" "Rebuild for $REL"; \
     DEB_BUILD_OPTIONS=nocheck dpkg-buildpackage --jobs=auto -sa; \

docs/config.en.md

@@ -21,7 +21,7 @@ Configuration parameters can be set in 3 places:
   mon, fio and QEMU options, OpenStack/Proxmox/etc configuration. The latter
   doesn't allow to set all variables directly, but it allows to override the
   configuration file and set everything you need inside it.
-- OSD superblocks created by [vitastor-disk](../usage/disk.en.md) contain
+- OSD superblocks created by [vitastor-disk](usage/disk.en.md) contain
   primarily disk layout parameters of specific OSDs. In fact, these parameters
   are automatically passed into the command line of vitastor-osd process, so
   they have the same "status" as command-line parameters.

docs/config.ru.md

@@ -23,7 +23,7 @@
   монитора, опциях fio и QEMU, настроек OpenStack, Proxmox и т.п. Последние,
   как правило, не включают полный набор параметров напрямую, но позволяют
   определить путь к файлу конфигурации и задать любые параметры в нём.
-- В суперблоке OSD, записываемом [vitastor-disk](../usage/disk.ru.md) - параметры,
+- В суперблоке OSD, записываемом [vitastor-disk](usage/disk.ru.md) - параметры,
   связанные с дисковым форматом и с этим конкретным OSD. На самом деле,
   при запуске OSD эти параметры автоматически передаются в командную строку
   процесса vitastor-osd, то есть по "статусу" они эквивалентны параметрам

docs/config/common.en.md

@@ -25,11 +25,16 @@ running if required parameters are specified.
 ## etcd_address
 
 - Type: string or array of strings
+- Can be changed online: yes
 
 etcd connection endpoint(s). Multiple endpoints may be delimited by "," or
 specified in a JSON array `["10.0.115.10:2379/v3","10.0.115.11:2379/v3"]`.
 Note that https is not supported for etcd connections yet.
 
+etcd connection endpoints can be changed online by updating global
+configuration in etcd itself - this allows to switch the cluster to new
+etcd addresses without downtime.
+
 ## etcd_prefix
 
 - Type: string
@@ -42,5 +47,6 @@ example, use a single etcd cluster for multiple Vitastor clusters.
 
 - Type: integer
 - Default: 0
+- Can be changed online: yes
 
 Log level. Raise if you want more verbose output.

docs/config/common.ru.md

@@ -24,10 +24,14 @@
 ## etcd_address
 
 - Тип: строка или массив строк
+- Можно менять на лету: да
 
 Адрес(а) подключения к etcd. Несколько адресов могут разделяться запятой
 или указываться в виде JSON-массива `["10.0.115.10:2379/v3","10.0.115.11:2379/v3"]`.
 
+Адреса подключения к etcd можно поменять на лету, обновив конфигурацию в
+самом etcd - это позволяет переключить кластер на новые etcd без остановки.
+
 ## etcd_prefix
 
 - Тип: строка
@@ -41,5 +45,6 @@
 
 - Тип: целое число
 - Значение по умолчанию: 0
+- Можно менять на лету: да
 
 Уровень логгирования. Повысьте, если хотите более подробный вывод.

docs/config/network.en.md

@@ -153,6 +153,7 @@ operations.
 - Type: seconds
 - Default: 5
 - Minimum: 1
+- Can be changed online: yes
 
 Interval before attempting to reconnect to an unavailable OSD.
 
@@ -161,6 +162,7 @@ Interval before attempting to reconnect to an unavailable OSD.
 - Type: seconds
 - Default: 5
 - Minimum: 1
+- Can be changed online: yes
 
 Timeout for OSD connection attempts.
 
@@ -169,6 +171,7 @@ Timeout for OSD connection attempts.
 - Type: seconds
 - Default: 5
 - Minimum: 1
+- Can be changed online: yes
 
 OSD connection inactivity time after which clients and other OSDs send
 keepalive requests to check state of the connection.
@@ -178,6 +181,7 @@ keepalive requests to check state of the connection.
 - Type: seconds
 - Default: 5
 - Minimum: 1
+- Can be changed online: yes
 
 Maximum time to wait for OSD keepalive responses. If an OSD doesn't respond
 within this time, the connection to it is dropped and a reconnection attempt
@@ -188,6 +192,7 @@ is scheduled.
 - Type: milliseconds
 - Default: 500
 - Minimum: 50
+- Can be changed online: yes
 
 OSDs respond to clients with a special error code when they receive I/O
 requests for a PG that's not synchronized and started. This parameter sets
@@ -197,6 +202,7 @@ the time for the clients to wait before re-attempting such I/O requests.
 
 - Type: integer
 - Default: 5
+- Can be changed online: yes
 
 Maximum number of attempts for etcd requests which can't be retried
 indefinitely.
@@ -205,6 +211,7 @@ indefinitely.
 
 - Type: milliseconds
 - Default: 1000
+- Can be changed online: yes
 
 Timeout for etcd requests which should complete quickly, like lease refresh.
 
@@ -212,6 +219,7 @@ Timeout for etcd requests which should complete quickly, like lease refresh.
 
 - Type: milliseconds
 - Default: 5000
+- Can be changed online: yes
 
 Timeout for etcd requests which are allowed to wait for some time.
 
@@ -219,6 +227,7 @@ Timeout for etcd requests which are allowed to wait for some time.
 
 - Type: seconds
 - Default: max(30, etcd_report_interval*2)
+- Can be changed online: yes
 
 Timeout for etcd connection HTTP Keep-Alive. Should be higher than
 etcd_report_interval to guarantee that keepalive actually works.
@@ -227,6 +236,7 @@ etcd_report_interval to guarantee that keepalive actually works.
 
 - Type: seconds
 - Default: 30
+- Can be changed online: yes
 
 etcd websocket ping interval required to keep the connection alive and
 detect disconnections quickly.
@@ -235,6 +245,7 @@ detect disconnections quickly.
 
 - Type: integer
 - Default: 33554432
+- Can be changed online: yes
 
 Without immediate_commit=all this parameter sets the limit of "dirty"
 (not committed by fsync) data allowed by the client before forcing an

docs/config/network.ru.md

@@ -161,6 +161,7 @@ OSD в любом случае согласовывают реальное зн
 - Тип: секунды
 - Значение по умолчанию: 5
 - Минимальное значение: 1
+- Можно менять на лету: да
 
 Время ожидания перед повторной попыткой соединиться с недоступным OSD.
 
@@ -169,6 +170,7 @@ OSD в любом случае согласовывают реальное зн
 - Тип: секунды
 - Значение по умолчанию: 5
 - Минимальное значение: 1
+- Можно менять на лету: да
 
 Максимальное время ожидания попытки соединения с OSD.
 
@@ -177,6 +179,7 @@ OSD в любом случае согласовывают реальное зн
 - Тип: секунды
 - Значение по умолчанию: 5
 - Минимальное значение: 1
+- Можно менять на лету: да
 
 Время неактивности соединения с OSD, после которого клиенты или другие OSD
 посылают запрос проверки состояния соединения.
@@ -186,6 +189,7 @@ OSD в любом случае согласовывают реальное зн
 - Тип: секунды
 - Значение по умолчанию: 5
 - Минимальное значение: 1
+- Можно менять на лету: да
 
 Максимальное время ожидания ответа на запрос проверки состояния соединения.
 Если OSD не отвечает за это время, соединение отключается и производится
@@ -196,6 +200,7 @@ OSD в любом случае согласовывают реальное зн
 - Тип: миллисекунды
 - Значение по умолчанию: 500
 - Минимальное значение: 50
+- Можно менять на лету: да
 
 Когда OSD получают от клиентов запросы ввода-вывода, относящиеся к не
 поднятым на данный момент на них PG, либо к PG в процессе синхронизации,
@@ -207,6 +212,7 @@ OSD в любом случае согласовывают реальное зн
 
 - Тип: целое число
 - Значение по умолчанию: 5
+- Можно менять на лету: да
 
 Максимальное число попыток выполнения запросов к etcd для тех запросов,
 которые нельзя повторять бесконечно.
@@ -215,6 +221,7 @@ OSD в любом случае согласовывают реальное зн
 
 - Тип: миллисекунды
 - Значение по умолчанию: 1000
+- Можно менять на лету: да
 
 Максимальное время выполнения запросов к etcd, которые должны завершаться
 быстро, таких, как обновление резервации (lease).
@@ -223,6 +230,7 @@ OSD в любом случае согласовывают реальное зн
 
 - Тип: миллисекунды
 - Значение по умолчанию: 5000
+- Можно менять на лету: да
 
 Максимальное время выполнения запросов к etcd, для которых не обязательно
 гарантировать быстрое выполнение.
@@ -231,6 +239,7 @@ OSD в любом случае согласовывают реальное зн
 
 - Тип: секунды
 - Значение по умолчанию: max(30, etcd_report_interval*2)
+- Можно менять на лету: да
 
 Таймаут для HTTP Keep-Alive в соединениях к etcd. Должен быть больше, чем
 etcd_report_interval, чтобы keepalive гарантированно работал.
@@ -239,6 +248,7 @@ etcd_report_interval, чтобы keepalive гарантированно рабо
 
 - Тип: секунды
 - Значение по умолчанию: 30
+- Можно менять на лету: да
 
 Интервал проверки живости вебсокет-подключений к etcd.
 
@@ -246,6 +256,7 @@ etcd_report_interval, чтобы keepalive гарантированно рабо
 
 - Тип: целое число
 - Значение по умолчанию: 33554432
+- Можно менять на лету: да
 
 При работе без immediate_commit=all - это лимит объёма "грязных" (не
 зафиксированных fsync-ом) данных, при достижении которого клиент будет

docs/config/osd.en.md

@@ -7,7 +7,8 @@
 # Runtime OSD Parameters
 
 These parameters only apply to OSDs, are not fixed at the moment of OSD drive
-initialization and can be changed with an OSD restart.
+initialization and can be changed - either with an OSD restart or, for some of
+them, even without restarting by updating configuration in etcd.
 
 - [etcd_report_interval](#etcd_report_interval)
 - [run_primary](#run_primary)
@@ -38,6 +39,14 @@ initialization and can be changed with an OSD restart.
 - [throttle_target_parallelism](#throttle_target_parallelism)
 - [throttle_threshold_us](#throttle_threshold_us)
 - [osd_memlock](#osd_memlock)
+- [auto_scrub](#auto_scrub)
+- [no_scrub](#no_scrub)
+- [scrub_interval](#scrub_interval)
+- [scrub_queue_depth](#scrub_queue_depth)
+- [scrub_sleep](#scrub_sleep)
+- [scrub_list_limit](#scrub_list_limit)
+- [scrub_find_best](#scrub_find_best)
+- [scrub_ec_max_bruteforce](#scrub_ec_max_bruteforce)
 
 ## etcd_report_interval
 
@@ -91,6 +100,7 @@ OSD by hand.
 
 - Type: seconds
 - Default: 5
+- Can be changed online: yes
 
 Time interval at which automatic fsyncs/flushes are issued by each OSD when
 the immediate_commit mode if disabled. fsyncs are required because without
@@ -103,6 +113,7 @@ issue fsyncs at all.
 
 - Type: integer
 - Default: 128
+- Can be changed online: yes
 
 Same as autosync_interval, but sets the maximum number of uncommitted write
 operations before issuing an fsync operation internally.
@@ -111,6 +122,7 @@ operations before issuing an fsync operation internally.
 
 - Type: integer
 - Default: 4
+- Can be changed online: yes
 
 Maximum recovery operations per one primary OSD at any given moment of time.
 Currently it's the only parameter available to tune the speed or recovery
@@ -120,6 +132,7 @@ and rebalancing, but it's planned to implement more.
 
 - Type: integer
 - Default: 128
+- Can be changed online: yes
 
 Number of recovery operations before switching to recovery of the next PG.
 The idea is to mix all PGs during recovery for more even space and load
@@ -130,6 +143,7 @@ Degraded PGs are anyway scanned first.
 
 - Type: integer
 - Default: 16
+- Can be changed online: yes
 
 Maximum number of recovery operations before issuing an additional fsync.
 
@@ -145,6 +159,7 @@ the underlying device. This may be useful for recovery purposes.
 
 - Type: boolean
 - Default: false
+- Can be changed online: yes
 
 Disable automatic background recovery of objects. Note that it doesn't
 affect implicit recovery of objects happening during writes - a write is
@@ -154,6 +169,7 @@ always made to a full set of at least pg_minsize OSDs.
 
 - Type: boolean
 - Default: false
+- Can be changed online: yes
 
 Disable background movement of data between different OSDs. Disabling it
 means that PGs in the `has_misplaced` state will be left in it indefinitely.
@@ -162,6 +178,7 @@ means that PGs in the `has_misplaced` state will be left in it indefinitely.
 
 - Type: seconds
 - Default: 3
+- Can be changed online: yes
 
 Time interval at which OSDs print simple human-readable operation
 statistics on stdout.
@@ -170,6 +187,7 @@ statistics on stdout.
 
 - Type: seconds
 - Default: 10
+- Can be changed online: yes
 
 Time interval at which OSDs dump slow or stuck operations on stdout, if
 they're any. Also it's the time after which an operation is considered
@@ -179,6 +197,7 @@ they're any. Also it's the time after which an operation is considered
 
 - Type: seconds
 - Default: 60
+- Can be changed online: yes
 
 Number of seconds after which a deleted inode is removed from OSD statistics.
 
@@ -186,6 +205,7 @@ Number of seconds after which a deleted inode is removed from OSD statistics.
 
 - Type: integer
 - Default: 128
+- Can be changed online: yes
 
 Parallel client write operation limit per one OSD. Operations that exceed
 this limit are pushed to a temporary queue instead of being executed
@@ -195,6 +215,7 @@ immediately.
 
 - Type: integer
 - Default: 1
+- Can be changed online: yes
 
 Flusher is a micro-thread that moves data from the journal to the data
 area of the device. Their number is auto-tuned between minimum and maximum.
@@ -204,6 +225,7 @@ Minimum number is set by this parameter.
 
 - Type: integer
 - Default: 256
+- Can be changed online: yes
 
 Maximum number of journal flushers (see above min_flusher_count).
 
@@ -260,6 +282,7 @@ Most (99%) other SSDs don't need this option.
 
 - Type: boolean
 - Default: false
+- Can be changed online: yes
 
 Enable soft throttling of small journaled writes. Useful for hybrid OSDs
 with fast journal/metadata devices and slow data devices. The idea is that
@@ -277,6 +300,7 @@ fills up.
 
 - Type: integer
 - Default: 100
+- Can be changed online: yes
 
 Target maximum number of throttled operations per second under the condition
 of full journal. Set it to approximate random write iops of your data devices
@@ -286,6 +310,7 @@ of full journal. Set it to approximate random write iops of your data devices
 
 - Type: integer
 - Default: 100
+- Can be changed online: yes
 
 Target maximum bandwidth in MB/s of throttled operations per second under
 the condition of full journal. Set it to approximate linear write
@@ -295,6 +320,7 @@ performance of your data devices (HDDs).
 
 - Type: integer
 - Default: 1
+- Can be changed online: yes
 
 Target maximum parallelism of throttled operations under the condition of
 full journal. Set it to approximate internal parallelism of your data
@@ -304,6 +330,7 @@ devices (1 for HDDs, 4-8 for SSDs).
 
 - Type: microseconds
 - Default: 50
+- Can be changed online: yes
 
 Minimal computed delay to be applied to throttled operations. Usually
 doesn't need to be changed.
@@ -313,4 +340,103 @@ doesn't need to be changed.
 - Type: boolean
 - Default: false
 
-Lock all OSD memory to prevent it from being unloaded into swap with mlockall(). Requires sufficient ulimit -l (max locked memory).
+Lock all OSD memory to prevent it from being unloaded into swap with
+mlockall(). Requires sufficient ulimit -l (max locked memory).
+
+## auto_scrub
+
+- Type: boolean
+- Default: false
+- Can be changed online: yes
+
+Data scrubbing is the process of background verification of copies to find
+and repair corrupted blocks. It's not run automatically by default since
+it's a new feature. Set this parameter to true to enable automatic scrubs.
+
+This parameter makes OSDs automatically schedule data scrubbing of clean PGs
+every `scrub_interval` (see below). You can also start/schedule scrubbing
+manually by setting `next_scrub` JSON key to the desired UNIX time of the
+next scrub in `/pg/history/...` values in etcd.
+
+## no_scrub
+
+- Type: boolean
+- Default: false
+- Can be changed online: yes
+
+Temporarily disable scrubbing and stop running scrubs.
+
+## scrub_interval
+
+- Type: string
+- Default: 30d
+- Can be changed online: yes
+
+Default automatic scrubbing interval for all pools. Numbers without suffix
+are treated as seconds, possible unit suffixes include 's' (seconds),
+'m' (minutes), 'h' (hours), 'd' (days), 'M' (months) and 'y' (years).
+
+## scrub_queue_depth
+
+- Type: integer
+- Default: 1
+- Can be changed online: yes
+
+Number of parallel scrubbing operations per one OSD.
+
+## scrub_sleep
+
+- Type: milliseconds
+- Default: 0
+- Can be changed online: yes
+
+Additional interval between two consecutive scrubbing operations on one OSD.
+Can be used to slow down scrubbing if it affects user load too much.
+
+## scrub_list_limit
+
+- Type: integer
+- Default: 1000
+- Can be changed online: yes
+
+Number of objects to list in one listing operation during scrub.
+
+## scrub_find_best
+
+- Type: boolean
+- Default: true
+- Can be changed online: yes
+
+Find and automatically restore best versions of objects with unmatched
+copies. In replicated setups, the best version is the version with most
+matching replicas. In EC setups, the best version is the subset of data
+and parity chunks without mismatches.
+
+The hypothetical situation where you might want to disable it is when
+you have 3 replicas and you are paranoid that 2 HDDs out of 3 may silently
+corrupt an object in the same way (for example, zero it out) and only
+1 HDD will remain good. In this case disabling scrub_find_best may help
+you to recover the data! See also scrub_ec_max_bruteforce below.
+
+## scrub_ec_max_bruteforce
+
+- Type: integer
+- Default: 100
+- Can be changed online: yes
+
+Vitastor can locate corrupted chunks in EC setups with more than 1 parity
+chunk by brute-forcing all possible error locations. This configuration
+value limits the maximum number of checked combinations. You can try to
+increase it if you have EC N+K setup with N and K large enough for
+combination count `C(N+K-1, K-1) = (N+K-1)! / (K-1)! / N!` to be greater
+than the default 100.
+
+If there are too many possible combinations or if multiple combinations give
+correct results then objects are marked inconsistent and aren't recovered
+automatically.
+
+In replicated setups bruteforcing isn't needed, Vitastor just assumes that
+the variant with most available equal copies is correct. For example, if
+you have 3 replicas and 1 of them differs, this one is considered to be
+corrupted. But if there is no "best" version with more copies than all
+others have then the object is also marked as inconsistent.

docs/config/osd.ru.md

@@ -8,7 +8,8 @@
 
 Данные параметры используются только OSD, но, в отличие от дисковых параметров,
 не фиксируются в момент инициализации дисков OSD и могут быть изменены в любой
-момент с перезапуском OSD.
+момент с помощью перезапуска OSD, а некоторые и без перезапуска, с помощью
+изменения конфигурации в etcd.
 
 - [etcd_report_interval](#etcd_report_interval)
 - [run_primary](#run_primary)
@@ -39,6 +40,14 @@
 - [throttle_target_parallelism](#throttle_target_parallelism)
 - [throttle_threshold_us](#throttle_threshold_us)
 - [osd_memlock](#osd_memlock)
+- [auto_scrub](#auto_scrub)
+- [no_scrub](#no_scrub)
+- [scrub_interval](#scrub_interval)
+- [scrub_queue_depth](#scrub_queue_depth)
+- [scrub_sleep](#scrub_sleep)
+- [scrub_list_limit](#scrub_list_limit)
+- [scrub_find_best](#scrub_find_best)
+- [scrub_ec_max_bruteforce](#scrub_ec_max_bruteforce)
 
 ## etcd_report_interval
 
@@ -93,6 +102,7 @@ RUNNING), подходящий под заданную маску. Также н
 
 - Тип: секунды
 - Значение по умолчанию: 5
+- Можно менять на лету: да
 
 Временной интервал отправки автоматических fsync-ов (операций очистки кэша)
 каждым OSD для случая, когда режим immediate_commit отключён. fsync-и нужны
@@ -105,6 +115,7 @@ OSD, чтобы успевать очищать журнал - без них OSD
 
 - Тип: целое число
 - Значение по умолчанию: 128
+- Можно менять на лету: да
 
 Аналогично autosync_interval, но задаёт не временной интервал, а
 максимальное количество незафиксированных операций записи перед
@@ -114,6 +125,7 @@ OSD, чтобы успевать очищать журнал - без них OSD
 
 - Тип: целое число
 - Значение по умолчанию: 4
+- Можно менять на лету: да
 
 Максимальное число операций восстановления на одном первичном OSD в любой
 момент времени. На данный момент единственный параметр, который можно менять
@@ -124,6 +136,7 @@ OSD, чтобы успевать очищать журнал - без них OSD
 
 - Тип: целое число
 - Значение по умолчанию: 128
+- Можно менять на лету: да
 
 Число операций восстановления перед переключением на восстановление другой PG.
 Идея заключается в том, чтобы восстанавливать все PG одновременно для более
@@ -135,6 +148,7 @@ OSD, чтобы успевать очищать журнал - без них OSD
 
 - Тип: целое число
 - Значение по умолчанию: 16
+- Можно менять на лету: да
 
 Максимальное число операций восстановления перед дополнительным fsync.
 
@@ -150,6 +164,7 @@ OSD, чтобы успевать очищать журнал - без них OSD
 
 - Тип: булево (да/нет)
 - Значение по умолчанию: false
+- Можно менять на лету: да
 
 Отключить автоматическое фоновое восстановление объектов. Обратите внимание,
 что эта опция не отключает восстановление объектов, происходящее при
@@ -160,6 +175,7 @@ OSD.
 
 - Тип: булево (да/нет)
 - Значение по умолчанию: false
+- Можно менять на лету: да
 
 Отключить фоновое перемещение объектов между разными OSD. Отключение
 означает, что PG, находящиеся в состоянии `has_misplaced`, будут оставлены
@@ -169,6 +185,7 @@ OSD.
 
 - Тип: секунды
 - Значение по умолчанию: 3
+- Можно менять на лету: да
 
 Временной интервал, с которым OSD печатают простую человекочитаемую
 статистику выполнения операций в стандартный вывод.
@@ -177,6 +194,7 @@ OSD.
 
 - Тип: секунды
 - Значение по умолчанию: 10
+- Можно менять на лету: да
 
 Временной интервал, с которым OSD выводят в стандартный вывод список
 медленных или зависших операций, если таковые имеются. Также время, при
@@ -186,6 +204,7 @@ OSD.
 
 - Тип: секунды
 - Значение по умолчанию: 60
+- Можно менять на лету: да
 
 Число секунд, через которое удалённые инод удаляется и из статистики OSD.
 
@@ -193,6 +212,7 @@ OSD.
 
 - Тип: целое число
 - Значение по умолчанию: 128
+- Можно менять на лету: да
 
 Максимальное число одновременных клиентских операций записи на один OSD.
 Операции, превышающие этот лимит, не исполняются сразу, а сохраняются во
@@ -202,6 +222,7 @@ OSD.
 
 - Тип: целое число
 - Значение по умолчанию: 1
+- Можно менять на лету: да
 
 Flusher - это микро-поток (корутина), которая копирует данные из журнала в
 основную область устройства данных. Их число настраивается динамически между
@@ -211,6 +232,7 @@ Flusher - это микро-поток (корутина), которая коп
 
 - Тип: целое число
 - Значение по умолчанию: 256
+- Можно менять на лету: да
 
 Максимальное число микро-потоков очистки журнала (см. выше min_flusher_count).
 
@@ -270,6 +292,7 @@ Flusher - это микро-поток (корутина), которая коп
 
 - Тип: булево (да/нет)
 - Значение по умолчанию: false
+- Можно менять на лету: да
 
 Разрешить мягкое ограничение скорости журналируемой записи. Полезно для
 гибридных OSD с быстрыми устройствами метаданных и медленными устройствами
@@ -288,6 +311,7 @@ Flusher - это микро-поток (корутина), которая коп
 
 - Тип: целое число
 - Значение по умолчанию: 100
+- Можно менять на лету: да
 
 Расчётное максимальное число ограничиваемых операций в секунду при условии
 отсутствия свободного места в журнале. Устанавливайте приблизительно равным
@@ -298,6 +322,7 @@ Flusher - это микро-поток (корутина), которая коп
 
 - Тип: целое число
 - Значение по умолчанию: 100
+- Можно менять на лету: да
 
 Расчётный максимальный размер в МБ/с ограничиваемых операций в секунду при
 условии отсутствия свободного места в журнале. Устанавливайте приблизительно
@@ -308,6 +333,7 @@ Flusher - это микро-поток (корутина), которая коп
 
 - Тип: целое число
 - Значение по умолчанию: 1
+- Можно менять на лету: да
 
 Расчётный максимальный параллелизм ограничиваемых операций в секунду при
 условии отсутствия свободного места в журнале. Устанавливайте приблизительно
@@ -318,6 +344,7 @@ Flusher - это микро-поток (корутина), которая коп
 
 - Тип: микросекунды
 - Значение по умолчанию: 50
+- Можно менять на лету: да
 
 Минимальная применимая к ограничиваемым операциям задержка. Обычно не
 требует изменений.
@@ -327,4 +354,113 @@ Flusher - это микро-поток (корутина), которая коп
 - Тип: булево (да/нет)
 - Значение по умолчанию: false
 
-Блокировать всю память OSD с помощью mlockall, чтобы запретить её выгрузку в пространство подкачки. Требует достаточного значения ulimit -l (лимита заблокированной памяти).
+Блокировать всю память OSD с помощью mlockall, чтобы запретить её выгрузку
+в пространство подкачки. Требует достаточного значения ulimit -l (лимита
+заблокированной памяти).
+
+## auto_scrub
+
+- Тип: булево (да/нет)
+- Значение по умолчанию: false
+- Можно менять на лету: да
+
+Скраб - процесс фоновой проверки копий данных, предназначенный, чтобы
+находить и исправлять повреждённые блоки. По умолчанию эти проверки ещё не
+запускаются автоматически, так как являются новой функцией. Чтобы включить
+автоматическое планирование скрабов, установите данный параметр в true.
+
+Включённый параметр заставляет OSD автоматически планировать фоновую
+проверку чистых PG раз в `scrub_interval` (см. ниже). Вы также можете
+запустить или запланировать проверку вручную, установив значение ключа JSON
+`next_scrub` внутри ключей etcd `/pg/history/...` в UNIX-время следующей
+желаемой проверки.
+
+## no_scrub
+
+- Тип: булево (да/нет)
+- Значение по умолчанию: false
+- Можно менять на лету: да
+
+Временно отключить и остановить запущенные скрабы.
+
+## scrub_interval
+
+- Тип: строка
+- Значение по умолчанию: 30d
+- Можно менять на лету: да
+
+Интервал автоматической фоновой проверки по умолчанию для всех пулов.
+Значения без указанной единицы измерения считаются в секундах, допустимые
+символы единиц измерения в конце: 's' (секунды),
+'m' (минуты), 'h' (часы), 'd' (дни), 'M' (месяца) или 'y' (годы).
+
+## scrub_queue_depth
+
+- Тип: целое число
+- Значение по умолчанию: 1
+- Можно менять на лету: да
+
+Число параллельных операций фоновой проверки на один OSD.
+
+## scrub_sleep
+
+- Тип: миллисекунды
+- Значение по умолчанию: 0
+- Можно менять на лету: да
+
+Дополнительный интервал ожидания после фоновой проверки каждого объекта на
+одном OSD. Может использоваться для замедления скраба, если он слишком
+сильно влияет на пользовательскую нагрузку.
+
+## scrub_list_limit
+
+- Тип: целое число
+- Значение по умолчанию: 1000
+- Можно менять на лету: да
+
+Размер загружаемых за одну операцию списков объектов в процессе фоновой
+проверки.
+
+## scrub_find_best
+
+- Тип: булево (да/нет)
+- Значение по умолчанию: true
+- Можно менять на лету: да
+
+Находить и автоматически восстанавливать "лучшие версии" объектов с
+несовпадающими копиями/частями. При использовании репликации "лучшая"
+версия - версия, доступная в большем числе экземпляров, чем другие. При
+использовании кодов коррекции ошибок "лучшая" версия - это подмножество
+частей данных и чётности, полностью соответствующих друг другу.
+
+Гипотетическая ситуация, в которой вы можете захотеть отключить этот
+поиск - это если у вас 3 реплики и вы боитесь, что 2 диска из 3 могут
+незаметно и одинаково повредить данные одного и того же объекта, например,
+занулив его, и только 1 диск останется неповреждённым. В этой ситуации
+отключение этого параметра поможет вам восстановить данные! Смотрите также
+описание следующего параметра - scrub_ec_max_bruteforce.
+
+## scrub_ec_max_bruteforce
+
+- Тип: целое число
+- Значение по умолчанию: 100
+- Можно менять на лету: да
+
+Vitastor старается определить повреждённые части объектов при использовании
+EC (кодов коррекции ошибок) с более, чем 1 диском чётности, путём перебора
+всех возможных комбинаций ошибочных частей. Данное значение конфигурации
+ограничивает число перебираемых комбинаций. Вы можете попробовать поднять
+его, если используете схему кодирования EC N+K с N и K, достаточно большими
+для того, чтобы число сочетаний `C(N+K-1, K-1) = (N+K-1)! / (K-1)! / N!`
+было больше, чем стандартное значение 100.
+
+Если возможных комбинаций слишком много или если корректная комбинаций не
+определяется однозначно, объекты помечаются неконсистентными (inconsistent)
+и не восстанавливаются автоматически.
+
+При использовании репликации перебор не нужен, Vitastor просто предполагает,
+что вариант объекта с наибольшим количеством одинаковых копий корректен.
+Например, если вы используете 3 реплики и 1 из них отличается, эта 1 копия
+считается некорректной. Однако, если "лучшую" версию с числом доступных
+копий большим, чем у всех других версий, найти невозможно, то объект тоже
+маркируется неконсистентным.

docs/config/pool.en.md

@@ -40,6 +40,7 @@ Parameters:
 - [root_node](#root_node)
 - [osd_tags](#osd_tags)
 - [primary_affinity_tags](#primary_affinity_tags)
+- [scrub_interval](#scrub_interval)
 
 Examples:
 
@@ -272,6 +273,13 @@ Specifies OSD tags to prefer putting primary OSDs in this pool to.
 Note that for EC/XOR pools Vitastor always prefers to put primary OSD on one
 of the OSDs containing a data chunk for a PG.
 
+## scrub_interval
+
+- Type: time interval (number + unit s/m/h/d/M/y)
+
+Automatic scrubbing interval for this pool. Overrides
+[global scrub_interval setting](osd.en.md#scrub_interval).
+
 # Examples
 
 ## Replicated pool

docs/config/pool.ru.md

@@ -39,6 +39,7 @@
 - [root_node](#root_node)
 - [osd_tags](#osd_tags)
 - [primary_affinity_tags](#primary_affinity_tags)
+- [scrub_interval](#scrub_interval)
 
 Примеры:
 
@@ -276,6 +277,13 @@ PG в Vitastor эферемерны, то есть вы можете менят
 для PG этого пула. Имейте в виду, что для EC-пулов Vitastor также всегда
 предпочитает помещать первичный OSD на один из OSD с данными, а не с чётностью.
 
+## scrub_interval
+
+- Тип: временной интервал (число + единица измерения s/m/h/d/M/y)
+
+Интервал скраба, то есть, автоматической фоновой проверки данных для данного пула.
+Переопределяет [глобальную настройку scrub_interval](osd.ru.md#scrub_interval).
+
 # Примеры
 
 ## Реплицированный пул

docs/config/src/common.yml

@@ -11,13 +11,21 @@
 - name: etcd_address
   type: string or array of strings
   type_ru: строка или массив строк
+  online: true
   info: |
     etcd connection endpoint(s). Multiple endpoints may be delimited by "," or
     specified in a JSON array `["10.0.115.10:2379/v3","10.0.115.11:2379/v3"]`.
     Note that https is not supported for etcd connections yet.
+
+    etcd connection endpoints can be changed online by updating global
+    configuration in etcd itself - this allows to switch the cluster to new
+    etcd addresses without downtime.
   info_ru: |
     Адрес(а) подключения к etcd. Несколько адресов могут разделяться запятой
     или указываться в виде JSON-массива `["10.0.115.10:2379/v3","10.0.115.11:2379/v3"]`.
+
+    Адреса подключения к etcd можно поменять на лету, обновив конфигурацию в
+    самом etcd - это позволяет переключить кластер на новые etcd без остановки.
 - name: etcd_prefix
   type: string
   default: "/vitastor"
@@ -31,5 +39,6 @@
 - name: log_level
   type: int
   default: 0
+  online: true
   info: Log level. Raise if you want more verbose output.
   info_ru: Уровень логгирования. Повысьте, если хотите более подробный вывод.

docs/config/src/include.js

@@ -0,0 +1,145 @@
+#!/usr/bin/nodejs
+
+const fsp = require('fs').promises;
+
+run(process.argv).catch(console.error);
+
+async function run(argv)
+{
+    if (argv.length < 3)
+    {
+        console.log('Markdown preprocessor\nUSAGE: ./include.js file.md');
+        return;
+    }
+    const index_file = await fsp.realpath(argv[2]);
+    const re = /(\{\{[\s\S]*?\}\}|\[[^\]]+\]\([^\)]+\)|(?:^|\n)#[^\n]+)/;
+    let text = await fsp.readFile(index_file, { encoding: 'utf-8' });
+    text = text.split(re);
+    let included = {};
+    let heading = 0, heading_name = '', m;
+    for (let i = 0; i < text.length; i++)
+    {
+        if (text[i].substr(0, 2) == '{{')
+        {
+            // Inclusion
+            let incfile = text[i].substr(2, text[i].length-4);
+            let section = null;
+            let indent = heading;
+            incfile = incfile.replace(/\s*\|\s*indent\s*=\s*(-?\d+)\s*$/, (m, m1) => { indent = parseInt(m1); return ''; });
+            incfile = incfile.replace(/\s*#\s*([^#]+)$/, (m, m1) => { section = m1; return ''; });
+            let inc_heading = section;
+            incfile = rel2abs(index_file, incfile);
+            let inc = await fsp.readFile(incfile, { encoding: 'utf-8' });
+            inc = inc.trim().replace(/^[\s\S]+?\n#/, '#'); // remove until the first header
+            inc = inc.split(re);
+            const indent_str = new Array(indent+1).join('#');
+            let section_start = -1, section_end = -1;
+            for (let j = 0; j < inc.length; j++)
+            {
+                if ((m = /^(\n?)(#+\s*)([\s\S]+)$/.exec(inc[j])))
+                {
+                    if (!inc_heading)
+                    {
+                        inc_heading = m[3].trim();
+                    }
+                    if (section)
+                    {
+                        if (m[3].trim() == section)
+                            section_start = j;
+                        else if (section_start >= 0)
+                        {
+                            section_end = j;
+                            break;
+                        }
+                    }
+                    inc[j] = m[1] + indent_str + m[2] + m[3];
+                }
+                else if ((m = /^(\[[^\]]+\]\()([^\)]+)(\))$/.exec(inc[j])) && !/^https?:(\/\/)|^#/.exec(m[2]))
+                {
+                    const abs_m2 = rel2abs(incfile, m[2]);
+                    const rel_m = abs2rel(__filename, abs_m2);
+                    if (rel_m.substr(0, 9) == '../../../') // outside docs
+                        inc[j] = m[1] + 'https://git.yourcmc.ru/vitalif/vitastor/src/branch/master/'+rel2abs('docs/config/src/include.js', rel_m) + m[3];
+                    else
+                        inc[j] = m[1] + abs_m2 + m[3];
+                }
+            }
+            if (section)
+            {
+                inc = section_start >= 0 ? inc.slice(section_start, section_end < 0 ? inc.length : section_end) : [];
+            }
+            if (inc.length)
+            {
+                if (!inc_heading)
+                    inc_heading = heading_name||'';
+                included[incfile+(section ? '#'+section : '')] = '#'+inc_heading.toLowerCase().replace(/\P{L}+/ug, '-').replace(/^-|-$/g, '');
+                inc[0] = inc[0].replace(/^\s+/, '');
+                inc[inc.length-1] = inc[inc.length-1].replace(/\s+$/, '');
+            }
+            text.splice(i, 1, ...inc);
+            i = i + inc.length - 1;
+        }
+        else if ((m = /^\n?(#+)\s*([\s\S]+)$/.exec(text[i])))
+        {
+            // Heading
+            heading = m[1].length;
+            heading_name = m[2].trim();
+        }
+    }
+    for (let i = 0; i < text.length; i++)
+    {
+        if ((m = /^(\[[^\]]+\]\()([^\)]+)(\))$/.exec(text[i])) && !/^https?:(\/\/)|^#/.exec(m[2]))
+        {
+            const p = m[2].indexOf('#');
+            if (included[m[2]])
+            {
+                text[i] = m[1]+included[m[2]]+m[3];
+            }
+            else if (p >= 0 && included[m[2].substr(0, p)])
+            {
+                text[i] = m[1]+m[2].substr(p)+m[3];
+            }
+        }
+    }
+    console.log(text.join(''));
+}
+
+function rel2abs(ref, rel)
+{
+    rel = [ ...ref.replace(/^(.*)\/[^\/]+$/, '$1').split(/\/+/), ...rel.split(/\/+/) ];
+    return killdots(rel).join('/');
+}
+
+function abs2rel(ref, abs)
+{
+    ref = ref.split(/\/+/);
+    abs = abs.split(/\/+/);
+    while (ref.length > 1 && ref[0] == abs[0])
+    {
+        ref.shift();
+        abs.shift();
+    }
+    for (let i = 1; i < ref.length; i++)
+    {
+        abs.unshift('..');
+    }
+    return killdots(abs).join('/');
+}
+
+function killdots(rel)
+{
+    for (let i = 0; i < rel.length; i++)
+    {
+        if (rel[i] == '.')
+        {
+            rel.splice(i, 1);
+            i--;
+        }
+        else if (i >= 1 && rel[i] == '..' && rel[i-1] != '..')
+        {
+            rel.splice(i-1, 2);
+            i -= 2;
+        }
+    }
+    return rel;
+}

docs/config/src/included.en.md

@@ -0,0 +1,65 @@
+# Vitastor
+
+{{../../../README.md#The Idea}}
+
+{{../../../README.md#Talks and presentations}}
+
+{{../../intro/features.en.md}}
+
+{{../../intro/quickstart.en.md}}
+
+{{../../intro/architecture.en.md}}
+
+## Installation
+
+{{../../installation/packages.en.md}}
+
+{{../../installation/proxmox.en.md}}
+
+{{../../installation/openstack.en.md}}
+
+{{../../installation/kubernetes.en.md}}
+
+{{../../installation/source.en.md}}
+
+{{../../config.en.md|indent=1}}
+
+{{../../config/common.en.md|indent=2}}
+
+{{../../config/network.en.md|indent=2}}
+
+{{../../config/layout-cluster.en.md|indent=2}}
+
+{{../../config/layout-osd.en.md|indent=2}}
+
+{{../../config/osd.en.md|indent=2}}
+
+{{../../config/monitor.en.md|indent=2}}
+
+{{../../config/pool.en.md|indent=2}}
+
+{{../../config/inode.en.md|indent=2}}
+
+## Usage
+
+{{../../usage/cli.en.md}}
+
+{{../../usage/disk.en.md}}
+
+{{../../usage/fio.en.md}}
+
+{{../../usage/nbd.en.md}}
+
+{{../../usage/qemu.en.md}}
+
+{{../../usage/nfs.en.md}}
+
+## Performance
+
+{{../../performance/understanding.en.md}}
+
+{{../../performance/theoretical.en.md}}
+
+{{../../performance/comparison1.en.md}}
+
+{{../../intro/author.en.md|indent=1}}

docs/config/src/included.ru.md

@@ -0,0 +1,65 @@
+# Vitastor
+
+{{../../../README-ru.md#Идея|indent=0}}
+
+{{../../../README-ru.md#Презентации и записи докладов|indent=0}}
+
+{{../../intro/features.ru.md}}
+
+{{../../intro/quickstart.ru.md}}
+
+{{../../intro/architecture.ru.md}}
+
+## Установка
+
+{{../../installation/packages.ru.md}}
+
+{{../../installation/proxmox.ru.md}}
+
+{{../../installation/openstack.ru.md}}
+
+{{../../installation/kubernetes.ru.md}}
+
+{{../../installation/source.ru.md}}
+
+{{../../config.ru.md|indent=1}}
+
+{{../../config/common.ru.md|indent=2}}
+
+{{../../config/network.ru.md|indent=2}}
+
+{{../../config/layout-cluster.ru.md|indent=2}}
+
+{{../../config/layout-osd.ru.md|indent=2}}
+
+{{../../config/osd.ru.md|indent=2}}
+
+{{../../config/monitor.ru.md|indent=2}}
+
+{{../../config/pool.ru.md|indent=2}}
+
+{{../../config/inode.ru.md|indent=2}}
+
+## Использование
+
+{{../../usage/cli.ru.md}}
+
+{{../../usage/disk.ru.md}}
+
+{{../../usage/fio.ru.md}}
+
+{{../../usage/nbd.ru.md}}
+
+{{../../usage/qemu.ru.md}}
+
+{{../../usage/nfs.ru.md}}
+
+## Производительность
+
+{{../../performance/understanding.ru.md}}
+
+{{../../performance/theoretical.ru.md}}
+
+{{../../performance/comparison1.ru.md}}
+
+{{../../intro/author.ru.md|indent=1}}

docs/config/src/make.js

@@ -14,6 +14,7 @@ const L = {
         toc_config: '[Configuration](../config.en.md)',
         toc_usage: 'Usage',
         toc_performance: 'Performance',
+        online: 'Can be changed online: yes',
     },
     ru: {
         Documentation: 'Документация',
@@ -28,6 +29,7 @@ const L = {
         toc_config: '[Конфигурация](../config.ru.md)',
         toc_usage: 'Использование',
         toc_performance: 'Производительность',
+        online: 'Можно менять на лету: да',
     },
 };
 const types = {
@@ -70,6 +72,8 @@ for (const file of params_files)
                 out += `- ${L[lang]['Default'] || 'Default'}: ${c.default}\n`;
             if (c.min !== undefined)
                 out += `- ${L[lang]['Minimum'] || 'Minimum'}: ${c.min}\n`;
+            if (c.online)
+                out += `- ${L[lang]['online'] || 'Can be changed online: yes'}\n`;
             out += `\n`+(c["info_"+lang] || c["info"]).replace(/\s+$/, '');
         }
         const head = fs.readFileSync(__dirname+'/'+file+'.'+lang+'.md', { encoding: 'utf-8' });

docs/config/src/network.yml

@@ -164,18 +164,21 @@
   type: sec
   min: 1
   default: 5
+  online: true
   info: Interval before attempting to reconnect to an unavailable OSD.
   info_ru: Время ожидания перед повторной попыткой соединиться с недоступным OSD.
 - name: peer_connect_timeout
   type: sec
   min: 1
   default: 5
+  online: true
   info: Timeout for OSD connection attempts.
   info_ru: Максимальное время ожидания попытки соединения с OSD.
 - name: osd_idle_timeout
   type: sec
   min: 1
   default: 5
+  online: true
   info: |
     OSD connection inactivity time after which clients and other OSDs send
     keepalive requests to check state of the connection.
@@ -186,6 +189,7 @@
   type: sec
   min: 1
   default: 5
+  online: true
   info: |
     Maximum time to wait for OSD keepalive responses. If an OSD doesn't respond
     within this time, the connection to it is dropped and a reconnection attempt
@@ -198,6 +202,7 @@
   type: ms
   min: 50
   default: 500
+  online: true
   info: |
     OSDs respond to clients with a special error code when they receive I/O
     requests for a PG that's not synchronized and started. This parameter sets
@@ -211,6 +216,7 @@
 - name: max_etcd_attempts
   type: int
   default: 5
+  online: true
   info: |
     Maximum number of attempts for etcd requests which can't be retried
     indefinitely.
@@ -220,6 +226,7 @@
 - name: etcd_quick_timeout
   type: ms
   default: 1000
+  online: true
   info: |
     Timeout for etcd requests which should complete quickly, like lease refresh.
   info_ru: |
@@ -228,6 +235,7 @@
 - name: etcd_slow_timeout
   type: ms
   default: 5000
+  online: true
   info: Timeout for etcd requests which are allowed to wait for some time.
   info_ru: |
     Максимальное время выполнения запросов к etcd, для которых не обязательно
@@ -235,6 +243,7 @@
 - name: etcd_keepalive_timeout
   type: sec
   default: max(30, etcd_report_interval*2)
+  online: true
   info: |
     Timeout for etcd connection HTTP Keep-Alive. Should be higher than
     etcd_report_interval to guarantee that keepalive actually works.
@@ -244,6 +253,7 @@
 - name: etcd_ws_keepalive_timeout
   type: sec
   default: 30
+  online: true
   info: |
     etcd websocket ping interval required to keep the connection alive and
     detect disconnections quickly.
@@ -252,6 +262,7 @@
 - name: client_dirty_limit
   type: int
   default: 33554432
+  online: true
   info: |
     Without immediate_commit=all this parameter sets the limit of "dirty"
     (not committed by fsync) data allowed by the client before forcing an

docs/config/src/osd.en.md

@@ -1,4 +1,5 @@
 # Runtime OSD Parameters
 
 These parameters only apply to OSDs, are not fixed at the moment of OSD drive
-initialization and can be changed with an OSD restart.
+initialization and can be changed - either with an OSD restart or, for some of
+them, even without restarting by updating configuration in etcd.

docs/config/src/osd.ru.md

@@ -2,4 +2,5 @@
 
 Данные параметры используются только OSD, но, в отличие от дисковых параметров,
 не фиксируются в момент инициализации дисков OSD и могут быть изменены в любой
-момент с перезапуском OSD.
+момент с помощью перезапуска OSD, а некоторые и без перезапуска, с помощью
+изменения конфигурации в etcd.

docs/config/src/osd.yml

@@ -66,6 +66,7 @@
 - name: autosync_interval
   type: sec
   default: 5
+  online: true
   info: |
     Time interval at which automatic fsyncs/flushes are issued by each OSD when
     the immediate_commit mode if disabled. fsyncs are required because without
@@ -83,6 +84,7 @@
 - name: autosync_writes
   type: int
   default: 128
+  online: true
   info: |
     Same as autosync_interval, but sets the maximum number of uncommitted write
     operations before issuing an fsync operation internally.
@@ -93,6 +95,7 @@
 - name: recovery_queue_depth
   type: int
   default: 4
+  online: true
   info: |
     Maximum recovery operations per one primary OSD at any given moment of time.
     Currently it's the only parameter available to tune the speed or recovery
@@ -105,6 +108,7 @@
 - name: recovery_pg_switch
   type: int
   default: 128
+  online: true
   info: |
     Number of recovery operations before switching to recovery of the next PG.
     The idea is to mix all PGs during recovery for more even space and load
@@ -119,6 +123,7 @@
 - name: recovery_sync_batch
   type: int
   default: 16
+  online: true
   info: Maximum number of recovery operations before issuing an additional fsync.
   info_ru: Максимальное число операций восстановления перед дополнительным fsync.
 - name: readonly
@@ -133,6 +138,7 @@
 - name: no_recovery
   type: bool
   default: false
+  online: true
   info: |
     Disable automatic background recovery of objects. Note that it doesn't
     affect implicit recovery of objects happening during writes - a write is
@@ -145,6 +151,7 @@
 - name: no_rebalance
   type: bool
   default: false
+  online: true
   info: |
     Disable background movement of data between different OSDs. Disabling it
     means that PGs in the `has_misplaced` state will be left in it indefinitely.
@@ -155,6 +162,7 @@
 - name: print_stats_interval
   type: sec
   default: 3
+  online: true
   info: |
     Time interval at which OSDs print simple human-readable operation
     statistics on stdout.
@@ -164,6 +172,7 @@
 - name: slow_log_interval
   type: sec
   default: 10
+  online: true
   info: |
     Time interval at which OSDs dump slow or stuck operations on stdout, if
     they're any. Also it's the time after which an operation is considered
@@ -175,6 +184,7 @@
 - name: inode_vanish_time
   type: sec
   default: 60
+  online: true
   info: |
     Number of seconds after which a deleted inode is removed from OSD statistics.
   info_ru: |
@@ -182,6 +192,7 @@
 - name: max_write_iodepth
   type: int
   default: 128
+  online: true
   info: |
     Parallel client write operation limit per one OSD. Operations that exceed
     this limit are pushed to a temporary queue instead of being executed
@@ -193,6 +204,7 @@
 - name: min_flusher_count
   type: int
   default: 1
+  online: true
   info: |
     Flusher is a micro-thread that moves data from the journal to the data
     area of the device. Their number is auto-tuned between minimum and maximum.
@@ -204,6 +216,7 @@
 - name: max_flusher_count
   type: int
   default: 256
+  online: true
   info: |
     Maximum number of journal flushers (see above min_flusher_count).
   info_ru: |
@@ -284,6 +297,7 @@
 - name: throttle_small_writes
   type: bool
   default: false
+  online: true
   info: |
     Enable soft throttling of small journaled writes. Useful for hybrid OSDs
     with fast journal/metadata devices and slow data devices. The idea is that
@@ -312,6 +326,7 @@
 - name: throttle_target_iops
   type: int
   default: 100
+  online: true
   info: |
     Target maximum number of throttled operations per second under the condition
     of full journal. Set it to approximate random write iops of your data devices
@@ -324,6 +339,7 @@
 - name: throttle_target_mbs
   type: int
   default: 100
+  online: true
   info: |
     Target maximum bandwidth in MB/s of throttled operations per second under
     the condition of full journal. Set it to approximate linear write
@@ -336,6 +352,7 @@
 - name: throttle_target_parallelism
   type: int
   default: 1
+  online: true
   info: |
     Target maximum parallelism of throttled operations under the condition of
     full journal. Set it to approximate internal parallelism of your data
@@ -348,6 +365,7 @@
 - name: throttle_threshold_us
   type: us
   default: 50
+  online: true
   info: |
     Minimal computed delay to be applied to throttled operations. Usually
     doesn't need to be changed.
@@ -357,10 +375,151 @@
 - name: osd_memlock
   type: bool
   default: false
-  info: >
+  info: |
     Lock all OSD memory to prevent it from being unloaded into swap with
     mlockall(). Requires sufficient ulimit -l (max locked memory).
-  info_ru: >
+  info_ru: |
     Блокировать всю память OSD с помощью mlockall, чтобы запретить её выгрузку
     в пространство подкачки. Требует достаточного значения ulimit -l (лимита
     заблокированной памяти).
+- name: auto_scrub
+  type: bool
+  default: false
+  online: true
+  info: |
+    Data scrubbing is the process of background verification of copies to find
+    and repair corrupted blocks. It's not run automatically by default since
+    it's a new feature. Set this parameter to true to enable automatic scrubs.
+
+    This parameter makes OSDs automatically schedule data scrubbing of clean PGs
+    every `scrub_interval` (see below). You can also start/schedule scrubbing
+    manually by setting `next_scrub` JSON key to the desired UNIX time of the
+    next scrub in `/pg/history/...` values in etcd.
+  info_ru: |
+    Скраб - процесс фоновой проверки копий данных, предназначенный, чтобы
+    находить и исправлять повреждённые блоки. По умолчанию эти проверки ещё не
+    запускаются автоматически, так как являются новой функцией. Чтобы включить
+    автоматическое планирование скрабов, установите данный параметр в true.
+
+    Включённый параметр заставляет OSD автоматически планировать фоновую
+    проверку чистых PG раз в `scrub_interval` (см. ниже). Вы также можете
+    запустить или запланировать проверку вручную, установив значение ключа JSON
+    `next_scrub` внутри ключей etcd `/pg/history/...` в UNIX-время следующей
+    желаемой проверки.
+- name: no_scrub
+  type: bool
+  default: false
+  online: true
+  info: |
+    Temporarily disable scrubbing and stop running scrubs.
+  info_ru: |
+    Временно отключить и остановить запущенные скрабы.
+- name: scrub_interval
+  type: string
+  default: 30d
+  online: true
+  info: |
+    Default automatic scrubbing interval for all pools. Numbers without suffix
+    are treated as seconds, possible unit suffixes include 's' (seconds),
+    'm' (minutes), 'h' (hours), 'd' (days), 'M' (months) and 'y' (years).
+  info_ru: |
+    Интервал автоматической фоновой проверки по умолчанию для всех пулов.
+    Значения без указанной единицы измерения считаются в секундах, допустимые
+    символы единиц измерения в конце: 's' (секунды),
+    'm' (минуты), 'h' (часы), 'd' (дни), 'M' (месяца) или 'y' (годы).
+- name: scrub_queue_depth
+  type: int
+  default: 1
+  online: true
+  info: |
+    Number of parallel scrubbing operations per one OSD.
+  info_ru: |
+    Число параллельных операций фоновой проверки на один OSD.
+- name: scrub_sleep
+  type: ms
+  default: 0
+  online: true
+  info: |
+    Additional interval between two consecutive scrubbing operations on one OSD.
+    Can be used to slow down scrubbing if it affects user load too much.
+  info_ru: |
+    Дополнительный интервал ожидания после фоновой проверки каждого объекта на
+    одном OSD. Может использоваться для замедления скраба, если он слишком
+    сильно влияет на пользовательскую нагрузку.
+- name: scrub_list_limit
+  type: int
+  default: 1000
+  online: true
+  info: |
+    Number of objects to list in one listing operation during scrub.
+  info_ru: |
+    Размер загружаемых за одну операцию списков объектов в процессе фоновой
+    проверки.
+- name: scrub_find_best
+  type: bool
+  default: true
+  online: true
+  info: |
+    Find and automatically restore best versions of objects with unmatched
+    copies. In replicated setups, the best version is the version with most
+    matching replicas. In EC setups, the best version is the subset of data
+    and parity chunks without mismatches.
+
+    The hypothetical situation where you might want to disable it is when
+    you have 3 replicas and you are paranoid that 2 HDDs out of 3 may silently
+    corrupt an object in the same way (for example, zero it out) and only
+    1 HDD will remain good. In this case disabling scrub_find_best may help
+    you to recover the data! See also scrub_ec_max_bruteforce below.
+  info_ru: |
+    Находить и автоматически восстанавливать "лучшие версии" объектов с
+    несовпадающими копиями/частями. При использовании репликации "лучшая"
+    версия - версия, доступная в большем числе экземпляров, чем другие. При
+    использовании кодов коррекции ошибок "лучшая" версия - это подмножество
+    частей данных и чётности, полностью соответствующих друг другу.
+
+    Гипотетическая ситуация, в которой вы можете захотеть отключить этот
+    поиск - это если у вас 3 реплики и вы боитесь, что 2 диска из 3 могут
+    незаметно и одинаково повредить данные одного и того же объекта, например,
+    занулив его, и только 1 диск останется неповреждённым. В этой ситуации
+    отключение этого параметра поможет вам восстановить данные! Смотрите также
+    описание следующего параметра - scrub_ec_max_bruteforce.
+- name: scrub_ec_max_bruteforce
+  type: int
+  default: 100
+  online: true
+  info: |
+    Vitastor can locate corrupted chunks in EC setups with more than 1 parity
+    chunk by brute-forcing all possible error locations. This configuration
+    value limits the maximum number of checked combinations. You can try to
+    increase it if you have EC N+K setup with N and K large enough for
+    combination count `C(N+K-1, K-1) = (N+K-1)! / (K-1)! / N!` to be greater
+    than the default 100.
+
+    If there are too many possible combinations or if multiple combinations give
+    correct results then objects are marked inconsistent and aren't recovered
+    automatically.
+
+    In replicated setups bruteforcing isn't needed, Vitastor just assumes that
+    the variant with most available equal copies is correct. For example, if
+    you have 3 replicas and 1 of them differs, this one is considered to be
+    corrupted. But if there is no "best" version with more copies than all
+    others have then the object is also marked as inconsistent.
+  info_ru: |
+    Vitastor старается определить повреждённые части объектов при использовании
+    EC (кодов коррекции ошибок) с более, чем 1 диском чётности, путём перебора
+    всех возможных комбинаций ошибочных частей. Данное значение конфигурации
+    ограничивает число перебираемых комбинаций. Вы можете попробовать поднять
+    его, если используете схему кодирования EC N+K с N и K, достаточно большими
+    для того, чтобы число сочетаний `C(N+K-1, K-1) = (N+K-1)! / (K-1)! / N!`
+    было больше, чем стандартное значение 100.
+
+    Если возможных комбинаций слишком много или если корректная комбинаций не
+    определяется однозначно, объекты помечаются неконсистентными (inconsistent)
+    и не восстанавливаются автоматически.
+
+    При использовании репликации перебор не нужен, Vitastor просто предполагает,
+    что вариант объекта с наибольшим количеством одинаковых копий корректен.
+    Например, если вы используете 3 реплики и 1 из них отличается, эта 1 копия
+    считается некорректной. Однако, если "лучшую" версию с числом доступных
+    копий большим, чем у всех других версий, найти невозможно, то объект тоже
+    маркируется неконсистентным.

docs/installation/kubernetes.ru.md

@@ -8,13 +8,13 @@
 
 У Vitastor есть CSI-плагин для Kubernetes, поддерживающий RWO, а также блочные RWX, тома.
 
-Для установки возьмите манифесты из директории [csi/deploy/](../csi/deploy/), поместите
-вашу конфигурацию подключения к Vitastor в [csi/deploy/001-csi-config-map.yaml](../csi/deploy/001-csi-config-map.yaml),
-настройте StorageClass в [csi/deploy/009-storage-class.yaml](../csi/deploy/009-storage-class.yaml)
+Для установки возьмите манифесты из директории [csi/deploy/](../../csi/deploy/), поместите
+вашу конфигурацию подключения к Vitastor в [csi/deploy/001-csi-config-map.yaml](../../csi/deploy/001-csi-config-map.yaml),
+настройте StorageClass в [csi/deploy/009-storage-class.yaml](../../csi/deploy/009-storage-class.yaml)
 и примените все `NNN-*.yaml` к вашей инсталляции Kubernetes.
 
 ```
 for i in ./???-*.yaml; do kubectl apply -f $i; done
 ```
 
-После этого вы сможете создавать PersistentVolume. Пример смотрите в файле [csi/deploy/example-pvc.yaml](../csi/deploy/example-pvc.yaml).
+После этого вы сможете создавать PersistentVolume. Пример смотрите в файле [csi/deploy/example-pvc.yaml](../../csi/deploy/example-pvc.yaml).

docs/installation/openstack.en.md

@@ -36,5 +36,5 @@ vitastor_pool_id = 1
 image_upload_use_cinder_backend = True
 ```
 
-To put Glance images in Vitastor, use [https://docs.openstack.org/cinder/pike/admin/blockstorage-volume-backed-image.html](volume-backed images),
+To put Glance images in Vitastor, use [volume-backed images](https://docs.openstack.org/cinder/pike/admin/blockstorage-volume-backed-image.html),
 although the support has not been verified yet.

docs/installation/openstack.ru.md

@@ -36,5 +36,5 @@ image_upload_use_cinder_backend = True
 ```
 
 Чтобы помещать в Vitastor Glance-образы, нужно использовать
-[https://docs.openstack.org/cinder/pike/admin/blockstorage-volume-backed-image.html](образы на основе томов Cinder),
+[образы на основе томов Cinder](https://docs.openstack.org/cinder/pike/admin/blockstorage-volume-backed-image.html),
 однако, поддержка этой функции ещё не проверялась.

docs/installation/packages.en.md

@@ -11,7 +11,8 @@
 - Trust Vitastor package signing key:
   `wget https://vitastor.io/debian/pubkey.gpg -O /etc/apt/trusted.gpg.d/vitastor.gpg`
 - Add Vitastor package repository to your /etc/apt/sources.list:
-  - Debian 11 (Bullseye/Sid): `deb https://vitastor.io/debian bullseye main`
+  - Debian 12 (Bookworm/Sid): `deb https://vitastor.io/debian bookworm main`
+  - Debian 11 (Bullseye): `deb https://vitastor.io/debian bullseye main`
   - Debian 10 (Buster): `deb https://vitastor.io/debian buster main`
 - For Debian 10 (Buster) also enable backports repository:
   `deb http://deb.debian.org/debian buster-backports main`
@@ -45,3 +46,10 @@
 - etcd 3.4.15 or newer. Earlier versions won't work because of various bugs,
   for example [#12402](https://github.com/etcd-io/etcd/pull/12402).
 - node.js 10 or newer
+
+## Version archive
+
+All previous Vitastor and other components (QEMU, etcd...) package builds
+can be found here:
+
+https://vitastor.io/archive/

docs/installation/packages.ru.md

@@ -11,7 +11,8 @@
 - Добавьте ключ репозитория Vitastor:
   `wget https://vitastor.io/debian/pubkey.gpg -O /etc/apt/trusted.gpg.d/vitastor.gpg`
 - Добавьте репозиторий Vitastor в /etc/apt/sources.list:
-  - Debian 11 (Bullseye/Sid): `deb https://vitastor.io/debian bullseye main`
+  - Debian 12 (Bookworm/Sid): `deb https://vitastor.io/debian bookworm main`
+  - Debian 11 (Bullseye): `deb https://vitastor.io/debian bullseye main`
   - Debian 10 (Buster): `deb https://vitastor.io/debian buster main`
 - Для Debian 10 (Buster) также включите репозиторий backports:
   `deb http://deb.debian.org/debian buster-backports main`
@@ -44,3 +45,10 @@
 - etcd 3.4.15 или новее. Более старые версии не будут работать из-за разных багов,
   например, [#12402](https://github.com/etcd-io/etcd/pull/12402).
 - node.js 10 или новее
+
+## Архив предыдущих версий
+
+Все предыдущие сборки пакетов Vitastor и других компонентов, таких, как QEMU
+и etcd, можно скачать по следующей ссылке:
+
+https://vitastor.io/archive/

docs/installation/proxmox.en.md

@@ -6,10 +6,10 @@
 
 # Proxmox VE
 
-To enable Vitastor support in Proxmox Virtual Environment (6.4-7.4 are supported):
+To enable Vitastor support in Proxmox Virtual Environment (6.4-8.0 are supported):
 
 - Add the corresponding Vitastor Debian repository into sources.list on Proxmox hosts:
-  buster for 6.4, bullseye for 7.4, pve7.1 for 7.1, pve7.2 for 7.2, pve7.3 for 7.3
+  bookworm for 8.0, bullseye for 7.4, pve7.3 for 7.3, pve7.2 for 7.2, pve7.1 for 7.1, buster for 6.4
 - Install vitastor-client, pve-qemu-kvm, pve-storage-vitastor (* or see note) packages from Vitastor repository
 - Define storage in `/etc/pve/storage.cfg` (see below)
 - Block network access from VMs to Vitastor network (to OSDs and etcd),
@@ -35,5 +35,5 @@ vitastor: vitastor
     vitastor_nbd 0
 ```
 
-\* Note: you can also manually copy [patches/VitastorPlugin.pm](patches/VitastorPlugin.pm) to Proxmox hosts
+\* Note: you can also manually copy [patches/VitastorPlugin.pm](../../patches/VitastorPlugin.pm) to Proxmox hosts
 as `/usr/share/perl5/PVE/Storage/Custom/VitastorPlugin.pm` instead of installing pve-storage-vitastor.

docs/installation/proxmox.ru.md

@@ -4,12 +4,12 @@
 
 [Read in English](proxmox.en.md)
 
-# Proxmox
+# Proxmox VE
 
-Чтобы подключить Vitastor к Proxmox Virtual Environment (поддерживаются версии 6.4-7.4):
+Чтобы подключить Vitastor к Proxmox Virtual Environment (поддерживаются версии 6.4-8.0):
 
 - Добавьте соответствующий Debian-репозиторий Vitastor в sources.list на хостах Proxmox:
-  buster для 6.4, bullseye для 7.4, pve7.1 для 7.1, pve7.2 для 7.2, pve7.3 для 7.3
+  bookworm для 8.0, bullseye для 7.4, pve7.3 для 7.3, pve7.2 для 7.2, pve7.1 для 7.1, buster для 6.4
 - Установите пакеты vitastor-client, pve-qemu-kvm, pve-storage-vitastor (* или см. сноску) из репозитория Vitastor
 - Определите тип хранилища в `/etc/pve/storage.cfg` (см. ниже)
 - Обязательно заблокируйте доступ от виртуальных машин к сети Vitastor (OSD и etcd), т.к. Vitastor (пока) не поддерживает аутентификацию
@@ -35,5 +35,5 @@ vitastor: vitastor
 ```
 
 \* Примечание: вместо установки пакета pve-storage-vitastor вы можете вручную скопировать файл
-[patches/VitastorPlugin.pm](patches/VitastorPlugin.pm) на хосты Proxmox как
+[patches/VitastorPlugin.pm](../../patches/VitastorPlugin.pm) на хосты Proxmox как
 `/usr/share/perl5/PVE/Storage/Custom/VitastorPlugin.pm`.

docs/intro/architecture.en.md

@@ -44,7 +44,7 @@
   depends linearly on drive capacity and data store block size which is 128 KB by default.
   With 128 KB blocks metadata takes around 512 MB per 1 TB (which is still less than Ceph wants).
   Journal is also kept in memory by default, but in SSD-only clusters it's only 32 MB, and in SSD+HDD
-  clusters, where it's beneficial to increase it, [inmemory_journal](docs/config/osd.en.md#inmemory_journal) can be disabled.
+  clusters, where it's beneficial to increase it, [inmemory_journal](../config/osd.en.md#inmemory_journal) can be disabled.
 - Vitastor storage layer doesn't have internal copy-on-write or redirect-write. I know that maybe
   it's possible to create a good copy-on-write storage, but it's much harder and makes performance
   less deterministic, so CoW isn't used in Vitastor.

docs/intro/architecture.ru.md

@@ -156,7 +156,7 @@
   блока хранилища (block_size, по умолчанию 128 КБ). С 128 КБ блоком потребление памяти
   составляет примерно 512 МБ на 1 ТБ данных. Журналы по умолчанию тоже хранятся в памяти,
   но в SSD-кластерах нужный размер журнала составляет всего 32 МБ, а в гибридных (SSD+HDD)
-  кластерах, в которых есть смысл делать журналы больше, можно отключить [inmemory_journal](../docs/config/osd.ru.md#inmemory_journal).
+  кластерах, в которых есть смысл делать журналы больше, можно отключить [inmemory_journal](../config/osd.ru.md#inmemory_journal).
 - В Vitastor нет внутреннего copy-on-write. Я считаю, что реализация CoW-хранилища гораздо сложнее,
   поэтому сложнее добиться устойчиво хороших результатов. Возможно, в один прекрасный день
   я придумаю красивый алгоритм для CoW-хранилища, но пока нет — внутреннего CoW в Vitastor не будет.

docs/intro/features.en.md

@@ -29,12 +29,13 @@
 - Snapshots and copy-on-write image clones
 - [Write throttling to smooth random write workloads in SSD+HDD configurations](../config/osd.en.md#throttle_small_writes)
 - [RDMA/RoCEv2 support via libibverbs](../config/network.en.md#rdma_device)
+- [Scrubbing without checksums](../config/osd.en.md#auto_scrub) (verification of copies)
 
 ## Plugins and tools
 
 - [Debian and CentOS packages](../installation/packages.en.md)
 - [Image management CLI (vitastor-cli)](../usage/cli.en.md)
-- [Disk management CLI (vitastor-disk)](docs/usage/disk.en.md)
+- [Disk management CLI (vitastor-disk)](../usage/disk.en.md)
 - Generic user-space client library
 - [Native QEMU driver](../usage/qemu.en.md)
 - [Loadable fio engine for benchmarks](../usage/fio.en.md)
@@ -54,7 +55,6 @@ The following features are planned for the future:
 - iSCSI proxy
 - Multi-threaded client
 - Faster failover
-- Scrubbing without checksums (verification of replicas)
 - Checksums
 - Tiered storage (SSD caching)
 - NVDIMM support

docs/intro/features.ru.md

@@ -13,7 +13,7 @@
 ## Серверные функции
 
 - Базовая часть - надёжное кластерное блочное хранилище без единой точки отказа
-- [Производительность](../comparison1.ru.md) ;-D
+- [Производительность](../performance/comparison1.ru.md) ;-D
 - [Несколько схем отказоустойчивости](../config/pool.ru.md#scheme): репликация, XOR n+1 (1 диск чётности), коды коррекции ошибок
   Рида-Соломона на основе библиотек jerasure и ISA-L с любым числом дисков данных и чётности в группе
 - Конфигурация через простые человекочитаемые JSON-структуры в etcd
@@ -31,12 +31,13 @@
 - Снапшоты и copy-on-write клоны
 - [Сглаживание производительности случайной записи в SSD+HDD конфигурациях](../config/osd.ru.md#throttle_small_writes)
 - [Поддержка RDMA/RoCEv2 через libibverbs](../config/network.ru.md#rdma_device)
+- [Фоновая проверка целостности без контрольных сумм](../config/osd.ru.md#auto_scrub) (сверка копий)
 
 ## Драйверы и инструменты
 
 - [Пакеты для Debian и CentOS](../installation/packages.ru.md)
 - [Консольный интерфейс управления образами (vitastor-cli)](../usage/cli.ru.md)
-- [Инструмент управления дисками (vitastor-disk)](docs/usage/disk.ru.md)
+- [Инструмент управления дисками (vitastor-disk)](../usage/disk.ru.md)
 - Общая пользовательская клиентская библиотека для работы с кластером
 - [Драйвер диска для QEMU](../usage/qemu.ru.md)
 - [Драйвер диска для утилиты тестирования производительности fio](../usage/fio.ru.md)
@@ -54,7 +55,6 @@
 - iSCSI-прокси
 - Многопоточный клиент
 - Более быстрое переключение при отказах
-- Фоновая проверка целостности без контрольных сумм (сверка реплик)
 - Контрольные суммы
 - Поддержка SSD-кэширования (tiered storage)
 - Поддержка NVDIMM

docs/usage/cli.en.md

@@ -20,6 +20,8 @@ It supports the following commands:
 - [flatten](#flatten)
 - [rm-data](#rm-data)
 - [merge-data](#merge-data)
+- [describe](#describe)
+- [fix](#fix)
 - [alloc-osd](#alloc-osd)
 - [rm-osd](#rm-osd)
 
@@ -174,6 +176,51 @@ Merge layer data without changing metadata. Merge `<from>`..`<to>` to `<target>`
 `<to>` must be a child of `<from>` and `<target>` may be one of the layers between
 `<from>` and `<to>`, including `<from>` and `<to>`.
 
+## describe
+
+`vitastor-cli describe [--osds <osds>] [--object-state <states>] [--pool <pool>]
+    [--inode <ino>] [--min-inode <ino>] [--max-inode <ino>]
+    [--min-offset <offset>] [--max-offset <offset>]`
+
+Describe unclean object locations in the cluster.
+
+```
+--osds <osds>
+    Only list objects from primary OSD(s) <osds>.
+--object-state <states>
+    Only list objects in given state(s). State(s) may include:
+    degraded, misplaced, incomplete, corrupted, inconsistent.
+--pool <pool name or number>
+    Only list objects in the given pool.
+--inode, --min-inode, --max-inode
+    Restrict listing to specific inode numbers.
+--min-offset, --max-offset
+    Restrict listing to specific offsets inside inodes.
+```
+
+## fix
+
+`vitastor-cli fix [--objects <objects>] [--bad-osds <osds>] [--part <part>] [--check no]`
+
+Fix inconsistent objects in the cluster by deleting some copies.
+
+```
+--objects <objects>
+    Objects to fix, either in plain text or JSON format. If not specified,
+    object list will be read from STDIN in one of the same formats.
+    Plain text format: 0x<inode>:0x<stripe> <any delimiter> 0x<inode>:0x<stripe> ...
+    JSON format: [{"inode":"0x...","stripe":"0x..."},...]
+--bad-osds <osds>
+    Remove inconsistent copies/parts of objects from these OSDs, effectively
+    marking them bad and allowing Vitastor to recover objects from other copies.
+--part <number>
+    Only remove EC part <number> (from 0 to pg_size-1), required for extreme
+    edge cases where one OSD has multiple parts of a EC object.
+--check no
+    Do not recheck that requested objects are actually inconsistent,
+    delete requested copies/parts anyway.
+```
+
 ## alloc-osd
 
 `vitastor-cli alloc-osd`

docs/usage/cli.ru.md

@@ -184,6 +184,59 @@ vitastor-cli snap-create [-p|--pool <id|name>] <image>@<snapshot>
 в целевой образ `<target>`. `<to>` должен быть дочерним образом `<from>`, а `<target>`
 должен быть одним из слоёв между `<from>` и `<to>`, включая сами `<from>` и `<to>`.
 
+## describe
+
+`vitastor-cli describe [--osds <osds>] [--object-state <состояния>] [--pool <пул>]
+    [--inode <номер>] [--min-inode <номер>] [--max-inode <номер>]
+    [--min-offset <смещение>] [--max-offset <смещение>]`
+
+Описать состояние "грязных" объектов в кластере, то есть таких объектов, копии
+или части которых хранятся на наборе OSD, не равном целевому.
+
+```
+--osds <osds>
+    Перечислять только объекты с первичных OSD из списка <osds>.
+--object-state <состояния>
+    Перечислять только объекты в указанных состояниях. Возможные состояния
+    объектов:
+    - degraded - деградированная избыточность
+    - misplaced - перемещённый
+    - incomplete - нечитаемый из-за потери большего числа частей, чем допустимо
+    - corrupted - с одной или более повреждённой частью
+    - inconsistent - неконсистентный, с неоднозначным расхождением копий/частей
+--pool <имя или ID пула>
+    Перечислять только объекты из заданного пула.
+--inode, --min-inode, --max-inode
+    Перечислять только объекты из указанных номеров инодов (образов).
+--min-offset, --max-offset
+    Перечислять только объекты с заданных смещений внутри образов.
+```
+
+## fix
+
+`vitastor-cli fix [--objects <объекты>] [--bad-osds <osds>] [--part <номер>] [--check no]`
+
+Исправить неконсистентные (неоднозначные) объекты путём удаления части копий.
+
+```
+--objects <объекты>
+    Объекты для исправления - в простом текстовом или JSON формате. Если опция
+    не указана, список объектов читается со стандартного ввода в тех же форматах.
+    Простой формат: 0x<инод>:0x<смещение> <любой разделитель> 0x<инод>:0x<смещение> ...
+    Формат JSON: [{"inode":"0x<инод>","stripe":"0x<смещение>"},...]
+--bad-osds <osds>
+    Удалить неконсистентные копии/части объектов с данных OSD, таким образом
+    признавая потерю этих копий и позволяя Vitastor-у восстановить объекты из
+    других копий.
+--part <номер>
+    Удалить только части EC с заданным номером (от 0 до pg_size-1). Нужно только
+    в редких граничных случаях, когда один и тот же OSD содержит несколько частей
+    одного EC-объекта.
+--check no
+    Не перепроверять, что заданные объекты действительно в неконсистентном
+    состоянии и просто удалять заданные части.
+```
+
 ## alloc-osd
 
 `vitastor-cli alloc-osd`

docs/usage/nbd.en.md

@@ -25,6 +25,23 @@ It will output a block device name like /dev/nbd0 which you can then use as a no
 
 You can also use `--pool <POOL> --inode <INODE> --size <SIZE>` instead of `--image <IMAGE>` if you want.
 
+Additional options for map command:
+
+* `--nbd_timeout 30` \
+  Timeout for I/O operations in seconds after exceeding which the kernel stops
+  the device. You can set it to 0 to disable the timeout, but beware that you
+  won't be able to stop the device at all if vitastor-nbd process dies.
+* `--nbd_max_devices 64 --nbd_max_part 3` \
+  Options for the `nbd` kernel module when modprobing it (`nbds_max` and `max_part`).
+  note that maximum allowed (nbds_max)*(1+max_part) is 256.
+* `--logfile /path/to/log/file.txt` \
+  Write log messages to the specified file instead of dropping them (in background mode)
+  or printing them to the standard output (in foreground mode).
+* `--dev_num N` \
+  Use the specified device /dev/nbdN instead of automatic selection.
+* `--foreground 1` \
+  Stay in foreground, do not daemonize.
+
 ## Unmap image
 
 To unmap the device run:
@@ -32,3 +49,27 @@ To unmap the device run:
 ```
 vitastor-nbd unmap /dev/nbd0
 ```
+
+## List mapped images
+
+```
+vitastor-nbd ls [--json]
+```
+
+Example output (normal format):
+
+```
+/dev/nbd0
+image: bench
+pid: 584536
+
+/dev/nbd1
+image: bench1
+pid: 584546
+```
+
+Example output (JSON format):
+
+```
+{"/dev/nbd0": {"image": "bench", "pid": 584536}, "/dev/nbd1": {"image": "bench1", "pid": 584546}}
+```

docs/usage/nbd.ru.md

@@ -30,6 +30,27 @@ vitastor-nbd map --etcd_address 10.115.0.10:2379/v3 --image testimg
 Для обращения по номеру инода, аналогично другим командам, можно использовать опции
 `--pool <POOL> --inode <INODE> --size <SIZE>` вместо `--image testimg`.
 
+Дополнительные опции для команды подключения NBD-устройства:
+
+* `--nbd_timeout 30` \
+  Максимальное время выполнения любой операции чтения/записи в секундах, при
+  превышении которого ядро остановит NBD-устройство. Вы можете установить опцию
+  в 0, чтобы отключить ограничение времени, но имейте в виду, что в этом случае
+  вы вообще не сможете отключить NBD-устройство при нештатном завершении процесса
+  vitastor-nbd.
+* `--nbd_max_devices 64 --nbd_max_part 3` \
+  Опции, передаваемые модулю ядра nbd, если его загружает vitastor-nbd
+  (`nbds_max` и `max_part`). Имейте в виду, что (nbds_max)*(1+max_part)
+  обычно не должно превышать 256.
+* `--logfile /path/to/log/file.txt` \
+  Писать сообщения о процессе работы в заданный файл, вместо пропуска их
+  при фоновом режиме запуска или печати на стандартный вывод при запуске
+  в консоли с `--foreground 1`.
+* `--dev_num N` \
+  Использовать заданное устройство `/dev/nbdN` вместо автоматического подбора.
+* `--foreground 1` \
+  Не уводить процесс в фоновый режим.
+
 ## Отключить устройство
 
 Для отключения устройства выполните:
@@ -37,3 +58,27 @@ vitastor-nbd map --etcd_address 10.115.0.10:2379/v3 --image testimg
 ```
 vitastor-nbd unmap /dev/nbd0
 ```
+
+## Вывести подключённые устройства
+
+```
+vitastor-nbd ls [--json]
+```
+
+Пример вывода в обычном формате:
+
+```
+/dev/nbd0
+image: bench
+pid: 584536
+
+/dev/nbd1
+image: bench1
+pid: 584546
+```
+
+Пример вывода в JSON-формате:
+
+```
+{"/dev/nbd0": {"image": "bench", "pid": 584536}, "/dev/nbd1": {"image": "bench1", "pid": 584546}}
+```

docs/usage/nfs.ru.md

@@ -29,7 +29,7 @@ vitastor-nfs [--etcd_address ADDR] [ДРУГИЕ ОПЦИИ]
 --bind <IP>       принимать соединения по адресу <IP> (по умолчанию 0.0.0.0 - на всех)
 --nfspath <PATH>  установить путь NFS-экспорта в <PATH> (по умолчанию /)
 --port <PORT>     использовать порт <PORT> для NFS-сервисов (по умолчанию 2049)
---pool <POOL>     исползовать пул <POOL> для новых образов (обязательно, если пул в кластере не один)
+--pool <POOL>     использовать пул <POOL> для новых образов (обязательно, если пул в кластере не один)
 --foreground 1    не уходить в фон после запуска
 ```
 

mon/lp-optimizer.js

@@ -50,8 +50,7 @@ async function lp_solve(text)
     return { score, vars };
 }
 
-async function optimize_initial({ osd_tree, pg_count, pg_size = 3, pg_minsize = 2, hier_sizes = null,
-    max_combinations = 10000, parity_space = 1, ordered = false, seq_layout = false })
+async function optimize_initial({ osd_tree, pg_count, pg_size = 3, pg_minsize = 2, max_combinations = 10000, parity_space = 1, ordered = false })
 {
     if (!pg_count || !osd_tree)
     {
@@ -59,7 +58,7 @@ async function optimize_initial({ osd_tree, pg_count, pg_size = 3, pg_minsize =
     }
     const all_weights = Object.assign({}, ...Object.values(osd_tree));
     const total_weight = Object.values(all_weights).reduce((a, c) => Number(a) + Number(c), 0);
-    const all_pgs = Object.values(random_hier_combinations(osd_tree, hier_sizes || [ pg_size, 1 ], max_combinations, parity_space > 1, seq_layout));
+    const all_pgs = Object.values(random_combinations(osd_tree, pg_size, max_combinations, parity_space > 1));
     const pg_per_osd = {};
     for (const pg of all_pgs)
     {
@@ -217,45 +216,39 @@ function calc_intersect_weights(old_pg_size, pg_size, pg_count, prev_weights, al
     return move_weights;
 }
 
-function build_parent_per_leaf(osd_tree, res = {}, parents = [])
-{
-    for (const item in osd_tree)
-    {
-        if (osd_tree[item] instanceof Object)
-            build_parent_per_leaf(osd_tree[item], res, [ ...parents, item ]);
-        else
-            res[item] = parents;
-    }
-    return res;
-}
-
-function add_valid_previous(osd_tree, prev_weights, all_pgs, hier_sizes)
+function add_valid_previous(osd_tree, prev_weights, all_pgs)
 {
     // Add previous combinations that are still valid
-    const parent_per_osd = build_parent_per_leaf(osd_tree);
+    const hosts = Object.keys(osd_tree).sort();
+    const host_per_osd = {};
+    for (const host in osd_tree)
+    {
+        for (const osd in osd_tree[host])
+        {
+            host_per_osd[osd] = host;
+        }
+    }
     skip_pg: for (const pg_name in prev_weights)
     {
-        const seen = [];
+        const seen_hosts = {};
         const pg = pg_name.substr(3).split(/_/);
         for (const osd of pg)
         {
-            if (!parent_per_osd[osd])
-                continue skip_pg;
-            for (let i = 0; i < parent_per_osd[osd].length; i++)
+            if (!host_per_osd[osd] || seen_hosts[host_per_osd[osd]])
             {
-                seen[parent_per_osd[osd][i]]++;
-                if (seen[parent_per_osd[osd][i]] > hier_sizes[i])
-                    continue skip_pg;
+                continue skip_pg;
             }
+            seen_hosts[host_per_osd[osd]] = true;
         }
         if (!all_pgs[pg_name])
+        {
             all_pgs[pg_name] = pg;
+        }
     }
 }
 
 // Try to minimize data movement
-async function optimize_change({ prev_pgs: prev_int_pgs, osd_tree, pg_size = 3, pg_minsize = 2,
-    hier_sizes = null, max_combinations = 10000, parity_space = 1, ordered = false, seq_layout = false })
+async function optimize_change({ prev_pgs: prev_int_pgs, osd_tree, pg_size = 3, pg_minsize = 2, max_combinations = 10000, parity_space = 1, ordered = false })
 {
     if (!osd_tree)
     {
@@ -280,10 +273,10 @@ async function optimize_change({ prev_pgs: prev_int_pgs, osd_tree, pg_size = 3,
     }
     const old_pg_size = prev_int_pgs[0].length;
     // Get all combinations
-    let all_pgs = random_hier_combinations(osd_tree, hier_sizes || [ pg_size, 1 ], max_combinations, parity_space > 1, seq_layout);
+    let all_pgs = random_combinations(osd_tree, pg_size, max_combinations, parity_space > 1);
     if (old_pg_size == pg_size)
     {
-        add_valid_previous(osd_tree, prev_weights, all_pgs, hier_sizes || [ pg_size, 1 ]);
+        add_valid_previous(osd_tree, prev_weights, all_pgs);
     }
     all_pgs = Object.values(all_pgs);
     const pg_per_osd = {};
@@ -509,147 +502,41 @@ function put_aligned_pgs(aligned_pgs, int_pgs, prev_int_pgs, keygen)
     }
 }
 
-// Convert multi-level tree_node = { level: number|string, id?: string, size?: number, children?: tree_node[] }
+// Convert multi-level osd_tree = { level: number|string, id?: string, size?: number, children?: osd_tree }[]
 // levels = { string: number }
-// to a multi-level OSD tree suitable for random_hier_combinations()
-// (or in case of just 2 levels - for all_combinations() / random_combinations())
-//
-// Example:
-// tree_node = { level: 'dc', children: [ { level: 'rack', children: [ { level: 'host', children: [ { level: 'osd', size: 10 } ] } ] } ] }
-// extract_levels = [ 'rack', 'osd' ]
-// level_defs = { dc: 1, rack: 2, host: 3, osd: 4 }
-//
-// Result:
-// { rack0: { osd1: 10 } }
-function extract_tree_levels(tree_node, extract_levels, level_defs, new_tree = { idx: 1, items: {} })
+// to a two-level osd_tree suitable for all_combinations()
+function flatten_tree(osd_tree, levels, failure_domain_level, osd_level, domains = {}, i = { i: 1 })
 {
-    const next_level = Number(level_defs[extract_levels[0]] || extract_levels[0]) || 0;
-    const level_name = level_defs[extract_levels[0]] ? extract_levels[0] : 'l'+extract_levels[0]+'_';
-    const is_leaf = extract_levels.length == 1;
-    if ((level_defs[tree_node.level] || tree_node.level) >= next_level)
+    osd_level = levels[osd_level] || osd_level;
+    failure_domain_level = levels[failure_domain_level] || failure_domain_level;
+    for (const node of osd_tree)
     {
-        if (!is_leaf)
+        if ((levels[node.level] || node.level) < failure_domain_level)
         {
-            // Insert a (possibly fake) level
-            const nt = { idx: 1, items: {} };
-            new_tree.items[level_name+(new_tree.idx++)] = nt.items;
-            extract_tree_levels(tree_node, extract_levels.slice(1), level_defs, nt);
+            flatten_tree(node.children||[], levels, failure_domain_level, osd_level, domains, i);
         }
         else
         {
-            // Insert a leaf node
-            const leaf_id = tree_node.id || (level_name+(new_tree.idx++));
-            new_tree.items[leaf_id] = tree_node.size;
+            domains['dom'+(i.i++)] = extract_osds([ node ], levels, osd_level);
         }
     }
-    else
-    {
-        for (const child_node of tree_node.children||[])
-        {
-            extract_tree_levels(child_node, extract_levels, level_defs, new_tree);
-        }
-    }
-    return new_tree.items;
+    return domains;
 }
 
-// generate random PGs with hierarchical failure domains, i.e. for example 3 DC each with 2 HOSTS
-// osd_tree = { level3_id: { level2_id: { level1_id: scalar_value } }, ... }
-//   osd_tree may contain arbitrary number of levels, but level count must be the same across the whole tree
-// size_per_level = number of items to select on each level, for example [3, 2, 1].
-//   must have the same number of items as the osd_tree level count.
-// count = PG count to generate
-// ordered = don't treat (x,y) and (y,x) as equal
-// seq_layout = true for the [DC1,DC1,DC2,DC2,DC3,DC3] layout, false for [DC1,DC2,DC3,DC1,DC2,DC3] layout
-function random_hier_combinations(osd_tree, size_per_level, count, ordered, seq_layout)
+function extract_osds(osd_tree, levels, osd_level, osds = {})
 {
-    let seed = 0x5f020e43;
-    const rng = () =>
+    for (const node of osd_tree)
     {
-        seed ^= seed << 13;
-        seed ^= seed >> 17;
-        seed ^= seed << 5;
-        return seed + 2147483648;
-    };
-    const get_max_level = (o) =>
-    {
-        let lvl = 0;
-        while (o instanceof Object)
+        if ((levels[node.level] || node.level) >= osd_level)
         {
-            for (const k in o)
-            {
-                lvl++;
-                o = o[k];
-                break;
-            }
+            osds[node.id] = node.size;
         }
-        return lvl;
-    };
-    const max_level = get_max_level(osd_tree);
-    const gen_pg = (select) =>
-    {
-        let pg = [ osd_tree ];
-        for (let level = 0; level < max_level; level++)
+        else
         {
-            let npg = [];
-            for (let i = 0; i < pg.length; i++)
-            {
-                const keys = pg[i] instanceof Object ? Object.keys(pg[i]) : [];
-                const max_keys = keys.length < size_per_level[level] ? keys.length : size_per_level[level];
-                for (let j = 0; j < max_keys; j++)
-                {
-                    const r = select(level, i, j, (ordered ? keys.length : (keys.length - (max_keys - j - 1))));
-                    const el = pg[i][keys[r]] instanceof Object ? pg[i][keys[r]] : keys[r];
-                    npg[seq_layout ? i*size_per_level[level]+j : j*pg.length+i] = el;
-                    keys.splice(ordered ? r : 0, ordered ? 1 : (r+1));
-                }
-                for (let j = max_keys; j < size_per_level[level]; j++)
-                    npg[seq_layout ? i*size_per_level[level]+j : j*pg.length+i] = NO_OSD;
-            }
-            pg = npg;
+            extract_osds(node.children||[], levels, osd_level, osds);
         }
-        return pg;
-    };
-    const r = {};
-    // Generate random combinations including each OSD at least once
-    let has_next = true;
-    let ctr = [];
-    while (has_next)
-    {
-        let pg = gen_pg((level, i, j, n) =>
-        {
-            if (i == 0 && j == 0)
-            {
-                // Select a pre-determined OSD in the first position on each level
-                const r = ctr[level] == null || ctr[level][1] != n ? 0 : ctr[level][0];
-                ctr[level] = [ r, n ];
-                return r;
-            }
-            return rng() % n;
-        });
-        for (let i = ctr.length-1; i >= 0; i--)
-        {
-            ctr[i][0]++;
-            if (ctr[i][0] < ctr[i][1])
-                break;
-            else
-                ctr[i] = null;
-        }
-        has_next = ctr[0] != null;
-        const cyclic_pgs = [ pg ];
-        if (ordered)
-            for (let i = 1; i < pg.size; i++)
-                cyclic_pgs.push([ ...pg.slice(i), ...pg.slice(0, i) ]);
-        for (const pg of cyclic_pgs)
-            r['pg_'+pg.join('_')] = pg;
     }
-    // Generate purely random combinations
-    while (count > 0)
-    {
-        let pg = gen_pg((l, i, j, n) => rng() % n);
-        r['pg_'+pg.join('_')] = pg;
-        count--;
-    }
-    return r;
+    return osds;
 }
 
 // ordered = don't treat (x,y) and (y,x) as equal
@@ -865,12 +752,11 @@ module.exports = {
     pg_weights_space_efficiency,
     pg_list_space_efficiency,
     pg_per_osd_space_efficiency,
-    extract_tree_levels,
+    flatten_tree,
 
     lp_solve,
     make_int_pgs,
     align_pgs,
     random_combinations,
-    random_hier_combinations,
     all_combinations,
 };

mon/make-etcd

@@ -63,7 +63,7 @@ Wants=network-online.target local-fs.target time-sync.target
 
 [Service]
 Restart=always
-ExecStart=/usr/local/bin/etcd -name etcd${num} --data-dir /var/lib/etcd${num}.etcd \\
+ExecStart=etcd -name etcd${num} --data-dir /var/lib/etcd${num}.etcd \\
     --advertise-client-urls http://${etcds[num]}:2379 --listen-client-urls http://${etcds[num]}:2379 \\
     --initial-advertise-peer-urls http://${etcds[num]}:2380 --listen-peer-urls http://${etcds[num]}:2380 \\
     --initial-cluster-token vitastor-etcd-1 --initial-cluster ${etcd_cluster} \\

mon/mon.js

@@ -104,12 +104,21 @@ const etcd_tree = {
             autosync_writes: 128,
             client_queue_depth: 128, // unused
             recovery_queue_depth: 4,
+            recovery_pg_switch: 128,
             recovery_sync_batch: 16,
             no_recovery: false,
             no_rebalance: false,
             print_stats_interval: 3,
             slow_log_interval: 10,
             inode_vanish_time: 60,
+            auto_scrub: false,
+            no_scrub: false,
+            scrub_interval: '30d', // 1s/1m/1h/1d
+            scrub_queue_depth: 1,
+            scrub_sleep: 0, // milliseconds
+            scrub_list_limit: 1000, // objects to list on one scrub iteration
+            scrub_find_best: true,
+            scrub_ec_max_bruteforce: 100, // maximum EC error locator brute-force iterators
             // blockstore - fixed in superblock
             block_size,
             disk_alignment,
@@ -159,10 +168,6 @@ const etcd_tree = {
                 // number of parity chunks, required for EC
                 parity_chunks?: 1,
                 pg_count: 100,
-                // failure_domain = string | { string: int }
-                // the second case specifies multiple failure domains. example:
-                // { datacenter: 3, host: 2 } - means 3 datacenters with 2 hosts each, for EC 4+2
-                // guarantees availability on outage of either 1 datacenter or 2 hosts
                 failure_domain: 'host',
                 max_osd_combinations: 10000,
                 // block_size, bitmap_granularity, immediate_commit must match all OSDs used in that pool
@@ -176,6 +181,8 @@ const etcd_tree = {
                 osd_tags?: 'nvme' | [ 'nvme', ... ],
                 // prefer to put primary on OSD with these tags
                 primary_affinity_tags?: 'nvme' | [ 'nvme', ... ],
+                // scrub interval
+                scrub_interval?: '30d',
             },
             ...
         }, */
@@ -271,7 +278,7 @@ const etcd_tree = {
                     primary: osd_num_t,
                     state: ("starting"|"peering"|"incomplete"|"active"|"repeering"|"stopping"|"offline"|
                         "degraded"|"has_incomplete"|"has_degraded"|"has_misplaced"|"has_unclean"|
-                        "has_invalid"|"left_on_dead")[],
+                        "has_invalid"|"has_inconsistent"|"has_corrupted"|"left_on_dead"|"scrubbing")[],
                 }
             }, */
         },
@@ -293,6 +300,7 @@ const etcd_tree = {
                     osd_sets: osd_num_t[][],
                     all_peers: osd_num_t[],
                     epoch: uint64_t,
+                    next_scrub: uint64_t,
                 },
             }, */
         },
@@ -383,6 +391,7 @@ class Mon
         this.etcd_start_timeout = (config.etcd_start_timeout || 5) * 1000;
         this.state = JSON.parse(JSON.stringify(this.constructor.etcd_tree));
         this.signals_set = false;
+        this.stat_time = Date.now();
         this.ws = null;
         this.ws_alive = false;
         this.ws_keepalive_timer = null;
@@ -1031,32 +1040,6 @@ class Mon
         pool_cfg.parity_chunks = Math.floor(pool_cfg.parity_chunks) || undefined;
         pool_cfg.pg_count = Math.floor(pool_cfg.pg_count);
         pool_cfg.failure_domain = pool_cfg.failure_domain || 'host';
-        if (pool_cfg.failure_domain instanceof Object)
-        {
-            for (const key in pool_cfg.failure_domain)
-            {
-                const cnt = parseInt(pool_cfg.failure_domain[key]);
-                if (!cnt || cnt <= 0)
-                {
-                    if (warn)
-                        console.log('Pool '+pool_id+' specifies invalid item count for failure domain \"'+key+'\"');
-                    return false;
-                }
-                if (key !== 'host' && key != 'osd' && !(key in this.config.placement_levels||{}))
-                {
-                    if (warn)
-                        console.log('Pool '+pool_id+' uses invalid failure domain \"'+key+'\"');
-                    return false;
-                }
-            }
-        }
-        else if (pool_cfg.failure_domain !== 'host' && pool_cfg.failure_domain != 'osd' &&
-            !(pool_cfg.failure_domain in this.config.placement_levels||{}))
-        {
-            if (warn)
-                console.log('Pool '+pool_id+' uses invalid failure domain \"'+pool_cfg.failure_domain+'\"');
-            return false;
-        }
         pool_cfg.max_osd_combinations = Math.floor(pool_cfg.max_osd_combinations) || 10000;
         if (!/^[1-9]\d*$/.exec(''+pool_id))
         {
@@ -1142,23 +1125,27 @@ class Mon
     filter_osds_by_tags(orig_tree, flat_tree, tags)
     {
         if (!tags)
-            return 1;
+        {
+            return;
+        }
         for (const tag of (tags instanceof Array ? tags : [ tags ]))
         {
-            for (const item in flat_tree)
+            for (const host in flat_tree)
             {
-                if (flat_tree[item] instanceof Object)
+                let found = 0;
+                for (const osd in flat_tree[host])
                 {
-                    if (!filter_osds_by_tags(orig_tree, flat_tree[item], tags))
-                        delete flat_tree[item];
+                    if (!orig_tree[osd].tags || !orig_tree[osd].tags[tag])
+                        delete flat_tree[host][osd];
+                    else
+                        found++;
+                }
+                if (!found)
+                {
+                    delete flat_tree[host];
                 }
-                else if (!orig_tree[item].tags || !orig_tree[item].tags[tag])
-                    delete flat_tree[item];
             }
         }
-        for (const item in flat_tree)
-            return 1;
-        return 0;
     }
 
     get_affinity_osds(pool_cfg, up_osds, osd_tree)
@@ -1217,11 +1204,9 @@ class Mon
                 {
                     continue;
                 }
-                let pool_tree = osd_tree[pool_cfg.root_node || ''] || {};
-                const failure_domains = pool_cfg.failure_domain instanceof Object
-                    ? [ ...Object.keys(pool_cfg.failure_domain), 'osd' ]
-                    : [ pool_cfg.failure_domain, 'osd' ];
-                pool_tree = LPOptimizer.extract_tree_levels(pool_tree, failure_domains, levels);
+                let pool_tree = osd_tree[pool_cfg.root_node || ''];
+                pool_tree = pool_tree ? pool_tree.children : [];
+                pool_tree = LPOptimizer.flatten_tree(pool_tree, levels, pool_cfg.failure_domain, 'osd');
                 this.filter_osds_by_tags(osd_tree, pool_tree, pool_cfg.osd_tags);
                 // These are for the purpose of building history.osd_sets
                 const real_prev_pgs = [];
@@ -1248,9 +1233,6 @@ class Mon
                     pg_count: pool_cfg.pg_count,
                     pg_size: pool_cfg.pg_size,
                     pg_minsize: pool_cfg.pg_minsize,
-                    hier_sizes: pool_cfg.failure_domain instanceof Object
-                        ? [ ...Object.values(pool_cfg.failure_domain), 1 ]
-                        : null,
                     max_combinations: pool_cfg.max_osd_combinations,
                     ordered: pool_cfg.scheme != 'replicated',
                 };
@@ -1306,7 +1288,7 @@ class Mon
                     } });
                 }
                 LPOptimizer.print_change_stats(optimize_result);
-                const pg_effsize = Math.min(pool_cfg.pg_size, Object.keys(pool_tree).length); // FIXME requires hier support too
+                const pg_effsize = Math.min(pool_cfg.pg_size, Object.keys(pool_tree).length);
                 this.state.pool.stats[pool_id] = {
                     used_raw_tb: (this.state.pool.stats[pool_id]||{}).used_raw_tb || 0,
                     total_raw_tb: optimize_result.space,
@@ -1429,65 +1411,75 @@ class Mon
         }
     }
 
+    derive_osd_stats(st, prev)
+    {
+        const zero_stats = { op: { bps: 0n, iops: 0n, lat: 0n }, subop: { iops: 0n, lat: 0n }, recovery: { bps: 0n, iops: 0n } };
+        const diff = { op_stats: {}, subop_stats: {}, recovery_stats: {} };
+        if (!st || !st.time || prev && (prev.time || this.stat_time/1000) >= st.time)
+        {
+            return diff;
+        }
+        const timediff = BigInt(st.time*1000 - (prev && prev.time*1000 || this.stat_time));
+        for (const op in st.op_stats||{})
+        {
+            const pr = prev && prev.op_stats && prev.op_stats[op];
+            let c = st.op_stats[op];
+            c = { bytes: BigInt(c.bytes||0), usec: BigInt(c.usec||0), count: BigInt(c.count||0) };
+            const b = c.bytes - BigInt(pr && pr.bytes||0);
+            const us = c.usec - BigInt(pr && pr.usec||0);
+            const n = c.count - BigInt(pr && pr.count||0);
+            if (n > 0)
+                diff.op_stats[op] = { ...c, bps: b*1000n/timediff, iops: n*1000n/timediff, lat: us/n };
+        }
+        for (const op in st.subop_stats||{})
+        {
+            const pr = prev && prev.subop_stats && prev.subop_stats[op];
+            let c = st.subop_stats[op];
+            c = { usec: BigInt(c.usec||0), count: BigInt(c.count||0) };
+            const us = c.usec - BigInt(pr && pr.usec||0);
+            const n = c.count - BigInt(pr && pr.count||0);
+            if (n > 0)
+                diff.subop_stats[op] = { ...c, iops: n*1000n/timediff, lat: us/n };
+        }
+        for (const op in st.recovery_stats||{})
+        {
+            const pr = prev && prev.recovery_stats && prev.recovery_stats[op];
+            let c = st.recovery_stats[op];
+            c = { bytes: BigInt(c.bytes||0), count: BigInt(c.count||0) };
+            const b = c.bytes - BigInt(pr && pr.bytes||0);
+            const n = c.count - BigInt(pr && pr.count||0);
+            if (n > 0)
+                diff.recovery_stats[op] = { ...c, bps: b*1000n/timediff, iops: n*1000n/timediff };
+        }
+        return diff;
+    }
+
     sum_op_stats(timestamp, prev_stats)
     {
-        const op_stats = {}, subop_stats = {}, recovery_stats = {};
+        const sum_diff = { op_stats: {}, subop_stats: {}, recovery_stats: {} };
+        if (!prev_stats || prev_stats.timestamp >= timestamp)
+        {
+            return sum_diff;
+        }
+        const tm = BigInt(timestamp - (prev_stats.timestamp || 0));
+        // Sum derived values instead of deriving summed
         for (const osd in this.state.osd.stats)
         {
-            const st = this.state.osd.stats[osd]||{};
-            for (const op in st.op_stats||{})
+            const derived = this.derive_osd_stats(this.state.osd.stats[osd],
+                this.prev_stats && this.prev_stats.osd_stats && this.prev_stats.osd_stats[osd]);
+            for (const type in derived)
             {
-                op_stats[op] = op_stats[op] || { count: 0n, usec: 0n, bytes: 0n };
-                op_stats[op].count += BigInt(st.op_stats[op].count||0);
-                op_stats[op].usec += BigInt(st.op_stats[op].usec||0);
-                op_stats[op].bytes += BigInt(st.op_stats[op].bytes||0);
-            }
-            for (const op in st.subop_stats||{})
-            {
-                subop_stats[op] = subop_stats[op] || { count: 0n, usec: 0n };
-                subop_stats[op].count += BigInt(st.subop_stats[op].count||0);
-                subop_stats[op].usec += BigInt(st.subop_stats[op].usec||0);
-            }
-            for (const op in st.recovery_stats||{})
-            {
-                recovery_stats[op] = recovery_stats[op] || { count: 0n, bytes: 0n };
-                recovery_stats[op].count += BigInt(st.recovery_stats[op].count||0);
-                recovery_stats[op].bytes += BigInt(st.recovery_stats[op].bytes||0);
+                for (const op in derived[type])
+                {
+                    for (const k in derived[type][op])
+                    {
+                        sum_diff[type][op] = sum_diff[type][op] || {};
+                        sum_diff[type][op][k] = (sum_diff[type][op][k] || 0n) + derived[type][op][k];
+                    }
+                }
             }
         }
-        if (prev_stats && prev_stats.timestamp >= timestamp)
-        {
-            prev_stats = null;
-        }
-        const tm = prev_stats ? BigInt(timestamp - prev_stats.timestamp) : 0;
-        for (const op in op_stats)
-        {
-            if (prev_stats && prev_stats.op_stats && prev_stats.op_stats[op])
-            {
-                op_stats[op].bps = (op_stats[op].bytes - prev_stats.op_stats[op].bytes) * 1000n / tm;
-                op_stats[op].iops = (op_stats[op].count - prev_stats.op_stats[op].count) * 1000n / tm;
-                op_stats[op].lat = (op_stats[op].usec - prev_stats.op_stats[op].usec)
-                    / ((op_stats[op].count - prev_stats.op_stats[op].count) || 1n);
-            }
-        }
-        for (const op in subop_stats)
-        {
-            if (prev_stats && prev_stats.subop_stats && prev_stats.subop_stats[op])
-            {
-                subop_stats[op].iops = (subop_stats[op].count - prev_stats.subop_stats[op].count) * 1000n / tm;
-                subop_stats[op].lat = (subop_stats[op].usec - prev_stats.subop_stats[op].usec)
-                    / ((subop_stats[op].count - prev_stats.subop_stats[op].count) || 1n);
-            }
-        }
-        for (const op in recovery_stats)
-        {
-            if (prev_stats && prev_stats.recovery_stats && prev_stats.recovery_stats[op])
-            {
-                recovery_stats[op].bps = (recovery_stats[op].bytes - prev_stats.recovery_stats[op].bytes) * 1000n / tm;
-                recovery_stats[op].iops = (recovery_stats[op].count - prev_stats.recovery_stats[op].count) * 1000n / tm;
-            }
-        }
-        return { op_stats, subop_stats, recovery_stats };
+        return sum_diff;
     }
 
     sum_object_counts()
@@ -1646,7 +1638,8 @@ class Mon
             this.prev_stats ? this.prev_stats.inode_stats : null,
             timestamp, this.prev_stats ? this.prev_stats.timestamp : null
         );
-        this.prev_stats = { timestamp, ...stats, inode_stats };
+        this.prev_stats = { timestamp, inode_stats, osd_stats: { ...this.state.osd.stats } };
+        this.stat_time = Date.now();
         stats.object_counts = object_counts;
         stats.object_bytes = object_bytes;
         stats = this.serialize_bigints(stats);
@@ -1762,13 +1755,14 @@ class Mon
         else if (key_parts[0] === 'osd' && key_parts[1] === 'stats')
         {
             // Recheck OSD tree on OSD addition/deletion
+            const osd_num = key_parts[2];
             if ((!old) != (!kv.value) || old && kv.value && old.size != kv.value.size)
             {
                 this.schedule_recheck();
             }
             // Recheck PGs <osd_out_time> after last OSD statistics report
             this.schedule_next_recheck_at(
-                !this.state.osd.stats[key[2]] ? 0 : this.state.osd.stats[key[2]].time+this.config.osd_out_time
+                !this.state.osd.stats[osd_num] ? 0 : this.state.osd.stats[osd_num].time+this.config.osd_out_time
             );
         }
     }

mon/test-optimize-undersized.js

@@ -36,7 +36,7 @@ const crush_tree = [
     ] },
 ];
 
-const osd_tree = LPOptimizer.extract_tree_levels({ level: -Infinity, children: crush_tree }, [ 1, 3 ], {});
+const osd_tree = LPOptimizer.flatten_tree(crush_tree, {}, 1, 3);
 console.log(osd_tree);
 
 async function run()
@@ -47,32 +47,32 @@ async function run()
     LPOptimizer.print_change_stats(res, false);
     assert(res.space == 0);
     console.log('\nAdding 1st failure domain:');
-    cur_tree['l1_1'] = osd_tree['l1_1'];
+    cur_tree['dom1'] = osd_tree['dom1'];
     res = await LPOptimizer.optimize_change({ prev_pgs: res.int_pgs, osd_tree: cur_tree, pg_size: 3 });
     LPOptimizer.print_change_stats(res, false);
     assert(res.space == 12 && res.total_space == 12);
     console.log('\nAdding 2nd failure domain:');
-    cur_tree['l1_2'] = osd_tree['l1_2'];
+    cur_tree['dom2'] = osd_tree['dom2'];
     res = await LPOptimizer.optimize_change({ prev_pgs: res.int_pgs, osd_tree: cur_tree, pg_size: 3 });
     LPOptimizer.print_change_stats(res, false);
     assert(res.space == 24 && res.total_space == 24);
     console.log('\nAdding 3rd failure domain:');
-    cur_tree['l1_3'] = osd_tree['l1_3'];
+    cur_tree['dom3'] = osd_tree['dom3'];
     res = await LPOptimizer.optimize_change({ prev_pgs: res.int_pgs, osd_tree: cur_tree, pg_size: 3 });
     LPOptimizer.print_change_stats(res, false);
     assert(res.space == 36 && res.total_space == 36);
     console.log('\nRemoving 3rd failure domain:');
-    delete cur_tree['l1_3'];
+    delete cur_tree['dom3'];
     res = await LPOptimizer.optimize_change({ prev_pgs: res.int_pgs, osd_tree: cur_tree, pg_size: 3 });
     LPOptimizer.print_change_stats(res, false);
     assert(res.space == 24 && res.total_space == 24);
     console.log('\nRemoving 2nd failure domain:');
-    delete cur_tree['l1_2'];
+    delete cur_tree['dom2'];
     res = await LPOptimizer.optimize_change({ prev_pgs: res.int_pgs, osd_tree: cur_tree, pg_size: 3 });
     LPOptimizer.print_change_stats(res, false);
     assert(res.space == 12 && res.total_space == 12);
     console.log('\nRemoving 1st failure domain:');
-    delete cur_tree['l1_1'];
+    delete cur_tree['dom1'];
     res = await LPOptimizer.optimize_change({ prev_pgs: res.int_pgs, osd_tree: cur_tree, pg_size: 3 });
     LPOptimizer.print_change_stats(res, false);
     assert(res.space == 0);

mon/test-optimize.js

@@ -108,11 +108,7 @@ async function run()
     LPOptimizer.print_change_stats(res, false);
 
     console.log('\n256 PGs, size=3, failure domain=rack');
-    res = await LPOptimizer.optimize_initial({
-        osd_tree: LPOptimizer.extract_tree_levels({ level: -Infinity, children: crush_tree }, [ 1, 3 ], {}),
-        pg_size: 3,
-        pg_count: 256,
-    });
+    res = await LPOptimizer.optimize_initial({ osd_tree: LPOptimizer.flatten_tree(crush_tree, {}, 1, 3), pg_size: 3, pg_count: 256 });
     LPOptimizer.print_change_stats(res, false);
 }
 

mon/test-random-hier.js

@@ -1,56 +0,0 @@
-// Copyright (c) Vitaliy Filippov, 2019+
-// License: VNPL-1.1 (see README.md for details)
-
-const LPOptimizer = require('./lp-optimizer.js');
-
-const osd_tree = {
-    100: { 110: { 111: 1, 112: 1 }, 120: { 121: 1, 122: 1 } },
-    200: { 210: { 211: 1, 212: 1 }, 220: { 221: 1, 222: 1 } },
-    300: { 310: { 311: 1, 312: 1 }, 320: { 321: 1, 322: 1 } },
-    400: { 410: { 411: 1, 412: 1 }, 420: { 421: 1, 422: 1 } },
-    500: { 510: { 511: 1, 512: 1 }, 520: { 521: 1, 522: 1 } },
-};
-
-const osd_tree2 = {
-    100: { 111: 1, 112: 1, 121: 1, 122: 1 },
-    200: { 211: 1, 212: 1, 221: 1, 222: 1 },
-    300: { 311: 1, 312: 1, 321: 1, 322: 1 },
-    400: { 411: 1, 412: 1, 421: 1, 422: 1 },
-    500: { 511: 1, 512: 1, 521: 1, 522: 1 },
-};
-
-const osd_tree3 = {
-    100: { 111: 1, 112: 1, 121: 1, 122: 1 },
-    200: { 211: 1, 212: 1, 221: 1, 222: 1 },
-    300: { 311: 1, 312: 1, 321: 1, 322: 1 },
-    400: { 411: 1, 412: 1, 421: 1, 422: 1 },
-    500: { 511: 1 },
-};
-
-async function run()
-{
-    let r;
-    console.log(r = LPOptimizer.random_hier_combinations(osd_tree, [ 3, 2, 1 ], 10000, false, true));
-    console.log(r = LPOptimizer.random_hier_combinations(osd_tree2, [ 3, 2 ], 0, false, true));
-    // Will contain 'Z':
-    console.log(r = LPOptimizer.random_combinations(osd_tree2, 6, 0, true));
-    console.log(r = LPOptimizer.extract_tree_levels(
-        { level: 'dc', children: [
-            { level: 'rack', children: [
-                { level: 'host', children: [
-                    { level: 'osd', id: 'OSD5', size: 10 },
-                ] },
-            ] },
-            { level: 'osd', id: 'OSD10', size: 10 },
-        ] },
-        [ 'rack', 'osd' ],
-        { dc: 1, rack: 2, host: 3, osd: 4 }
-    ));
-    if (JSON.stringify(r) != '{"rack1":{"OSD5":10},"rack2":{"OSD10":10}}')
-        throw new Error('extract_tree_levels failed');
-    // should not contain Z:
-    console.log(r = LPOptimizer.random_hier_combinations(osd_tree3, [ 3, 2 ], 0, false, true));
-    console.log('OK');
-}
-
-run().catch(console.error);

patches/VitastorPlugin.pm

@@ -388,8 +388,6 @@ sub unmap_volume
     my ($class, $storeid, $scfg, $volname, $snapname) = @_;
     my $prefix = defined $scfg->{vitastor_prefix} ? $scfg->{vitastor_prefix} : 'pve/';
 
-    return 1 if !$scfg->{vitastor_nbd};
-
     my ($vtype, $name, $vmid) = $class->parse_volname($volname);
     $name .= '@'.$snapname if $snapname;
 
@@ -413,7 +411,7 @@ sub activate_volume
 sub deactivate_volume
 {
     my ($class, $storeid, $scfg, $volname, $snapname, $cache) = @_;
-    $class->unmap_volume($storeid, $scfg, $volname, $snapname);
+    $class->unmap_volume($storeid, $scfg, $volname, $snapname) if $scfg->{vitastor_nbd};
     return 1;
 }
 

patches/cinder-vitastor.py

@@ -50,7 +50,7 @@ from cinder.volume import configuration
 from cinder.volume import driver
 from cinder.volume import volume_utils
 
-VERSION = '0.8.9'
+VERSION = '0.9.2'
 
 LOG = logging.getLogger(__name__)
 

patches/libvirt-9.0-vitastor.diff

@@ -0,0 +1,644 @@
+commit e6f935157944279c2c0634915c3c00feeec748c9
+Author: Vitaliy Filippov <vitalif@yourcmc.ru>
+Date:   Mon Jun 19 00:58:19 2023 +0300
+
+    Add Vitastor support
+
+diff --git a/include/libvirt/libvirt-storage.h b/include/libvirt/libvirt-storage.h
+index aaad4a3..5f5daa8 100644
+--- a/include/libvirt/libvirt-storage.h
++++ b/include/libvirt/libvirt-storage.h
+@@ -326,6 +326,7 @@ typedef enum {
+     VIR_CONNECT_LIST_STORAGE_POOLS_ZFS           = 1 << 17, /* (Since: 1.2.8) */
+     VIR_CONNECT_LIST_STORAGE_POOLS_VSTORAGE      = 1 << 18, /* (Since: 3.1.0) */
+     VIR_CONNECT_LIST_STORAGE_POOLS_ISCSI_DIRECT  = 1 << 19, /* (Since: 5.6.0) */
++    VIR_CONNECT_LIST_STORAGE_POOLS_VITASTOR      = 1 << 20, /* (Since: 5.0.0) */
+ } virConnectListAllStoragePoolsFlags;
+ 
+ int                     virConnectListAllStoragePools(virConnectPtr conn,
+diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c
+index 45965fa..b7c23d3 100644
+--- a/src/conf/domain_conf.c
++++ b/src/conf/domain_conf.c
+@@ -7103,7 +7103,8 @@ virDomainDiskSourceNetworkParse(xmlNodePtr node,
+     src->configFile = virXPathString("string(./config/@file)", ctxt);
+ 
+     if (src->protocol == VIR_STORAGE_NET_PROTOCOL_HTTP ||
+-        src->protocol == VIR_STORAGE_NET_PROTOCOL_HTTPS)
++        src->protocol == VIR_STORAGE_NET_PROTOCOL_HTTPS ||
++        src->protocol == VIR_STORAGE_NET_PROTOCOL_VITASTOR)
+         src->query = virXMLPropString(node, "query");
+ 
+     if (virDomainStorageNetworkParseHosts(node, ctxt, &src->hosts, &src->nhosts) < 0)
+@@ -30121,6 +30122,7 @@ virDomainStorageSourceTranslateSourcePool(virStorageSource *src,
+ 
+     case VIR_STORAGE_POOL_MPATH:
+     case VIR_STORAGE_POOL_RBD:
++    case VIR_STORAGE_POOL_VITASTOR:
+     case VIR_STORAGE_POOL_SHEEPDOG:
+     case VIR_STORAGE_POOL_GLUSTER:
+     case VIR_STORAGE_POOL_LAST:
+diff --git a/src/conf/domain_validate.c b/src/conf/domain_validate.c
+index 5a9bf20..05058b8 100644
+--- a/src/conf/domain_validate.c
++++ b/src/conf/domain_validate.c
+@@ -494,6 +494,7 @@ virDomainDiskDefValidateSourceChainOne(const virStorageSource *src)
+         case VIR_STORAGE_NET_PROTOCOL_RBD:
+             break;
+ 
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+         case VIR_STORAGE_NET_PROTOCOL_NBD:
+         case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+         case VIR_STORAGE_NET_PROTOCOL_GLUSTER:
+@@ -541,7 +542,7 @@ virDomainDiskDefValidateSourceChainOne(const virStorageSource *src)
+         }
+     }
+ 
+-    /* internal snapshots and config files are currently supported only with rbd: */
++    /* internal snapshots are currently supported only with rbd: */
+     if (virStorageSourceGetActualType(src) != VIR_STORAGE_TYPE_NETWORK &&
+         src->protocol != VIR_STORAGE_NET_PROTOCOL_RBD) {
+         if (src->snapshot) {
+@@ -550,11 +551,15 @@ virDomainDiskDefValidateSourceChainOne(const virStorageSource *src)
+                              "only with 'rbd' disks"));
+             return -1;
+         }
+-
++    }
++    /* config files are currently supported only with rbd and vitastor: */
++    if (virStorageSourceGetActualType(src) != VIR_STORAGE_TYPE_NETWORK &&
++        src->protocol != VIR_STORAGE_NET_PROTOCOL_RBD &&
++        src->protocol != VIR_STORAGE_NET_PROTOCOL_VITASTOR) {
+         if (src->configFile) {
+             virReportError(VIR_ERR_XML_ERROR, "%s",
+                            _("<config> element is currently supported "
+-                             "only with 'rbd' disks"));
++                             "only with 'rbd' and 'vitastor' disks"));
+             return -1;
+         }
+     }
+diff --git a/src/conf/schemas/domaincommon.rng b/src/conf/schemas/domaincommon.rng
+index 6cb0a20..8bf7de9 100644
+--- a/src/conf/schemas/domaincommon.rng
++++ b/src/conf/schemas/domaincommon.rng
+@@ -1972,6 +1972,35 @@
+     </element>
+   </define>
+ 
++  <define name="diskSourceNetworkProtocolVitastor">
++    <element name="source">
++      <interleave>
++        <attribute name="protocol">
++          <value>vitastor</value>
++        </attribute>
++        <ref name="diskSourceCommon"/>
++        <optional>
++          <attribute name="name"/>
++        </optional>
++        <optional>
++          <attribute name="query"/>
++        </optional>
++        <zeroOrMore>
++          <ref name="diskSourceNetworkHost"/>
++        </zeroOrMore>
++        <optional>
++          <element name="config">
++            <attribute name="file">
++              <ref name="absFilePath"/>
++            </attribute>
++            <empty/>
++          </element>
++        </optional>
++        <empty/>
++      </interleave>
++    </element>
++  </define>
++
+   <define name="diskSourceNetworkProtocolISCSI">
+     <element name="source">
+       <attribute name="protocol">
+@@ -2264,6 +2293,7 @@
+       <ref name="diskSourceNetworkProtocolSimple"/>
+       <ref name="diskSourceNetworkProtocolVxHS"/>
+       <ref name="diskSourceNetworkProtocolNFS"/>
++      <ref name="diskSourceNetworkProtocolVitastor"/>
+     </choice>
+   </define>
+ 
+diff --git a/src/conf/storage_conf.c b/src/conf/storage_conf.c
+index f5a9636..8339bc4 100644
+--- a/src/conf/storage_conf.c
++++ b/src/conf/storage_conf.c
+@@ -56,7 +56,7 @@ VIR_ENUM_IMPL(virStoragePool,
+               "logical", "disk", "iscsi",
+               "iscsi-direct", "scsi", "mpath",
+               "rbd", "sheepdog", "gluster",
+-              "zfs", "vstorage",
++              "zfs", "vstorage", "vitastor",
+ );
+ 
+ VIR_ENUM_IMPL(virStoragePoolFormatFileSystem,
+@@ -242,6 +242,18 @@ static virStoragePoolTypeInfo poolTypeInfo[] = {
+           .formatToString = virStorageFileFormatTypeToString,
+       }
+     },
++    {.poolType = VIR_STORAGE_POOL_VITASTOR,
++     .poolOptions = {
++         .flags = (VIR_STORAGE_POOL_SOURCE_HOST |
++                   VIR_STORAGE_POOL_SOURCE_NETWORK |
++                   VIR_STORAGE_POOL_SOURCE_NAME),
++      },
++      .volOptions = {
++          .defaultFormat = VIR_STORAGE_FILE_RAW,
++          .formatFromString = virStorageVolumeFormatFromString,
++          .formatToString = virStorageFileFormatTypeToString,
++      }
++    },
+     {.poolType = VIR_STORAGE_POOL_SHEEPDOG,
+      .poolOptions = {
+          .flags = (VIR_STORAGE_POOL_SOURCE_HOST |
+@@ -542,6 +554,11 @@ virStoragePoolDefParseSource(xmlXPathContextPtr ctxt,
+                        _("element 'name' is mandatory for RBD pool"));
+         return -1;
+     }
++    if (pool_type == VIR_STORAGE_POOL_VITASTOR && source->name == NULL) {
++        virReportError(VIR_ERR_XML_ERROR, "%s",
++                       _("element 'name' is mandatory for Vitastor pool"));
++        return -1;
++    }
+ 
+     if (options->formatFromString) {
+         g_autofree char *format = NULL;
+@@ -1132,6 +1149,7 @@ virStoragePoolDefFormatBuf(virBuffer *buf,
+     /* RBD, Sheepdog, Gluster and Iscsi-direct devices are not local block devs nor
+      * files, so they don't have a target */
+     if (def->type != VIR_STORAGE_POOL_RBD &&
++        def->type != VIR_STORAGE_POOL_VITASTOR &&
+         def->type != VIR_STORAGE_POOL_SHEEPDOG &&
+         def->type != VIR_STORAGE_POOL_GLUSTER &&
+         def->type != VIR_STORAGE_POOL_ISCSI_DIRECT) {
+diff --git a/src/conf/storage_conf.h b/src/conf/storage_conf.h
+index fc67957..720c07e 100644
+--- a/src/conf/storage_conf.h
++++ b/src/conf/storage_conf.h
+@@ -103,6 +103,7 @@ typedef enum {
+     VIR_STORAGE_POOL_GLUSTER,  /* Gluster device */
+     VIR_STORAGE_POOL_ZFS,      /* ZFS */
+     VIR_STORAGE_POOL_VSTORAGE, /* Virtuozzo Storage */
++    VIR_STORAGE_POOL_VITASTOR, /* Vitastor */
+ 
+     VIR_STORAGE_POOL_LAST,
+ } virStoragePoolType;
+@@ -454,6 +455,7 @@ VIR_ENUM_DECL(virStoragePartedFs);
+                  VIR_CONNECT_LIST_STORAGE_POOLS_SCSI     | \
+                  VIR_CONNECT_LIST_STORAGE_POOLS_MPATH    | \
+                  VIR_CONNECT_LIST_STORAGE_POOLS_RBD      | \
++                 VIR_CONNECT_LIST_STORAGE_POOLS_VITASTOR | \
+                  VIR_CONNECT_LIST_STORAGE_POOLS_SHEEPDOG | \
+                  VIR_CONNECT_LIST_STORAGE_POOLS_GLUSTER  | \
+                  VIR_CONNECT_LIST_STORAGE_POOLS_ZFS      | \
+diff --git a/src/conf/storage_source_conf.c b/src/conf/storage_source_conf.c
+index cecd7e8..d7b79a4 100644
+--- a/src/conf/storage_source_conf.c
++++ b/src/conf/storage_source_conf.c
+@@ -87,6 +87,7 @@ VIR_ENUM_IMPL(virStorageNetProtocol,
+               "ssh",
+               "vxhs",
+               "nfs",
++              "vitastor",
+ );
+ 
+ 
+@@ -1286,6 +1287,7 @@ virStorageSourceNetworkDefaultPort(virStorageNetProtocol protocol)
+         case VIR_STORAGE_NET_PROTOCOL_GLUSTER:
+             return 24007;
+ 
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+         case VIR_STORAGE_NET_PROTOCOL_RBD:
+             /* we don't provide a default for RBD */
+             return 0;
+diff --git a/src/conf/storage_source_conf.h b/src/conf/storage_source_conf.h
+index 14a6825..eb4acac 100644
+--- a/src/conf/storage_source_conf.h
++++ b/src/conf/storage_source_conf.h
+@@ -128,6 +128,7 @@ typedef enum {
+     VIR_STORAGE_NET_PROTOCOL_SSH,
+     VIR_STORAGE_NET_PROTOCOL_VXHS,
+     VIR_STORAGE_NET_PROTOCOL_NFS,
++    VIR_STORAGE_NET_PROTOCOL_VITASTOR,
+ 
+     VIR_STORAGE_NET_PROTOCOL_LAST
+ } virStorageNetProtocol;
+diff --git a/src/conf/virstorageobj.c b/src/conf/virstorageobj.c
+index e6c187e..035b423 100644
+--- a/src/conf/virstorageobj.c
++++ b/src/conf/virstorageobj.c
+@@ -1433,6 +1433,7 @@ virStoragePoolObjSourceFindDuplicateCb(const void *payload,
+             return 1;
+         break;
+ 
++    case VIR_STORAGE_POOL_VITASTOR:
+     case VIR_STORAGE_POOL_ISCSI_DIRECT:
+     case VIR_STORAGE_POOL_RBD:
+     case VIR_STORAGE_POOL_LAST:
+@@ -1918,6 +1919,8 @@ virStoragePoolObjMatch(virStoragePoolObj *obj,
+                (obj->def->type == VIR_STORAGE_POOL_MPATH))   ||
+               (MATCH(VIR_CONNECT_LIST_STORAGE_POOLS_RBD) &&
+                (obj->def->type == VIR_STORAGE_POOL_RBD))     ||
++              (MATCH(VIR_CONNECT_LIST_STORAGE_POOLS_VITASTOR) &&
++               (obj->def->type == VIR_STORAGE_POOL_VITASTOR)) ||
+               (MATCH(VIR_CONNECT_LIST_STORAGE_POOLS_SHEEPDOG) &&
+                (obj->def->type == VIR_STORAGE_POOL_SHEEPDOG)) ||
+               (MATCH(VIR_CONNECT_LIST_STORAGE_POOLS_GLUSTER) &&
+diff --git a/src/libvirt-storage.c b/src/libvirt-storage.c
+index 8490034..ab2cdaa 100644
+--- a/src/libvirt-storage.c
++++ b/src/libvirt-storage.c
+@@ -94,6 +94,7 @@ virStoragePoolGetConnect(virStoragePoolPtr pool)
+  * VIR_CONNECT_LIST_STORAGE_POOLS_SCSI
+  * VIR_CONNECT_LIST_STORAGE_POOLS_MPATH
+  * VIR_CONNECT_LIST_STORAGE_POOLS_RBD
++ * VIR_CONNECT_LIST_STORAGE_POOLS_VITASTOR
+  * VIR_CONNECT_LIST_STORAGE_POOLS_SHEEPDOG
+  * VIR_CONNECT_LIST_STORAGE_POOLS_GLUSTER
+  * VIR_CONNECT_LIST_STORAGE_POOLS_ZFS
+diff --git a/src/libxl/libxl_conf.c b/src/libxl/libxl_conf.c
+index 17ac880..59711b5 100644
+--- a/src/libxl/libxl_conf.c
++++ b/src/libxl/libxl_conf.c
+@@ -970,6 +970,7 @@ libxlMakeNetworkDiskSrcStr(virStorageSource *src,
+     case VIR_STORAGE_NET_PROTOCOL_SSH:
+     case VIR_STORAGE_NET_PROTOCOL_VXHS:
+     case VIR_STORAGE_NET_PROTOCOL_NFS:
++    case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+     case VIR_STORAGE_NET_PROTOCOL_LAST:
+     case VIR_STORAGE_NET_PROTOCOL_NONE:
+         virReportError(VIR_ERR_NO_SUPPORT,
+diff --git a/src/libxl/xen_xl.c b/src/libxl/xen_xl.c
+index 6919325..55ffc32 100644
+--- a/src/libxl/xen_xl.c
++++ b/src/libxl/xen_xl.c
+@@ -1445,6 +1445,7 @@ xenFormatXLDiskSrcNet(virStorageSource *src)
+     case VIR_STORAGE_NET_PROTOCOL_SSH:
+     case VIR_STORAGE_NET_PROTOCOL_VXHS:
+     case VIR_STORAGE_NET_PROTOCOL_NFS:
++    case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+     case VIR_STORAGE_NET_PROTOCOL_LAST:
+     case VIR_STORAGE_NET_PROTOCOL_NONE:
+         virReportError(VIR_ERR_NO_SUPPORT,
+diff --git a/src/qemu/qemu_block.c b/src/qemu/qemu_block.c
+index e865aa1..40162af 100644
+--- a/src/qemu/qemu_block.c
++++ b/src/qemu/qemu_block.c
+@@ -604,6 +604,38 @@ qemuBlockStorageSourceGetRBDProps(virStorageSource *src,
+ }
+ 
+ 
++static virJSONValue *
++qemuBlockStorageSourceGetVitastorProps(virStorageSource *src)
++{
++    virJSONValue *ret = NULL;
++    virStorageNetHostDef *host;
++    size_t i;
++    g_auto(virBuffer) buf = VIR_BUFFER_INITIALIZER;
++    g_autofree char *etcd = NULL;
++
++    for (i = 0; i < src->nhosts; i++) {
++        host = src->hosts + i;
++        if ((virStorageNetHostTransport)host->transport != VIR_STORAGE_NET_HOST_TRANS_TCP) {
++            return NULL;
++        }
++        virBufferAsprintf(&buf, i > 0 ? ",%s:%u" : "%s:%u", host->name, host->port);
++    }
++    if (src->nhosts > 0) {
++        etcd = virBufferContentAndReset(&buf);
++    }
++
++    if (virJSONValueObjectAdd(&ret,
++                              "S:etcd-host", etcd,
++                              "S:etcd-prefix", src->query,
++                              "S:config-path", src->configFile,
++                              "s:image", src->path,
++                              NULL) < 0)
++        return NULL;
++
++    return ret;
++}
++
++
+ static virJSONValue *
+ qemuBlockStorageSourceGetSheepdogProps(virStorageSource *src)
+ {
+@@ -917,6 +949,12 @@ qemuBlockStorageSourceGetBackendProps(virStorageSource *src,
+                 return NULL;
+             break;
+ 
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
++            driver = "vitastor";
++            if (!(fileprops = qemuBlockStorageSourceGetVitastorProps(src)))
++                return NULL;
++            break;
++
+         case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+             driver = "sheepdog";
+             if (!(fileprops = qemuBlockStorageSourceGetSheepdogProps(src)))
+@@ -1860,6 +1898,7 @@ qemuBlockGetBackingStoreString(virStorageSource *src,
+ 
+             case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+             case VIR_STORAGE_NET_PROTOCOL_RBD:
++            case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+             case VIR_STORAGE_NET_PROTOCOL_VXHS:
+             case VIR_STORAGE_NET_PROTOCOL_NFS:
+             case VIR_STORAGE_NET_PROTOCOL_SSH:
+@@ -2242,6 +2281,12 @@ qemuBlockStorageSourceCreateGetStorageProps(virStorageSource *src,
+                 return -1;
+             break;
+ 
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
++            driver = "vitastor";
++            if (!(location = qemuBlockStorageSourceGetVitastorProps(src)))
++                return -1;
++            break;
++
+         case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+             driver = "sheepdog";
+             if (!(location = qemuBlockStorageSourceGetSheepdogProps(src)))
+diff --git a/src/qemu/qemu_domain.c b/src/qemu/qemu_domain.c
+index 2eb5653..60ee82d 100644
+--- a/src/qemu/qemu_domain.c
++++ b/src/qemu/qemu_domain.c
+@@ -4958,7 +4958,8 @@ qemuDomainValidateStorageSource(virStorageSource *src,
+     if (src->query &&
+         (actualType != VIR_STORAGE_TYPE_NETWORK ||
+          (src->protocol != VIR_STORAGE_NET_PROTOCOL_HTTPS &&
+-          src->protocol != VIR_STORAGE_NET_PROTOCOL_HTTP))) {
++          src->protocol != VIR_STORAGE_NET_PROTOCOL_HTTP &&
++          src->protocol != VIR_STORAGE_NET_PROTOCOL_VITASTOR))) {
+         virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
+                        _("query is supported only with HTTP(S) protocols"));
+         return -1;
+@@ -10129,6 +10130,7 @@ qemuDomainPrepareStorageSourceTLS(virStorageSource *src,
+         break;
+ 
+     case VIR_STORAGE_NET_PROTOCOL_RBD:
++    case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+     case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+     case VIR_STORAGE_NET_PROTOCOL_GLUSTER:
+     case VIR_STORAGE_NET_PROTOCOL_ISCSI:
+diff --git a/src/qemu/qemu_snapshot.c b/src/qemu/qemu_snapshot.c
+index b841680..a6be771 100644
+--- a/src/qemu/qemu_snapshot.c
++++ b/src/qemu/qemu_snapshot.c
+@@ -373,6 +373,7 @@ qemuSnapshotPrepareDiskExternalInactive(virDomainSnapshotDiskDef *snapdisk,
+         case VIR_STORAGE_NET_PROTOCOL_NONE:
+         case VIR_STORAGE_NET_PROTOCOL_NBD:
+         case VIR_STORAGE_NET_PROTOCOL_RBD:
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+         case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+         case VIR_STORAGE_NET_PROTOCOL_GLUSTER:
+         case VIR_STORAGE_NET_PROTOCOL_ISCSI:
+@@ -578,6 +579,7 @@ qemuSnapshotPrepareDiskInternal(virDomainDiskDef *disk,
+         case VIR_STORAGE_NET_PROTOCOL_NONE:
+         case VIR_STORAGE_NET_PROTOCOL_NBD:
+         case VIR_STORAGE_NET_PROTOCOL_RBD:
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+         case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+         case VIR_STORAGE_NET_PROTOCOL_GLUSTER:
+         case VIR_STORAGE_NET_PROTOCOL_ISCSI:
+diff --git a/src/storage/storage_driver.c b/src/storage/storage_driver.c
+index d90c1c9..e853457 100644
+--- a/src/storage/storage_driver.c
++++ b/src/storage/storage_driver.c
+@@ -1627,6 +1627,7 @@ storageVolLookupByPathCallback(virStoragePoolObj *obj,
+ 
+         case VIR_STORAGE_POOL_GLUSTER:
+         case VIR_STORAGE_POOL_RBD:
++        case VIR_STORAGE_POOL_VITASTOR:
+         case VIR_STORAGE_POOL_SHEEPDOG:
+         case VIR_STORAGE_POOL_ZFS:
+         case VIR_STORAGE_POOL_LAST:
+diff --git a/src/storage_file/storage_source_backingstore.c b/src/storage_file/storage_source_backingstore.c
+index e48ae72..2017ccc 100644
+--- a/src/storage_file/storage_source_backingstore.c
++++ b/src/storage_file/storage_source_backingstore.c
+@@ -284,6 +284,75 @@ virStorageSourceParseRBDColonString(const char *rbdstr,
+ }
+ 
+ 
++static int
++virStorageSourceParseVitastorColonString(const char *colonstr,
++                                         virStorageSource *src)
++{
++    char *p, *e, *next;
++    g_autofree char *options = NULL;
++
++    /* optionally skip the "vitastor:" prefix if provided */
++    if (STRPREFIX(colonstr, "vitastor:"))
++        colonstr += strlen("vitastor:");
++
++    options = g_strdup(colonstr);
++
++    p = options;
++    while (*p) {
++        /* find : delimiter or end of string */
++        for (e = p; *e && *e != ':'; ++e) {
++            if (*e == '\\') {
++                e++;
++                if (*e == '\0')
++                    break;
++            }
++        }
++        if (*e == '\0') {
++            next = e;    /* last kv pair */
++        } else {
++            next = e + 1;
++            *e = '\0';
++        }
++
++        if (STRPREFIX(p, "image=")) {
++            src->path = g_strdup(p + strlen("image="));
++        } else if (STRPREFIX(p, "etcd-prefix=")) {
++            src->query = g_strdup(p + strlen("etcd-prefix="));
++        } else if (STRPREFIX(p, "config-path=")) {
++            src->configFile = g_strdup(p + strlen("config-path="));
++        } else if (STRPREFIX(p, "etcd-host=")) {
++            char *h, *sep;
++
++            h = p + strlen("etcd-host=");
++            while (h < e) {
++                for (sep = h; sep < e; ++sep) {
++                    if (*sep == '\\' && (sep[1] == ',' ||
++                                         sep[1] == ';' ||
++                                         sep[1] == ' ')) {
++                        *sep = '\0';
++                        sep += 2;
++                        break;
++                    }
++                }
++
++                if (virStorageSourceRBDAddHost(src, h) < 0)
++                    return -1;
++
++                h = sep;
++            }
++        }
++
++        p = next;
++    }
++
++    if (!src->path) {
++        return -1;
++    }
++
++    return 0;
++}
++
++
+ static int
+ virStorageSourceParseNBDColonString(const char *nbdstr,
+                                     virStorageSource *src)
+@@ -396,6 +465,11 @@ virStorageSourceParseBackingColon(virStorageSource *src,
+             return -1;
+         break;
+ 
++    case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
++        if (virStorageSourceParseVitastorColonString(path, src) < 0)
++            return -1;
++        break;
++
+     case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+     case VIR_STORAGE_NET_PROTOCOL_LAST:
+     case VIR_STORAGE_NET_PROTOCOL_NONE:
+@@ -984,6 +1058,54 @@ virStorageSourceParseBackingJSONRBD(virStorageSource *src,
+     return 0;
+ }
+ 
++static int
++virStorageSourceParseBackingJSONVitastor(virStorageSource *src,
++                                         virJSONValue *json,
++                                         const char *jsonstr G_GNUC_UNUSED,
++                                         int opaque G_GNUC_UNUSED)
++{
++    const char *filename;
++    const char *image = virJSONValueObjectGetString(json, "image");
++    const char *conf = virJSONValueObjectGetString(json, "config-path");
++    const char *etcd_prefix = virJSONValueObjectGetString(json, "etcd-prefix");
++    virJSONValue *servers = virJSONValueObjectGetArray(json, "server");
++    size_t nservers;
++    size_t i;
++
++    src->type = VIR_STORAGE_TYPE_NETWORK;
++    src->protocol = VIR_STORAGE_NET_PROTOCOL_VITASTOR;
++
++    /* legacy syntax passed via 'filename' option */
++    if ((filename = virJSONValueObjectGetString(json, "filename")))
++        return virStorageSourceParseVitastorColonString(filename, src);
++
++    if (!image) {
++        virReportError(VIR_ERR_INVALID_ARG, "%s",
++                       _("missing image name in Vitastor backing volume "
++                         "JSON specification"));
++        return -1;
++    }
++
++    src->path = g_strdup(image);
++    src->configFile = g_strdup(conf);
++    src->query = g_strdup(etcd_prefix);
++
++    if (servers) {
++        nservers = virJSONValueArraySize(servers);
++
++        src->hosts = g_new0(virStorageNetHostDef, nservers);
++        src->nhosts = nservers;
++
++        for (i = 0; i < nservers; i++) {
++            if (virStorageSourceParseBackingJSONInetSocketAddress(src->hosts + i,
++                                                                  virJSONValueArrayGet(servers, i)) < 0)
++                return -1;
++        }
++    }
++
++    return 0;
++}
++
+ static int
+ virStorageSourceParseBackingJSONRaw(virStorageSource *src,
+                                     virJSONValue *json,
+@@ -1162,6 +1284,7 @@ static const struct virStorageSourceJSONDriverParser jsonParsers[] = {
+     {"sheepdog", false, virStorageSourceParseBackingJSONSheepdog, 0},
+     {"ssh", false, virStorageSourceParseBackingJSONSSH, 0},
+     {"rbd", false, virStorageSourceParseBackingJSONRBD, 0},
++    {"vitastor", false, virStorageSourceParseBackingJSONVitastor, 0},
+     {"raw", true, virStorageSourceParseBackingJSONRaw, 0},
+     {"nfs", false, virStorageSourceParseBackingJSONNFS, 0},
+     {"vxhs", false, virStorageSourceParseBackingJSONVxHS, 0},
+diff --git a/src/test/test_driver.c b/src/test/test_driver.c
+index bd6f063..cce34e1 100644
+--- a/src/test/test_driver.c
++++ b/src/test/test_driver.c
+@@ -7338,6 +7338,7 @@ testStorageVolumeTypeForPool(int pooltype)
+     case VIR_STORAGE_POOL_ISCSI_DIRECT:
+     case VIR_STORAGE_POOL_GLUSTER:
+     case VIR_STORAGE_POOL_RBD:
++    case VIR_STORAGE_POOL_VITASTOR:
+         return VIR_STORAGE_VOL_NETWORK;
+     case VIR_STORAGE_POOL_LOGICAL:
+     case VIR_STORAGE_POOL_DISK:
+diff --git a/tests/storagepoolcapsschemadata/poolcaps-fs.xml b/tests/storagepoolcapsschemadata/poolcaps-fs.xml
+index eee75af..8bd0a57 100644
+--- a/tests/storagepoolcapsschemadata/poolcaps-fs.xml
++++ b/tests/storagepoolcapsschemadata/poolcaps-fs.xml
+@@ -204,4 +204,11 @@
+       </enum>
+     </volOptions>
+   </pool>
++  <pool type='vitastor' supported='no'>
++    <volOptions>
++      <defaultFormat type='raw'/>
++      <enum name='targetFormatType'>
++      </enum>
++    </volOptions>
++  </pool>
+ </storagepoolCapabilities>
+diff --git a/tests/storagepoolcapsschemadata/poolcaps-full.xml b/tests/storagepoolcapsschemadata/poolcaps-full.xml
+index 805950a..852df0d 100644
+--- a/tests/storagepoolcapsschemadata/poolcaps-full.xml
++++ b/tests/storagepoolcapsschemadata/poolcaps-full.xml
+@@ -204,4 +204,11 @@
+       </enum>
+     </volOptions>
+   </pool>
++  <pool type='vitastor' supported='yes'>
++    <volOptions>
++      <defaultFormat type='raw'/>
++      <enum name='targetFormatType'>
++      </enum>
++    </volOptions>
++  </pool>
+ </storagepoolCapabilities>
+diff --git a/tests/storagepoolxml2argvtest.c b/tests/storagepoolxml2argvtest.c
+index e8e40d6..db55fe5 100644
+--- a/tests/storagepoolxml2argvtest.c
++++ b/tests/storagepoolxml2argvtest.c
+@@ -65,6 +65,7 @@ testCompareXMLToArgvFiles(bool shouldFail,
+     case VIR_STORAGE_POOL_GLUSTER:
+     case VIR_STORAGE_POOL_ZFS:
+     case VIR_STORAGE_POOL_VSTORAGE:
++    case VIR_STORAGE_POOL_VITASTOR:
+     case VIR_STORAGE_POOL_LAST:
+     default:
+         VIR_TEST_DEBUG("pool type '%s' has no xml2argv test", defTypeStr);
+diff --git a/tools/virsh-pool.c b/tools/virsh-pool.c
+index 8a98c6a..4b1bbd4 100644
+--- a/tools/virsh-pool.c
++++ b/tools/virsh-pool.c
+@@ -1221,6 +1221,9 @@ cmdPoolList(vshControl *ctl, const vshCmd *cmd G_GNUC_UNUSED)
+             case VIR_STORAGE_POOL_VSTORAGE:
+                 flags |= VIR_CONNECT_LIST_STORAGE_POOLS_VSTORAGE;
+                 break;
++            case VIR_STORAGE_POOL_VITASTOR:
++                flags |= VIR_CONNECT_LIST_STORAGE_POOLS_VITASTOR;
++                break;
+             case VIR_STORAGE_POOL_LAST:
+                 break;
+             }

patches/pve-qemu-8.0-vitastor.patch

@@ -0,0 +1,190 @@
+diff --git a/block/meson.build b/block/meson.build
+index 382bec0e7d..af6207dbce 100644
+--- a/block/meson.build
++++ b/block/meson.build
+@@ -114,6 +114,7 @@ foreach m : [
+   [libnfs, 'nfs', files('nfs.c')],
+   [libssh, 'ssh', files('ssh.c')],
+   [rbd, 'rbd', files('rbd.c')],
++  [vitastor, 'vitastor', files('vitastor.c')],
+ ]
+   if m[0].found()
+     module_ss = ss.source_set()
+diff --git a/meson.build b/meson.build
+index c44d05a13f..ebedb42843 100644
+--- a/meson.build
++++ b/meson.build
+@@ -1028,6 +1028,26 @@ if not get_option('rbd').auto() or have_block
+   endif
+ endif
+ 
++vitastor = not_found
++if not get_option('vitastor').auto() or have_block
++  libvitastor_client = cc.find_library('vitastor_client', has_headers: ['vitastor_c.h'],
++    required: get_option('vitastor'), kwargs: static_kwargs)
++  if libvitastor_client.found()
++    if cc.links('''
++      #include <vitastor_c.h>
++      int main(void) {
++        vitastor_c_create_qemu(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0);
++        return 0;
++      }''', dependencies: libvitastor_client)
++      vitastor = declare_dependency(dependencies: libvitastor_client)
++    elif get_option('vitastor').enabled()
++      error('could not link libvitastor_client')
++    else
++      warning('could not link libvitastor_client, disabling')
++    endif
++  endif
++endif
++
+ glusterfs = not_found
+ glusterfs_ftruncate_has_stat = false
+ glusterfs_iocb_has_stat = false
+@@ -1882,6 +1902,7 @@ endif
+ config_host_data.set('CONFIG_OPENGL', opengl.found())
+ config_host_data.set('CONFIG_PROFILER', get_option('profiler'))
+ config_host_data.set('CONFIG_RBD', rbd.found())
++config_host_data.set('CONFIG_VITASTOR', vitastor.found())
+ config_host_data.set('CONFIG_RDMA', rdma.found())
+ config_host_data.set('CONFIG_SDL', sdl.found())
+ config_host_data.set('CONFIG_SDL_IMAGE', sdl_image.found())
+@@ -4020,6 +4041,7 @@ if spice_protocol.found()
+   summary_info += {'  spice server support': spice}
+ endif
+ summary_info += {'rbd support':       rbd}
++summary_info += {'vitastor support':  vitastor}
+ summary_info += {'smartcard support': cacard}
+ summary_info += {'U2F support':       u2f}
+ summary_info += {'libusb':            libusb}
+diff --git a/meson_options.txt b/meson_options.txt
+index fc9447d267..c4ac55c283 100644
+--- a/meson_options.txt
++++ b/meson_options.txt
+@@ -173,6 +173,8 @@ option('lzo', type : 'feature', value : 'auto',
+        description: 'lzo compression support')
+ option('rbd', type : 'feature', value : 'auto',
+        description: 'Ceph block device driver')
++option('vitastor', type : 'feature', value : 'auto',
++       description: 'Vitastor block device driver')
+ option('opengl', type : 'feature', value : 'auto',
+        description: 'OpenGL support')
+ option('rdma', type : 'feature', value : 'auto',
+diff --git a/qapi/block-core.json b/qapi/block-core.json
+index c05ad0c07e..f5eb701604 100644
+--- a/qapi/block-core.json
++++ b/qapi/block-core.json
+@@ -3308,7 +3308,7 @@
+             'raw', 'rbd',
+             { 'name': 'replication', 'if': 'CONFIG_REPLICATION' },
+             'pbs',
+-            'ssh', 'throttle', 'vdi', 'vhdx',
++            'ssh', 'throttle', 'vdi', 'vhdx', 'vitastor',
+             { 'name': 'virtio-blk-vfio-pci', 'if': 'CONFIG_BLKIO' },
+             { 'name': 'virtio-blk-vhost-user', 'if': 'CONFIG_BLKIO' },
+             { 'name': 'virtio-blk-vhost-vdpa', 'if': 'CONFIG_BLKIO' },
+@@ -4338,6 +4338,28 @@
+             '*key-secret': 'str',
+             '*server': ['InetSocketAddressBase'] } }
+ 
++##
++# @BlockdevOptionsVitastor:
++#
++# Driver specific block device options for vitastor
++#
++# @image:       Image name
++# @inode:       Inode number
++# @pool:        Pool ID
++# @size:        Desired image size in bytes
++# @config-path: Path to Vitastor configuration
++# @etcd-host:   etcd connection address(es)
++# @etcd-prefix: etcd key/value prefix
++##
++{ 'struct': 'BlockdevOptionsVitastor',
++  'data': { '*inode': 'uint64',
++            '*pool': 'uint64',
++            '*size': 'uint64',
++            '*image': 'str',
++            '*config-path': 'str',
++            '*etcd-host': 'str',
++            '*etcd-prefix': 'str' } }
++
+ ##
+ # @ReplicationMode:
+ #
+@@ -4787,6 +4809,7 @@
+       'throttle':   'BlockdevOptionsThrottle',
+       'vdi':        'BlockdevOptionsGenericFormat',
+       'vhdx':       'BlockdevOptionsGenericFormat',
++      'vitastor':   'BlockdevOptionsVitastor',
+       'virtio-blk-vfio-pci':
+                     { 'type': 'BlockdevOptionsVirtioBlkVfioPci',
+                       'if': 'CONFIG_BLKIO' },
+@@ -5187,6 +5210,17 @@
+             '*cluster-size' :   'size',
+             '*encrypt' :        'RbdEncryptionCreateOptions' } }
+ 
++##
++# @BlockdevCreateOptionsVitastor:
++#
++# Driver specific image creation options for Vitastor.
++#
++# @size: Size of the virtual disk in bytes
++##
++{ 'struct': 'BlockdevCreateOptionsVitastor',
++  'data': { 'location':         'BlockdevOptionsVitastor',
++            'size':             'size' } }
++
+ ##
+ # @BlockdevVmdkSubformat:
+ #
+@@ -5385,6 +5419,7 @@
+       'ssh':            'BlockdevCreateOptionsSsh',
+       'vdi':            'BlockdevCreateOptionsVdi',
+       'vhdx':           'BlockdevCreateOptionsVhdx',
++      'vitastor':       'BlockdevCreateOptionsVitastor',
+       'vmdk':           'BlockdevCreateOptionsVmdk',
+       'vpc':            'BlockdevCreateOptionsVpc'
+   } }
+diff --git a/scripts/ci/org.centos/stream/8/x86_64/configure b/scripts/ci/org.centos/stream/8/x86_64/configure
+index 6e8983f39c..1b0b9fcf3e 100755
+--- a/scripts/ci/org.centos/stream/8/x86_64/configure
++++ b/scripts/ci/org.centos/stream/8/x86_64/configure
+@@ -32,7 +32,7 @@
+ --with-git=meson \
+ --with-git-submodules=update \
+ --target-list="x86_64-softmmu" \
+---block-drv-rw-whitelist="qcow2,raw,file,host_device,nbd,iscsi,rbd,blkdebug,luks,null-co,nvme,copy-on-read,throttle,gluster" \
++--block-drv-rw-whitelist="qcow2,raw,file,host_device,nbd,iscsi,rbd,vitastor,blkdebug,luks,null-co,nvme,copy-on-read,throttle,gluster" \
+ --audio-drv-list="" \
+ --block-drv-ro-whitelist="vmdk,vhdx,vpc,https,ssh" \
+ --with-coroutine=ucontext \
+@@ -179,6 +179,7 @@
+ --enable-opengl \
+ --enable-pie \
+ --enable-rbd \
++--enable-vitastor \
+ --enable-rdma \
+ --enable-seccomp \
+ --enable-snappy \
+diff --git a/scripts/meson-buildoptions.sh b/scripts/meson-buildoptions.sh
+index 009fab1515..95914e6ebc 100644
+--- a/scripts/meson-buildoptions.sh
++++ b/scripts/meson-buildoptions.sh
+@@ -144,6 +144,7 @@ meson_options_help() {
+   printf "%s\n" '  qed             qed image format support'
+   printf "%s\n" '  qga-vss         build QGA VSS support (broken with MinGW)'
+   printf "%s\n" '  rbd             Ceph block device driver'
++  printf "%s\n" '  vitastor        Vitastor block device driver'
+   printf "%s\n" '  rdma            Enable RDMA-based migration'
+   printf "%s\n" '  replication     replication support'
+   printf "%s\n" '  sdl             SDL user interface'
+@@ -392,6 +393,8 @@ _meson_option_parse() {
+     --disable-qom-cast-debug) printf "%s" -Dqom_cast_debug=false ;;
+     --enable-rbd) printf "%s" -Drbd=enabled ;;
+     --disable-rbd) printf "%s" -Drbd=disabled ;;
++    --enable-vitastor) printf "%s" -Dvitastor=enabled ;;
++    --disable-vitastor) printf "%s" -Dvitastor=disabled ;;
+     --enable-rdma) printf "%s" -Drdma=enabled ;;
+     --disable-rdma) printf "%s" -Drdma=disabled ;;
+     --enable-replication) printf "%s" -Dreplication=enabled ;;

rpm/build-tarball.sh

@@ -24,4 +24,4 @@ rm fio
 mv fio-copy fio
 FIO=`rpm -qi fio | perl -e 'while(<>) { /^Epoch[\s:]+(\S+)/ && print "$1:"; /^Version[\s:]+(\S+)/ && print $1; /^Release[\s:]+(\S+)/ && print "-$1"; }'`
 perl -i -pe 's/(Requires:\s*fio)([^\n]+)?/$1 = '$FIO'/' $VITASTOR/rpm/vitastor-el$EL.spec
-tar --transform 's#^#vitastor-0.8.9/#' --exclude 'rpm/*.rpm' -czf $VITASTOR/../vitastor-0.8.9$(rpm --eval '%dist').tar.gz *
+tar --transform 's#^#vitastor-0.9.2/#' --exclude 'rpm/*.rpm' -czf $VITASTOR/../vitastor-0.9.2$(rpm --eval '%dist').tar.gz *

rpm/vitastor-el7.Dockerfile

@@ -35,7 +35,7 @@ ADD . /root/vitastor
 RUN set -e; \
     cd /root/vitastor/rpm; \
     sh build-tarball.sh; \
-    cp /root/vitastor-0.8.9.el7.tar.gz ~/rpmbuild/SOURCES; \
+    cp /root/vitastor-0.9.2.el7.tar.gz ~/rpmbuild/SOURCES; \
     cp vitastor-el7.spec ~/rpmbuild/SPECS/vitastor.spec; \
     cd ~/rpmbuild/SPECS/; \
     rpmbuild -ba vitastor.spec; \

rpm/vitastor-el7.spec

@@ -1,11 +1,11 @@
 Name:           vitastor
-Version:        0.8.9
+Version:        0.9.2
 Release:        1%{?dist}
 Summary:        Vitastor, a fast software-defined clustered block storage
 
 License:        Vitastor Network Public License 1.1
 URL:            https://vitastor.io/
-Source0:        vitastor-0.8.9.el7.tar.gz
+Source0:        vitastor-0.9.2.el7.tar.gz
 
 BuildRequires:  liburing-devel >= 0.6
 BuildRequires:  gperftools-devel

rpm/vitastor-el8.Dockerfile

@@ -35,7 +35,7 @@ ADD . /root/vitastor
 RUN set -e; \
     cd /root/vitastor/rpm; \
     sh build-tarball.sh; \
-    cp /root/vitastor-0.8.9.el8.tar.gz ~/rpmbuild/SOURCES; \
+    cp /root/vitastor-0.9.2.el8.tar.gz ~/rpmbuild/SOURCES; \
     cp vitastor-el8.spec ~/rpmbuild/SPECS/vitastor.spec; \
     cd ~/rpmbuild/SPECS/; \
     rpmbuild -ba vitastor.spec; \

rpm/vitastor-el8.spec

@@ -1,11 +1,11 @@
 Name:           vitastor
-Version:        0.8.9
+Version:        0.9.2
 Release:        1%{?dist}
 Summary:        Vitastor, a fast software-defined clustered block storage
 
 License:        Vitastor Network Public License 1.1
 URL:            https://vitastor.io/
-Source0:        vitastor-0.8.9.el8.tar.gz
+Source0:        vitastor-0.9.2.el8.tar.gz
 
 BuildRequires:  liburing-devel >= 0.6
 BuildRequires:  gperftools-devel

rpm/vitastor-el9.Dockerfile

@@ -18,7 +18,7 @@ ADD . /root/vitastor
 RUN set -e; \
     cd /root/vitastor/rpm; \
     sh build-tarball.sh; \
-    cp /root/vitastor-0.8.9.el9.tar.gz ~/rpmbuild/SOURCES; \
+    cp /root/vitastor-0.9.2.el9.tar.gz ~/rpmbuild/SOURCES; \
     cp vitastor-el9.spec ~/rpmbuild/SPECS/vitastor.spec; \
     cd ~/rpmbuild/SPECS/; \
     rpmbuild -ba vitastor.spec; \

rpm/vitastor-el9.spec

@@ -1,11 +1,11 @@
 Name:           vitastor
-Version:        0.8.9
+Version:        0.9.2
 Release:        1%{?dist}
 Summary:        Vitastor, a fast software-defined clustered block storage
 
 License:        Vitastor Network Public License 1.1
 URL:            https://vitastor.io/
-Source0:        vitastor-0.8.9.el9.tar.gz
+Source0:        vitastor-0.9.2.el9.tar.gz
 
 BuildRequires:  liburing-devel >= 0.6
 BuildRequires:  gperftools-devel
@@ -73,7 +73,7 @@ Vitastor library headers for development.
 Summary:        Vitastor - fio drivers
 Group:          Development/Libraries
 Requires:       vitastor-client = %{version}-%{release}
-Requires:       fio = 3.27-7.el9
+Requires:       fio = 3.27-8.el9
 
 
 %description -n vitastor-fio

src/CMakeLists.txt

@@ -16,7 +16,7 @@ if("${CMAKE_INSTALL_PREFIX}" MATCHES "^/usr/local/?$")
 	set(CMAKE_INSTALL_RPATH "${CMAKE_INSTALL_PREFIX}/${CMAKE_INSTALL_LIBDIR}")
 endif()
 
-add_definitions(-DVERSION="0.8.9")
+add_definitions(-DVERSION="0.9.2")
 add_definitions(-Wall -Wno-sign-compare -Wno-comment -Wno-parentheses -Wno-pointer-arith -fdiagnostics-color=always -I ${CMAKE_SOURCE_DIR}/src)
 if (${WITH_ASAN})
 	add_definitions(-fsanitize=address -fno-omit-frame-pointer)
@@ -111,7 +111,7 @@ target_compile_options(vitastor_common PUBLIC -fPIC)
 add_executable(vitastor-osd
 	osd_main.cpp osd.cpp osd_secondary.cpp osd_peering.cpp osd_flush.cpp osd_peering_pg.cpp
 	osd_primary.cpp osd_primary_chain.cpp osd_primary_sync.cpp osd_primary_write.cpp osd_primary_subops.cpp
-	osd_cluster.cpp osd_rmw.cpp
+	osd_cluster.cpp osd_rmw.cpp osd_scrub.cpp osd_primary_describe.cpp
 )
 target_link_libraries(vitastor-osd
 	vitastor_common
@@ -141,6 +141,8 @@ add_library(vitastor_client SHARED
 	cli_common.cpp
 	cli_alloc_osd.cpp
 	cli_status.cpp
+	cli_describe.cpp
+	cli_fix.cpp
 	cli_df.cpp
 	cli_ls.cpp
 	cli_create.cpp
@@ -299,7 +301,7 @@ add_executable(test_cluster_client
 	EXCLUDE_FROM_ALL
 	test_cluster_client.cpp
 	pg_states.cpp osd_ops.cpp cluster_client.cpp cluster_client_list.cpp msgr_op.cpp mock/messenger.cpp msgr_stop.cpp
-	etcd_state_client.cpp timerfd_manager.cpp ../json11/json11.cpp
+	etcd_state_client.cpp timerfd_manager.cpp str_util.cpp ../json11/json11.cpp
 )
 target_compile_definitions(test_cluster_client PUBLIC -D__MOCK__)
 target_include_directories(test_cluster_client PUBLIC ${CMAKE_SOURCE_DIR}/src/mock)

src/blockstore.h

@@ -73,7 +73,10 @@ Input:
   write request is copied into the metadata area bitwise and stored there.
 
 Output:
-- retval = number of bytes actually read/written or negative error number (-EINVAL or -ENOSPC)
+- retval = number of bytes actually read/written or negative error number
+  -EINVAL = invalid input parameters
+  -ENOENT = requested object/version does not exist for reads
+  -ENOSPC = no space left in the store for writes
 - version = the version actually read or written
 
 ## BS_OP_DELETE
@@ -122,11 +125,14 @@ Output:
 Get a list of all objects in this Blockstore.
 
 Input:
-- oid.stripe = PG alignment
-- len = PG count or 0 to list all objects
-- offset = PG number
-- oid.inode = min inode number or 0 to list all inodes
-- version = max inode number or 0 to list all inodes
+- pg_alignment = PG alignment
+- pg_count = PG count or 0 to list all objects
+- pg_number = PG number
+- list_stable_limit = max number of clean objects in the reply
+  it's guaranteed that dirty objects are returned from the same interval,
+  i.e. from (min_oid .. min(max_oid, max(returned stable OIDs)))
+- min_oid = min inode/stripe or 0 to list all objects
+- max_oid = max inode/stripe or 0 to list all objects
 
 Output:
 - retval = total obj_ver_id count
@@ -143,10 +149,27 @@ struct blockstore_op_t
     uint64_t opcode;
     // finish callback
     std::function<void (blockstore_op_t*)> callback;
-    object_id oid;
-    uint64_t version;
-    uint32_t offset;
-    uint32_t len;
+    union __attribute__((__packed__))
+    {
+        // R/W
+        struct __attribute__((__packed__))
+        {
+            object_id oid;
+            uint64_t version;
+            uint32_t offset;
+            uint32_t len;
+        };
+        // List
+        struct __attribute__((__packed__))
+        {
+            object_id min_oid;
+            object_id max_oid;
+            uint32_t pg_alignment;
+            uint32_t pg_count;
+            uint32_t pg_number;
+            uint32_t list_stable_limit;
+        };
+    };
     void *buf;
     void *bitmap;
     int retval;

src/blockstore_flush.cpp

@@ -536,14 +536,27 @@ resume_1:
                 return false;
             }
             // zero out old metadata entry
+            {
+                clean_disk_entry *old_entry = (clean_disk_entry*)((uint8_t*)meta_old.buf + meta_old.pos*bs->dsk.clean_entry_size);
+                if (old_entry->oid.inode != 0 && old_entry->oid != cur.oid)
+                {
+                    printf("Fatal error (metadata corruption or bug): tried to wipe metadata entry %lu (%lx:%lx v%lu) as old location of %lx:%lx\n",
+                        old_clean_loc >> bs->dsk.block_order, old_entry->oid.inode, old_entry->oid.stripe,
+                        old_entry->version, cur.oid.inode, cur.oid.stripe);
+                    exit(1);
+                }
+            }
             memset((uint8_t*)meta_old.buf + meta_old.pos*bs->dsk.clean_entry_size, 0, bs->dsk.clean_entry_size);
-            await_sqe(15);
-            data->iov = (struct iovec){ meta_old.buf, bs->dsk.meta_block_size };
-            data->callback = simple_callback_w;
-            my_uring_prep_writev(
-                sqe, bs->dsk.meta_fd, &data->iov, 1, bs->dsk.meta_offset + bs->dsk.meta_block_size + meta_old.sector
-            );
-            wait_count++;
+            if (meta_old.sector != meta_new.sector)
+            {
+                await_sqe(15);
+                data->iov = (struct iovec){ meta_old.buf, bs->dsk.meta_block_size };
+                data->callback = simple_callback_w;
+                my_uring_prep_writev(
+                    sqe, bs->dsk.meta_fd, &data->iov, 1, bs->dsk.meta_offset + bs->dsk.meta_block_size + meta_old.sector
+                );
+                wait_count++;
+            }
         }
         if (has_delete)
         {
@@ -662,53 +675,58 @@ resume_1:
                     return false;
                 }
                 flusher->trimming = true;
-                // First update journal "superblock" and only then update <used_start> in memory
-                await_sqe(12);
-                *((journal_entry_start*)flusher->journal_superblock) = {
-                    .crc32 = 0,
-                    .magic = JOURNAL_MAGIC,
-                    .type = JE_START,
-                    .size = sizeof(journal_entry_start),
-                    .reserved = 0,
-                    .journal_start = new_trim_pos,
-                    .version = JOURNAL_VERSION,
-                };
-                ((journal_entry_start*)flusher->journal_superblock)->crc32 = je_crc32((journal_entry*)flusher->journal_superblock);
-                data->iov = (struct iovec){ flusher->journal_superblock, bs->dsk.journal_block_size };
-                data->callback = simple_callback_w;
-                my_uring_prep_writev(sqe, bs->dsk.journal_fd, &data->iov, 1, bs->journal.offset);
-                wait_count++;
-            resume_13:
-                if (wait_count > 0)
+                // Recheck the position with the "lock" taken
+                new_trim_pos = bs->journal.get_trim_pos();
+                if (new_trim_pos != bs->journal.used_start)
                 {
-                    wait_state = 13;
-                    return false;
-                }
-                if (!bs->disable_journal_fsync)
-                {
-                    await_sqe(20);
-                    my_uring_prep_fsync(sqe, bs->dsk.journal_fd, IORING_FSYNC_DATASYNC);
-                    data->iov = { 0 };
+                    // First update journal "superblock" and only then update <used_start> in memory
+                    await_sqe(12);
+                    *((journal_entry_start*)flusher->journal_superblock) = {
+                        .crc32 = 0,
+                        .magic = JOURNAL_MAGIC,
+                        .type = JE_START,
+                        .size = sizeof(journal_entry_start),
+                        .reserved = 0,
+                        .journal_start = new_trim_pos,
+                        .version = JOURNAL_VERSION,
+                    };
+                    ((journal_entry_start*)flusher->journal_superblock)->crc32 = je_crc32((journal_entry*)flusher->journal_superblock);
+                    data->iov = (struct iovec){ flusher->journal_superblock, bs->dsk.journal_block_size };
                     data->callback = simple_callback_w;
-                resume_21:
+                    my_uring_prep_writev(sqe, bs->dsk.journal_fd, &data->iov, 1, bs->journal.offset);
+                    wait_count++;
+                resume_13:
                     if (wait_count > 0)
                     {
-                        wait_state = 21;
+                        wait_state = 13;
                         return false;
                     }
-                }
-                bs->journal.used_start = new_trim_pos;
+                    if (!bs->disable_journal_fsync)
+                    {
+                        await_sqe(20);
+                        my_uring_prep_fsync(sqe, bs->dsk.journal_fd, IORING_FSYNC_DATASYNC);
+                        data->iov = { 0 };
+                        data->callback = simple_callback_w;
+                    resume_21:
+                        if (wait_count > 0)
+                        {
+                            wait_state = 21;
+                            return false;
+                        }
+                    }
+                    bs->journal.used_start = new_trim_pos;
 #ifdef BLOCKSTORE_DEBUG
-                printf("Journal trimmed to %08lx (next_free=%08lx)\n", bs->journal.used_start, bs->journal.next_free);
+                    printf("Journal trimmed to %08lx (next_free=%08lx)\n", bs->journal.used_start, bs->journal.next_free);
 #endif
+                    if (bs->journal.flush_journal && !flusher->flush_queue.size())
+                    {
+                        assert(bs->journal.used_start == bs->journal.next_free);
+                        printf("Journal flushed\n");
+                        exit(0);
+                    }
+                }
                 flusher->trimming = false;
             }
-            if (bs->journal.flush_journal && !flusher->flush_queue.size())
-            {
-                assert(bs->journal.used_start == bs->journal.next_free);
-                printf("Journal flushed\n");
-                exit(0);
-            }
         }
         // All done
         flusher->active_flushers--;

src/blockstore_impl.cpp

@@ -462,11 +462,11 @@ void blockstore_impl_t::reshard_clean_db(pool_id_t pool, uint32_t pg_count, uint
 
 void blockstore_impl_t::process_list(blockstore_op_t *op)
 {
-    uint32_t list_pg = op->offset+1;
-    uint32_t pg_count = op->len;
-    uint64_t pg_stripe_size = op->oid.stripe;
-    uint64_t min_inode = op->oid.inode;
-    uint64_t max_inode = op->version;
+    uint32_t list_pg = op->pg_number+1;
+    uint32_t pg_count = op->pg_count;
+    uint64_t pg_stripe_size = op->pg_alignment;
+    uint64_t min_inode = op->min_oid.inode;
+    uint64_t max_inode = op->max_oid.inode;
     // Check PG
     if (pg_count != 0 && (pg_stripe_size < MIN_DATA_BLOCK_SIZE || list_pg > pg_count))
     {
@@ -513,7 +513,13 @@ void blockstore_impl_t::process_list(blockstore_op_t *op)
             stable_alloc += clean_db.size();
         }
     }
-    else
+    if (op->list_stable_limit > 0)
+    {
+        stable_alloc = op->list_stable_limit;
+        if (stable_alloc > 1024*1024)
+            stable_alloc = 1024*1024;
+    }
+    if (stable_alloc < 32768)
     {
         stable_alloc = 32768;
     }
@@ -524,22 +530,22 @@ void blockstore_impl_t::process_list(blockstore_op_t *op)
         FINISH_OP(op);
         return;
     }
+    auto max_oid = op->max_oid;
+    bool limited = false;
+    pool_pg_id_t last_shard_id = 0;
     for (auto shard_it = clean_db_shards.lower_bound(first_shard);
         shard_it != clean_db_shards.end() && shard_it->first <= last_shard;
         shard_it++)
     {
         auto & clean_db = shard_it->second;
         auto clean_it = clean_db.begin(), clean_end = clean_db.end();
-        if ((min_inode != 0 || max_inode != 0) && min_inode <= max_inode)
+        if (op->min_oid.inode != 0 || op->min_oid.stripe != 0)
         {
-            clean_it = clean_db.lower_bound({
-                .inode = min_inode,
-                .stripe = 0,
-            });
-            clean_end = clean_db.upper_bound({
-                .inode = max_inode,
-                .stripe = UINT64_MAX,
-            });
+            clean_it = clean_db.lower_bound(op->min_oid);
+        }
+        if ((max_oid.inode != 0 || max_oid.stripe != 0) && !(max_oid < op->min_oid))
+        {
+            clean_end = clean_db.upper_bound(max_oid);
         }
         for (; clean_it != clean_end; clean_it++)
         {
@@ -558,11 +564,29 @@ void blockstore_impl_t::process_list(blockstore_op_t *op)
                 .oid = clean_it->first,
                 .version = clean_it->second.version,
             };
+            if (op->list_stable_limit > 0 && stable_count >= op->list_stable_limit)
+            {
+                if (!limited)
+                {
+                    limited = true;
+                    max_oid = stable[stable_count-1].oid;
+                }
+                break;
+            }
         }
+        if (op->list_stable_limit > 0)
+        {
+            // To maintain the order, we have to include objects in the same range from other shards
+            if (last_shard_id != 0 && last_shard_id != shard_it->first)
+                std::sort(stable, stable+stable_count);
+            if (stable_count > op->list_stable_limit)
+                stable_count = op->list_stable_limit;
+        }
+        last_shard_id = shard_it->first;
     }
-    if (first_shard != last_shard)
+    if (op->list_stable_limit == 0 && first_shard != last_shard)
     {
-        // If that's not a per-PG listing, sort clean entries
+        // If that's not a per-PG listing, sort clean entries (already sorted if list_stable_limit != 0)
         std::sort(stable, stable+stable_count);
     }
     int clean_stable_count = stable_count;
@@ -571,20 +595,17 @@ void blockstore_impl_t::process_list(blockstore_op_t *op)
     obj_ver_id *unstable = NULL;
     {
         auto dirty_it = dirty_db.begin(), dirty_end = dirty_db.end();
-        if ((min_inode != 0 || max_inode != 0) && min_inode <= max_inode)
+        if (op->min_oid.inode != 0 || op->min_oid.stripe != 0)
         {
             dirty_it = dirty_db.lower_bound({
-                .oid = {
-                    .inode = min_inode,
-                    .stripe = 0,
-                },
+                .oid = op->min_oid,
                 .version = 0,
             });
+        }
+        if ((max_oid.inode != 0 || max_oid.stripe != 0) && !(max_oid < op->min_oid))
+        {
             dirty_end = dirty_db.upper_bound({
-                .oid = {
-                    .inode = max_inode,
-                    .stripe = UINT64_MAX,
-                },
+                .oid = max_oid,
                 .version = UINT64_MAX,
             });
         }
@@ -628,6 +649,11 @@ void blockstore_impl_t::process_list(blockstore_op_t *op)
                             stable[stable_count++] = dirty_it->first;
                         }
                     }
+                    if (op->list_stable_limit > 0 && stable_count >= op->list_stable_limit)
+                    {
+                        // Stop here
+                        break;
+                    }
                 }
                 else
                 {

src/blockstore_journal.cpp

@@ -236,14 +236,6 @@ journal_t::~journal_t()
 uint64_t journal_t::get_trim_pos()
 {
     auto journal_used_it = used_sectors.lower_bound(used_start);
-#ifdef BLOCKSTORE_DEBUG
-    printf(
-        "Trimming journal (used_start=%08lx, next_free=%08lx, dirty_start=%08lx, new_start=%08lx, new_refcount=%ld)\n",
-        used_start, next_free, dirty_start,
-        journal_used_it == used_sectors.end() ? 0 : journal_used_it->first,
-        journal_used_it == used_sectors.end() ? 0 : journal_used_it->second
-    );
-#endif
     if (journal_used_it == used_sectors.end())
     {
         // Journal is cleared to its end, restart from the beginning
@@ -256,12 +248,26 @@ uint64_t journal_t::get_trim_pos()
         else
         {
             // next_free does not need updating during trim
+#ifdef BLOCKSTORE_DEBUG
+            printf(
+                "Trimming journal (used_start=%08lx, next_free=%08lx, dirty_start=%08lx, new_start=%08lx, new_refcount=%ld)\n",
+                used_start, next_free, dirty_start,
+                journal_used_it->first, journal_used_it->second
+            );
+#endif
             return journal_used_it->first;
         }
     }
     else if (journal_used_it->first > used_start)
     {
         // Journal is cleared up to <journal_used_it>
+#ifdef BLOCKSTORE_DEBUG
+        printf(
+            "Trimming journal (used_start=%08lx, next_free=%08lx, dirty_start=%08lx, new_start=%08lx, new_refcount=%ld)\n",
+            used_start, next_free, dirty_start,
+            journal_used_it->first, journal_used_it->second
+        );
+#endif
         return journal_used_it->first;
     }
     // Can't trim journal

src/blockstore_read.cpp

@@ -124,10 +124,8 @@ int blockstore_impl_t::dequeue_read(blockstore_op_t *read_op)
     bool dirty_found = (dirty_it != dirty_db.end() && dirty_it->first.oid == read_op->oid);
     if (!clean_found && !dirty_found)
     {
-        // region is not allocated - return zeroes
-        memset(read_op->buf, 0, read_op->len);
         read_op->version = 0;
-        read_op->retval = read_op->len;
+        read_op->retval = -ENOENT;
         FINISH_OP(read_op);
         return 2;
     }
@@ -140,14 +138,16 @@ int blockstore_impl_t::dequeue_read(blockstore_op_t *read_op)
         {
             dirty_entry& dirty = dirty_it->second;
             bool version_ok = !IS_IN_FLIGHT(dirty.state) && read_op->version >= dirty_it->first.version;
-            if (IS_SYNCED(dirty.state))
-            {
-                if (!version_ok && read_op->version != 0)
-                    read_op->version = dirty_it->first.version;
-                version_ok = true;
-            }
             if (version_ok)
             {
+                if (IS_DELETE(dirty.state))
+                {
+                    assert(!result_version);
+                    read_op->version = 0;
+                    read_op->retval = -ENOENT;
+                    FINISH_OP(read_op);
+                    return 2;
+                }
                 if (!result_version)
                 {
                     result_version = dirty_it->first.version;
@@ -234,12 +234,19 @@ int blockstore_impl_t::dequeue_read(blockstore_op_t *read_op)
             }
         }
     }
-    else if (fulfilled < read_op->len)
+    if (!result_version)
     {
-        // fill remaining parts with zeroes
-        assert(fulfill_read(read_op, fulfilled, 0, dsk.data_block_size, (BS_ST_DELETE | BS_ST_STABLE), 0, 0, 0));
+        // May happen if there are entries in dirty_db but all of them are !version_ok
+        read_op->version = 0;
+        read_op->retval = -ENOENT;
+        FINISH_OP(read_op);
+        return 2;
+    }
+    if (fulfilled < read_op->len)
+    {
+        assert(fulfill_read(read_op, fulfilled, 0, dsk.data_block_size, (BS_ST_DELETE | BS_ST_STABLE), 0, 0, 0));
+        assert(fulfilled == read_op->len);
     }
-    assert(fulfilled == read_op->len);
     read_op->version = result_version;
     if (!PRIV(read_op)->pending_ops)
     {

src/blockstore_rollback.cpp

@@ -179,7 +179,7 @@ void blockstore_impl_t::erase_dirty(blockstore_dirty_db_t::iterator dirty_start,
     {
         object_id oid = dirty_it->first.oid;
 #ifdef BLOCKSTORE_DEBUG
-        printf("Unblock writes-after-delete %lx:%lx v%lx\n", oid.inode, oid.stripe, dirty_it->first.version);
+        printf("Unblock writes-after-delete %lx:%lx v%lu\n", oid.inode, oid.stripe, dirty_it->first.version);
 #endif
         dirty_it = dirty_end;
         // Unblock operations blocked by delete flushing

src/blockstore_stable.cpp

@@ -103,7 +103,7 @@ blockstore_op_t* blockstore_impl_t::selective_sync(blockstore_op_t *op)
     blockstore_op_t *sync_op = new blockstore_op_t;
     sync_op->opcode = BS_OP_SYNC;
     sync_op->buf = NULL;
-    sync_op->callback = [this](blockstore_op_t *sync_op)
+    sync_op->callback = [](blockstore_op_t *sync_op)
     {
         delete sync_op;
     };
@@ -244,7 +244,7 @@ int blockstore_impl_t::split_stab_op(blockstore_op_t *op, std::function<int(obj_
         // Make a wrapped callback
         int *split_op_counter = (int*)malloc_or_die(sizeof(int));
         *split_op_counter = (sync_op ? 1 : 0) + (split_stab_op ? 1 : 0) + (todo ? 1 : 0);
-        auto cb = [this, op, good_items = good_vers.items,
+        auto cb = [op, good_items = good_vers.items,
             bad_items = bad_vers.items, split_op_counter,
             orig_buf, real_cb = op->callback](blockstore_op_t *split_op)
         {
@@ -458,6 +458,16 @@ void blockstore_impl_t::mark_stable(const obj_ver_id & v, bool forget_dirty)
                     big_to_flush++;
                 }
             }
+            else if (IS_IN_FLIGHT(dirty_it->second.state))
+            {
+                // mark_stable should never be called for in-flight or submitted writes
+                printf(
+                    "BUG: Attempt to mark_stable object %lx:%lx v%lu state of which is %x\n",
+                    dirty_it->first.oid.inode, dirty_it->first.oid.stripe, dirty_it->first.version,
+                    dirty_it->second.state
+                );
+                exit(1);
+            }
             if (forget_dirty && (IS_BIG_WRITE(dirty_it->second.state) ||
                 IS_DELETE(dirty_it->second.state)))
             {

src/blockstore_write.cpp

@@ -6,7 +6,7 @@
 bool blockstore_impl_t::enqueue_write(blockstore_op_t *op)
 {
     // Check or assign version number
-    bool found = false, deleted = false, is_del = (op->opcode == BS_OP_DELETE);
+    bool found = false, deleted = false, unsynced = false, is_del = (op->opcode == BS_OP_DELETE);
     bool wait_big = false, wait_del = false;
     void *bmp = NULL;
     uint64_t version = 1;
@@ -26,6 +26,7 @@ bool blockstore_impl_t::enqueue_write(blockstore_op_t *op)
             found = true;
             version = dirty_it->first.version + 1;
             deleted = IS_DELETE(dirty_it->second.state);
+            unsynced = !IS_SYNCED(dirty_it->second.state);
             wait_del = ((dirty_it->second.state & BS_ST_WORKFLOW_MASK) == BS_ST_WAIT_DEL);
             wait_big = (dirty_it->second.state & BS_ST_TYPE_MASK) == BS_ST_BIG_WRITE
                 ? !IS_SYNCED(dirty_it->second.state)
@@ -81,10 +82,28 @@ bool blockstore_impl_t::enqueue_write(blockstore_op_t *op)
             wait_del = true;
             PRIV(op)->real_version = op->version;
             op->version = version;
-            flusher->unshift_flush((obj_ver_id){
-                .oid = op->oid,
-                .version = version-1,
-            }, true);
+            if (unsynced)
+            {
+                // Issue an additional sync so the delete reaches the journal
+                blockstore_op_t *sync_op = new blockstore_op_t;
+                sync_op->opcode = BS_OP_SYNC;
+                sync_op->callback = [this, op](blockstore_op_t *sync_op)
+                {
+                    flusher->unshift_flush((obj_ver_id){
+                        .oid = op->oid,
+                        .version = op->version-1,
+                    }, true);
+                    delete sync_op;
+                };
+                enqueue_op(sync_op);
+            }
+            else
+            {
+                flusher->unshift_flush((obj_ver_id){
+                    .oid = op->oid,
+                    .version = version-1,
+                }, true);
+            }
         }
         else
         {
@@ -390,7 +409,23 @@ int blockstore_impl_t::dequeue_write(blockstore_op_t *op)
         );
 #endif
         // Figure out where data will be
-        journal.next_free = (journal.next_free + op->len) <= journal.len ? journal.next_free : dsk.journal_block_size;
+        auto next_next_free = (journal.next_free + op->len) <= journal.len ? journal.next_free : dsk.journal_block_size;
+        if (op->len > 0)
+        {
+            auto journal_used_it = journal.used_sectors.lower_bound(next_next_free);
+            if (journal_used_it != journal.used_sectors.end() &&
+                journal_used_it->first < next_next_free + op->len)
+            {
+                printf(
+                    "BUG: Attempt to overwrite used offset (%lx, %lu refs) of the journal with the object %lx:%lx v%lu: data at %lx, len %x!"
+                    " Journal used_start=%08lx (%lu refs), next_free=%08lx, dirty_start=%08lx\n",
+                    journal_used_it->first, journal_used_it->second, op->oid.inode, op->oid.stripe, op->version, next_next_free, op->len,
+                    journal.used_start, journal.used_sectors[journal.used_start], journal.next_free, journal.dirty_start
+                );
+                exit(1);
+            }
+        }
+        journal.next_free = next_next_free;
         je->oid = op->oid;
         je->version = op->version;
         je->offset = op->offset;

src/cli.cpp

@@ -73,6 +73,37 @@ static const char* help_text =
     "  <to> must be a child of <from> and <target> may be one of the layers between\n"
     "  <from> and <to>, including <from> and <to>.\n"
     "\n"
+    "vitastor-cli describe [--osds <osds>] [--object-state <states>] [--pool <pool>] [--inode <ino>] [--min-inode <ino>] [--max-inode <ino>] [--min-offset <offset>] [--max-offset <offset>]\n"
+    "  Describe unclean object locations in the cluster.\n"
+    "  --osds <osds>\n"
+    "      Only list objects from primary OSD(s) <osds>.\n"
+    "  --object-state <states>\n"
+    "      Only list objects in given state(s). State(s) may include:\n"
+    "      degraded, misplaced, incomplete, corrupted, inconsistent.\n"
+    "  --pool <pool name or number>\n"
+    "      Only list objects in the given pool.\n"
+    "  --inode, --min-inode, --max-inode\n"
+    "      Restrict listing to specific inode numbers.\n"
+    "  --min-offset, --max-offset\n"
+    "      Restrict listing to specific offsets inside inodes.\n"
+    "\n"
+    "vitastor-cli fix [--objects <objects>] [--bad-osds <osds>] [--part <part>] [--check no]\n"
+    "  Fix inconsistent objects in the cluster by deleting some copies.\n"
+    "  --objects <objects>\n"
+    "      Objects to fix, either in plain text or JSON format. If not specified,\n"
+    "      object list will be read from STDIN in one of the same formats.\n"
+    "      Plain text format: 0x<inode>:0x<stripe> <any delimiter> 0x<inode>:0x<stripe> ...\n"
+    "      JSON format: [{\"inode\":\"0x...\",\"stripe\":\"0x...\"},...]\n"
+    "  --bad-osds <osds>\n"
+    "      Remove inconsistent copies/parts of objects from these OSDs, effectively\n"
+    "      marking them bad and allowing Vitastor to recover objects from other copies.\n"
+    "  --part <number>\n"
+    "      Only remove EC part <number> (from 0 to pg_size-1), required for extreme\n"
+    "      edge cases where one OSD has multiple parts of a EC object.\n"
+    "  --check no\n"
+    "      Do not recheck that requested objects are actually inconsistent,\n"
+    "      delete requested copies/parts anyway.\n"
+    "\n"
     "vitastor-cli alloc-osd\n"
     "  Allocate a new OSD number and reserve it by creating empty /osd/stats/<n> key.\n"
     "\n"
@@ -168,6 +199,7 @@ static json11::Json::object parse_args(int narg, const char *args[])
 static int run(cli_tool_t *p, json11::Json::object cfg)
 {
     cli_result_t result = {};
+    p->is_command_line = true;
     p->parse_config(cfg);
     json11::Json::array cmd = cfg["command"].array_items();
     cfg.erase("command");
@@ -276,6 +308,16 @@ static int run(cli_tool_t *p, json11::Json::object cfg)
         }
         action_cb = p->start_rm(cfg);
     }
+    else if (cmd[0] == "describe")
+    {
+        // Describe unclean objects
+        action_cb = p->start_describe(cfg);
+    }
+    else if (cmd[0] == "fix")
+    {
+        // Fix inconsistent objects (by deleting some copies)
+        action_cb = p->start_fix(cfg);
+    }
     else if (cmd[0] == "alloc-osd")
     {
         // Allocate a new OSD number

src/cli.h

@@ -34,12 +34,12 @@ public:
     bool list_first = false;
     bool json_output = false;
     int log_level = 0;
+    bool is_command_line = false;
     bool color = false;
 
     ring_loop_t *ringloop = NULL;
     epoll_manager_t *epmgr = NULL;
     cluster_client_t *cli = NULL;
-    bool no_recovery = false, no_rebalance = false, readonly = false;
 
     int waiting = 0;
     cli_result_t etcd_err;
@@ -56,6 +56,8 @@ public:
     friend struct snap_remover_t;
 
     std::function<bool(cli_result_t &)> start_status(json11::Json);
+    std::function<bool(cli_result_t &)> start_describe(json11::Json);
+    std::function<bool(cli_result_t &)> start_fix(json11::Json);
     std::function<bool(cli_result_t &)> start_df(json11::Json);
     std::function<bool(cli_result_t &)> start_ls(json11::Json);
     std::function<bool(cli_result_t &)> start_create(json11::Json);

src/cli_describe.cpp

@@ -0,0 +1,256 @@
+// Copyright (c) Vitaliy Filippov, 2019+
+// License: VNPL-1.1 (see README.md for details)
+
+#include "cli_fix.h"
+#include "cluster_client.h"
+#include "pg_states.h"
+#include "str_util.h"
+
+std::vector<uint64_t> parse_uint64_list(json11::Json val)
+{
+    std::vector<uint64_t> ret;
+    if (val.is_number())
+        ret.push_back(val.uint64_value());
+    else if (val.is_string())
+    {
+        const std::string & s = val.string_value();
+        for (int i = 0, p = -1; i <= s.size(); i++)
+        {
+            if (p < 0 && i < s.size() && (isdigit(s[i]) || s[i] == 'x'))
+                p = i;
+            else if (p >= 0 && (i >= s.size() || !isdigit(s[i]) && s[i] != 'x'))
+            {
+                ret.push_back(stoull_full(s.substr(p, i-p), 0));
+                p = -1;
+            }
+        }
+    }
+    else if (val.is_array())
+    {
+        for (auto & pg_num: val.array_items())
+            ret.push_back(pg_num.uint64_value());
+    }
+    return ret;
+}
+
+struct cli_describe_t
+{
+    uint64_t object_state = 0;
+    pool_id_t only_pool = 0;
+    std::vector<uint64_t> only_osds;
+    uint64_t min_inode = 0, max_inode = 0;
+    uint64_t min_offset = 0, max_offset = 0;
+
+    cli_tool_t *parent = NULL;
+    int state = 0;
+    int count = 0;
+
+    json11::Json options;
+    cli_result_t result;
+    json11::Json::array describe_items;
+
+    bool is_done()
+    {
+        return state == 100;
+    }
+
+    void parse_options(json11::Json cfg)
+    {
+        only_pool = cfg["pool"].uint64_value();
+        if (!only_pool && cfg["pool"].is_string())
+        {
+            for (auto & pp: parent->cli->st_cli.pool_config)
+            {
+                if (pp.second.name == cfg["pool"].string_value())
+                {
+                    only_pool = pp.first;
+                    break;
+                }
+            }
+        }
+        min_inode = cfg["inode"].uint64_value();
+        if (min_inode)
+        {
+            if (!INODE_POOL(min_inode))
+                min_inode |= (uint64_t)only_pool << (64-POOL_ID_BITS);
+            max_inode = min_inode;
+            min_offset = max_offset = 0;
+        }
+        else
+        {
+            min_inode = stoull_full(cfg["min_inode"].string_value(), 0); // to support 0x...
+            max_inode = stoull_full(cfg["max_inode"].string_value(), 0);
+            min_offset = stoull_full(cfg["min_offset"].string_value(), 0);
+            max_offset = stoull_full(cfg["max_offset"].string_value(), 0);
+            if (!min_inode && !max_inode && only_pool)
+            {
+                min_inode = (uint64_t)only_pool << (64-POOL_ID_BITS);
+                max_inode = ((uint64_t)only_pool << (64-POOL_ID_BITS)) |
+                    (((uint64_t)1 << (64-POOL_ID_BITS)) - 1);
+            }
+        }
+        only_osds = parse_uint64_list(cfg["osds"]);
+        object_state = stoull_full(cfg["object_state"].string_value(), 0);
+        if (!object_state && cfg["object_state"].is_string())
+        {
+            if (cfg["object_state"].string_value().find("inconsistent") != std::string::npos)
+                object_state |= OBJ_INCONSISTENT;
+            if (cfg["object_state"].string_value().find("corrupted") != std::string::npos)
+                object_state |= OBJ_CORRUPTED;
+            if (cfg["object_state"].string_value().find("incomplete") != std::string::npos)
+                object_state |= OBJ_INCOMPLETE;
+            if (cfg["object_state"].string_value().find("degraded") != std::string::npos)
+                object_state |= OBJ_DEGRADED;
+            if (cfg["object_state"].string_value().find("misplaced") != std::string::npos)
+                object_state |= OBJ_MISPLACED;
+        }
+    }
+
+    void loop()
+    {
+        if (state == 1)
+            goto resume_1;
+        if (state == 100)
+            return;
+        parse_options(options);
+        if (min_inode && !INODE_POOL(min_inode))
+        {
+            result = (cli_result_t){ .err = EINVAL, .text = "Pool is not specified" };
+            state = 100;
+            return;
+        }
+        if (!only_osds.size())
+        {
+            uint64_t min_pool = min_inode >> (64-POOL_ID_BITS);
+            uint64_t max_pool = max_inode >> (64-POOL_ID_BITS);
+            for (auto & pp: parent->cli->st_cli.pool_config)
+            {
+                if (pp.first >= min_pool && (!max_pool || pp.first <= max_pool))
+                {
+                    for (auto & pgp: pp.second.pg_config)
+                        only_osds.push_back(pgp.second.cur_primary);
+                }
+            }
+        }
+        remove_duplicates(only_osds);
+        parent->cli->init_msgr();
+        if (parent->json_output && parent->is_command_line)
+        {
+            printf("[\n");
+        }
+        for (int i = 0; i < only_osds.size(); i++)
+        {
+            osd_op_t *op = new osd_op_t;
+            op->req = (osd_any_op_t){
+                .describe = {
+                    .header = {
+                        .magic = SECONDARY_OSD_OP_MAGIC,
+                        .id = parent->cli->next_op_id(),
+                        .opcode = OSD_OP_DESCRIBE,
+                    },
+                    .object_state = object_state,
+                    .min_inode = min_inode,
+                    .min_offset = min_offset,
+                    .max_inode = max_inode,
+                    .max_offset = max_offset,
+                },
+            };
+            op->callback = [this, osd_num = only_osds[i]](osd_op_t *op)
+            {
+                if (op->reply.hdr.retval < 0)
+                {
+                    fprintf(
+                        stderr, "Failed to describe objects on OSD %lu (retval=%ld)\n",
+                        osd_num, op->reply.hdr.retval
+                    );
+                }
+                else if (op->reply.describe.result_bytes != op->reply.hdr.retval * sizeof(osd_reply_describe_item_t))
+                {
+                    fprintf(
+                        stderr, "Invalid response size from OSD %lu (expected %lu bytes, got %lu bytes)\n",
+                        osd_num, op->reply.hdr.retval * sizeof(osd_reply_describe_item_t), op->reply.describe.result_bytes
+                    );
+                }
+                else
+                {
+                    osd_reply_describe_item_t *items = (osd_reply_describe_item_t *)op->buf;
+                    for (int i = 0; i < op->reply.hdr.retval; i++)
+                    {
+                        if (!parent->json_output || parent->is_command_line)
+                        {
+#define FMT "{\"inode\":\"0x%lx\",\"stripe\":\"0x%lx\",\"part\":%u,\"osd_num\":%lu%s%s%s}"
+                            printf(
+                                (parent->json_output
+                                    ? (count > 0 ? ",\n  " FMT : "  " FMT)
+                                    : "%lx:%lx part %u on OSD %lu%s%s%s\n"),
+#undef FMT
+                                items[i].inode, items[i].stripe,
+                                items[i].role, items[i].osd_num,
+                                (items[i].loc_bad & LOC_CORRUPTED ? (parent->json_output ? ",\"corrupted\":true" : " corrupted") : ""),
+                                (items[i].loc_bad & LOC_INCONSISTENT ? (parent->json_output ? ",\"inconsistent\":true" : " inconsistent") : ""),
+                                (items[i].loc_bad & LOC_OUTDATED ? (parent->json_output ? ",\"outdated\":true" : " outdated") : "")
+                            );
+                        }
+                        else
+                        {
+                            auto json_item = json11::Json::object {
+                                { "inode", (uint64_t)items[i].inode },
+                                { "stripe", (uint64_t)items[i].stripe },
+                                { "part", (uint64_t)items[i].role },
+                                { "osd_num", (uint64_t)items[i].osd_num },
+                            };
+                            if (items[i].loc_bad & LOC_CORRUPTED)
+                                json_item["corrupted"] = true;
+                            if (items[i].loc_bad & LOC_INCONSISTENT)
+                                json_item["inconsistent"] = true;
+                            if (items[i].loc_bad & LOC_OUTDATED)
+                                json_item["outdated"] = true;
+                            describe_items.push_back(json_item);
+                        }
+                        count++;
+                    }
+                }
+                delete op;
+                parent->waiting--;
+                if (!parent->waiting)
+                    loop();
+            };
+            parent->waiting++;
+            parent->cli->execute_raw(only_osds[i], op);
+        }
+    resume_1:
+        state = 1;
+        if (parent->waiting > 0)
+        {
+            return;
+        }
+        if (parent->json_output && parent->is_command_line)
+        {
+            printf(count > 0 ? "\n]\n" : "]\n");
+        }
+        else
+        {
+            result.data = describe_items;
+        }
+        state = 100;
+        describe_items.clear();
+    }
+};
+
+std::function<bool(cli_result_t &)> cli_tool_t::start_describe(json11::Json cfg)
+{
+    auto describer = new cli_describe_t();
+    describer->parent = this;
+    describer->options = cfg;
+    return [describer](cli_result_t & result)
+    {
+        describer->loop();
+        if (describer->is_done())
+        {
+            result = describer->result;
+            delete describer;
+            return true;
+        }
+        return false;
+    };
+}

src/cli_fix.cpp

@@ -0,0 +1,313 @@
+// Copyright (c) Vitaliy Filippov, 2019+
+// License: VNPL-1.1 (see README.md for details)
+
+#include "cli_fix.h"
+#include "cluster_client.h"
+#include "pg_states.h"
+#include "str_util.h"
+
+struct cli_fix_t
+{
+    std::vector<object_id> objects;
+    int part = -1;
+    int processed_count = 0;
+    std::set<osd_num_t> bad_osds;
+    bool no_check = false;
+
+    cli_tool_t *parent = NULL;
+    int state = 0;
+
+    json11::Json options;
+    cli_result_t result;
+    json11::Json::array fix_result;
+
+    bool is_done()
+    {
+        return state == 100;
+    }
+
+    void parse_objects_str(std::string str)
+    {
+        str = trim(str);
+        if (str[0] == '[')
+        {
+            std::string json_err;
+            json11::Json list = json11::Json::parse(str, json_err);
+            if (json_err != "")
+                fprintf(stderr, "Invalid JSON object list input: %s\n", json_err.c_str());
+            else
+                parse_object_list(list);
+        }
+        else
+        {
+            const char *s = str.c_str();
+            char *e = NULL;
+            int len = str.size();
+            object_id oid;
+            for (int p = 0; p < len; p++)
+            {
+                if (isdigit(s[p]))
+                {
+                    int p0 = p;
+                    oid.inode = strtoull(s+p, &e, 0);
+                    p = e-s;
+                    while (p < len && !isdigit(s[p]) && s[p] != ':')
+                        p++;
+                    if (s[p] != ':')
+                    {
+                        fprintf(stderr, "Invalid object ID in input: %s\n", std::string(s+p0, p-p0).c_str());
+                        continue;
+                    }
+                    p++;
+                    while (p < len && !isdigit(s[p]))
+                        p++;
+                    oid.stripe = strtoull(s+p, &e, 0) & ~STRIPE_MASK;
+                    p = e-s;
+                    if (oid.inode)
+                        objects.push_back(oid);
+                    else
+                        fprintf(stderr, "Invalid object ID in input: %s\n", std::string(s+p0, p-p0).c_str());
+                }
+            }
+        }
+    }
+
+    void parse_object_list(json11::Json list)
+    {
+        for (auto & obj: list.array_items())
+        {
+            object_id oid = (object_id){
+                .inode = stoull_full(obj["inode"].string_value(), 0),
+                .stripe = stoull_full(obj["stripe"].string_value(), 0) & ~STRIPE_MASK,
+            };
+            if (oid.inode)
+                objects.push_back(oid);
+            else
+                fprintf(stderr, "Invalid JSON object ID in input: %s, bad or missing \"inode\" field\n", obj.dump().c_str());
+        }
+    }
+
+    void parse_options(json11::Json cfg)
+    {
+        json11::Json object_list;
+        if (cfg["objects"].is_null())
+            parse_objects_str(read_all_fd(0));
+        else if (cfg["objects"].is_string())
+            parse_objects_str(cfg["objects"].string_value());
+        else
+            parse_object_list(cfg["objects"].array_items());
+        for (auto osd_num: parse_uint64_list(cfg["bad_osds"]))
+            bad_osds.insert(osd_num);
+        no_check = json_is_false(cfg["check"]);
+        if (cfg["part"].is_number() || cfg["part"].is_string())
+            part = cfg["part"].uint64_value();
+    }
+
+    void loop()
+    {
+        if (state == 1)
+            goto resume_1;
+        if (state == 100)
+            return;
+        parse_options(options);
+        if (!objects.size())
+        {
+            result = (cli_result_t){ .err = EINVAL, .text = "Object list is not specified" };
+            state = 100;
+            return;
+        }
+        if (!bad_osds.size())
+        {
+            result = (cli_result_t){ .err = EINVAL, .text = "OSDs are not specified" };
+            state = 100;
+            return;
+        }
+        remove_duplicates(objects);
+        parent->cli->init_msgr();
+    resume_1:
+        state = 1;
+        while (processed_count < objects.size())
+        {
+            if (parent->waiting >= parent->iodepth*parent->parallel_osds)
+            {
+                return;
+            }
+            auto & obj = objects[processed_count++];
+            auto pool_cfg_it = parent->cli->st_cli.pool_config.find(INODE_POOL(obj.inode));
+            if (pool_cfg_it == parent->cli->st_cli.pool_config.end())
+            {
+                fprintf(stderr, "Object %lx:%lx is from unknown pool\n", obj.inode, obj.stripe);
+                continue;
+            }
+            auto & pool_cfg = pool_cfg_it->second;
+            pg_num_t pg_num = (obj.stripe/pool_cfg.pg_stripe_size) % pool_cfg.real_pg_count + 1; // like map_to_pg()
+            auto pg_it = pool_cfg.pg_config.find(pg_num);
+            if (pg_it == pool_cfg.pg_config.end() ||
+                !pg_it->second.cur_primary || !(pg_it->second.cur_state & PG_ACTIVE))
+            {
+                fprintf(
+                    stderr, "Object %lx:%lx is from PG %u/%u which is not currently active\n",
+                    obj.inode, obj.stripe, pool_cfg_it->first, pg_num
+                );
+                continue;
+            }
+            osd_num_t primary_osd = pg_it->second.cur_primary;
+            // Describe -> Remove some copies -> Scrub again
+            osd_op_t *op = new osd_op_t;
+            op->req = (osd_any_op_t){
+                .describe = {
+                    .header = {
+                        .magic = SECONDARY_OSD_OP_MAGIC,
+                        .id = parent->cli->next_op_id(),
+                        .opcode = OSD_OP_DESCRIBE,
+                    },
+                    .min_inode = obj.inode,
+                    .min_offset = obj.stripe,
+                    .max_inode = obj.inode,
+                    .max_offset = obj.stripe,
+                },
+            };
+            op->callback = [this, primary_osd, &obj](osd_op_t *op)
+            {
+                if (op->reply.hdr.retval < 0 || op->reply.describe.result_bytes != op->reply.hdr.retval * sizeof(osd_reply_describe_item_t))
+                {
+                    fprintf(stderr, "Failed to describe objects on OSD %lu (retval=%ld)\n", primary_osd, op->reply.hdr.retval);
+                    parent->waiting--;
+                    loop();
+                }
+                else
+                {
+                    osd_reply_describe_item_t *items = (osd_reply_describe_item_t *)op->buf;
+                    int *rm_count = (int*)malloc_or_die(sizeof(int));
+                    *rm_count = 1; // just in case if anything gets called instantly
+                    for (int i = 0; i < op->reply.hdr.retval; i++)
+                    {
+                        if (((items[i].loc_bad & LOC_INCONSISTENT) || no_check) &&
+                            bad_osds.find(items[i].osd_num) != bad_osds.end() &&
+                            (part == -1 || items[i].role == part))
+                        {
+                            // Remove
+                            uint64_t rm_osd_num = items[i].osd_num;
+                            osd_op_t *rm_op = new osd_op_t;
+                            rm_op->req = (osd_any_op_t){
+                                .sec_del = {
+                                    .header = {
+                                        .magic = SECONDARY_OSD_OP_MAGIC,
+                                        .id = parent->cli->next_op_id(),
+                                        .opcode = OSD_OP_SEC_DELETE,
+                                    },
+                                    .oid = {
+                                        .inode = op->req.describe.min_inode,
+                                        .stripe = op->req.describe.min_offset | items[i].role,
+                                    },
+                                    .version = 0,
+                                },
+                            };
+                            rm_op->callback = [this, primary_osd, rm_osd_num, rm_count, &obj](osd_op_t *rm_op)
+                            {
+                                (*rm_count)--;
+                                if (rm_op->reply.hdr.retval < 0)
+                                {
+                                    fprintf(
+                                        stderr, "Failed to remove object %lx:%lx from OSD %lu (retval=%ld)\n",
+                                        rm_op->req.sec_del.oid.inode, rm_op->req.sec_del.oid.stripe,
+                                        rm_osd_num, rm_op->reply.hdr.retval
+                                    );
+                                }
+                                else if (parent->json_output)
+                                {
+                                    fix_result.push_back(json11::Json::object {
+                                        { "inode", (uint64_t)rm_op->req.sec_del.oid.inode },
+                                        { "stripe", (uint64_t)rm_op->req.sec_del.oid.stripe & ~STRIPE_MASK },
+                                        { "part", (uint64_t)rm_op->req.sec_del.oid.stripe & STRIPE_MASK },
+                                        { "osd_num", (uint64_t)rm_osd_num },
+                                    });
+                                }
+                                else
+                                {
+                                    printf(
+                                        "Removed %lx:%lx (part %lu) from OSD %lu\n",
+                                        rm_op->req.sec_del.oid.inode, rm_op->req.sec_del.oid.stripe & ~STRIPE_MASK,
+                                        rm_op->req.sec_del.oid.stripe & STRIPE_MASK, rm_osd_num
+                                    );
+                                }
+                                delete rm_op;
+                                if (!(*rm_count))
+                                {
+                                    // Scrub
+                                    free(rm_count);
+                                    osd_op_t *scrub_op = new osd_op_t;
+                                    scrub_op->req = (osd_any_op_t){
+                                        .rw = {
+                                            .header = {
+                                                .magic = SECONDARY_OSD_OP_MAGIC,
+                                                .id = parent->cli->next_op_id(),
+                                                .opcode = OSD_OP_SCRUB,
+                                            },
+                                            .inode = obj.inode,
+                                            .offset = obj.stripe,
+                                            .len = 0,
+                                        },
+                                    };
+                                    scrub_op->callback = [this, primary_osd, &obj](osd_op_t *scrub_op)
+                                    {
+                                        if (scrub_op->reply.hdr.retval < 0 && scrub_op->reply.hdr.retval != -ENOENT)
+                                        {
+                                            fprintf(
+                                                stderr, "Failed to scrub %lx:%lx on OSD %lu (retval=%ld)\n",
+                                                obj.inode, obj.stripe, primary_osd, scrub_op->reply.hdr.retval
+                                            );
+                                        }
+                                        delete scrub_op;
+                                        parent->waiting--;
+                                        loop();
+                                    };
+                                    parent->cli->execute_raw(primary_osd, scrub_op);
+                                }
+                            };
+                            (*rm_count)++;
+                            parent->cli->execute_raw(rm_osd_num, rm_op);
+                        }
+                    }
+                    (*rm_count)--;
+                    if (!*rm_count)
+                    {
+                        free(rm_count);
+                        parent->waiting--;
+                        loop();
+                    }
+                }
+                delete op;
+            };
+            parent->waiting++;
+            parent->cli->execute_raw(primary_osd, op);
+        }
+        if (parent->waiting > 0)
+        {
+            return;
+        }
+        if (parent->json_output)
+        {
+            result.data = fix_result;
+        }
+        state = 100;
+    }
+};
+
+std::function<bool(cli_result_t &)> cli_tool_t::start_fix(json11::Json cfg)
+{
+    auto fixer = new cli_fix_t();
+    fixer->parent = this;
+    fixer->options = cfg;
+    return [fixer](cli_result_t & result)
+    {
+        fixer->loop();
+        if (fixer->is_done())
+        {
+            result = fixer->result;
+            delete fixer;
+            return true;
+        }
+        return false;
+    };
+}

src/cli_fix.h

@@ -0,0 +1,26 @@
+// Copyright (c) Vitaliy Filippov, 2019+
+// License: VNPL-1.1 (see README.md for details)
+
+#pragma once
+
+#include "cli.h"
+#include <algorithm>
+
+std::vector<uint64_t> parse_uint64_list(json11::Json val);
+
+template<class T> void remove_duplicates(std::vector<T> & ret)
+{
+    if (!ret.size())
+        return;
+    std::sort(ret.begin(), ret.end());
+    int j = 0;
+    for (int i = 1; i < ret.size(); i++)
+    {
+        if (ret[i] != ret[j])
+            ret[++j] = ret[i];
+    }
+    ret.resize(j+1);
+}
+
+// from http_client.cpp...
+bool json_is_false(const json11::Json & val);

src/cli_merge.cpp

@@ -41,7 +41,7 @@ struct snap_merger_t
     int fsync_interval = 128;
 
     // -- STATE --
-    inode_t target;
+    inode_t target, to_num;
     int target_rank;
     bool inside_continue = false;
     int state = 0;
@@ -98,6 +98,7 @@ struct snap_merger_t
             state = 100;
             return;
         }
+        to_num = to_cfg->num;
         // Check that to_cfg is actually a child of from_cfg and target_cfg is somewhere between them
         std::vector<inode_t> chain_list;
         inode_config_t *cur = to_cfg;
@@ -451,7 +452,7 @@ struct snap_merger_t
     {
         cluster_op_t *op = &rwo->op;
         op->opcode = OSD_OP_READ;
-        op->inode = target;
+        op->inode = to_num;
         op->offset = rwo->offset;
         op->len = target_block_size;
         op->iov.push_back(rwo->buf, target_block_size);
@@ -483,7 +484,7 @@ struct snap_merger_t
                 {
                     // write start->end
                     rwo->todo++;
-                    write_subop(rwo, rwo->start*gran, rwo->end*gran, use_cas ? 1+rwo->op.version : 0);
+                    write_subop(rwo, rwo->start*gran, rwo->end*gran, use_cas && to_num == target ? 1+rwo->op.version : 0);
                     rwo->start = rwo->end;
                     if (use_cas)
                     {
@@ -502,7 +503,7 @@ struct snap_merger_t
         {
             // write start->end
             rwo->todo++;
-            write_subop(rwo, rwo->start*gran, rwo->end*gran, use_cas ? 1+rwo->op.version : 0);
+            write_subop(rwo, rwo->start*gran, rwo->end*gran, use_cas && to_num == target ? 1+rwo->op.version : 0);
             rwo->start = rwo->end;
             if (use_cas)
             {
@@ -532,7 +533,7 @@ struct snap_merger_t
                 if (use_cas && subop->retval == -EINTR)
                 {
                     // CAS failure - reread and repeat optimistically
-                    rwo->start = subop->offset - rwo->offset;
+                    rwo->start = rwo->end = 0;
                     rwo_read(rwo);
                     delete subop;
                     return;
@@ -542,7 +543,7 @@ struct snap_merger_t
                 rwo->error_read = false;
             }
             // Increment CAS version
-            rwo->op.version++;
+            rwo->op.version = subop->version;
             if (use_cas)
                 next_write(rwo);
             else

src/cli_rm.cpp

@@ -65,6 +65,9 @@ struct snap_remover_t
     int current_child = 0;
     std::function<bool(cli_result_t &)> cb;
 
+    std::vector<std::string> rebased_images, deleted_images;
+    std::vector<uint64_t> deleted_ids;
+    std::string inverse_child_name, inverse_parent_name;
     cli_result_t result;
 
     bool is_done()
@@ -122,6 +125,7 @@ resume_1:
         {
             if (merge_children[current_child] == inverse_child)
                 continue;
+            rebased_images.push_back(parent->cli->st_cli.inode_config.at(merge_children[current_child]).name);
             start_merge_child(merge_children[current_child], merge_children[current_child]);
             if (state == 100)
                 return;
@@ -134,9 +138,12 @@ resume_2:
             cb = NULL;
             if (result.err)
             {
+                result.data = my_result(result.data);
                 state = 100;
                 return;
             }
+            else if (parent->progress)
+                printf("%s\n", result.text.c_str());
             parent->change_parent(merge_children[current_child], new_parent, &result);
             state = 3;
 resume_3:
@@ -144,6 +151,7 @@ resume_3:
                 return;
             if (result.err)
             {
+                result.data = my_result(result.data);
                 state = 100;
                 return;
             }
@@ -165,9 +173,12 @@ resume_4:
             cb = NULL;
             if (result.err)
             {
+                result.data = my_result(result.data);
                 state = 100;
                 return;
             }
+            else if (parent->progress)
+                printf("%s\n", result.text.c_str());
             // Delete "inverse" child data
             start_delete_source(inverse_child);
             if (state == 100)
@@ -181,9 +192,12 @@ resume_5:
             cb = NULL;
             if (result.err)
             {
+                result.data = my_result(result.data);
                 state = 100;
                 return;
             }
+            else if (parent->progress)
+                printf("%s\n", result.text.c_str());
             // Delete "inverse" child metadata, rename parent over it,
             // and also change parent links of the previous "inverse" child
             rename_inverse_parent();
@@ -199,6 +213,12 @@ resume_6:
         {
             if (chain_list[current_child] == inverse_parent)
                 continue;
+            {
+                auto parent_it = parent->cli->st_cli.inode_config.find(chain_list[current_child]);
+                if (parent_it != parent->cli->st_cli.inode_config.end())
+                    deleted_images.push_back(parent_it->second.name);
+                deleted_ids.push_back(chain_list[current_child]);
+            }
             start_delete_source(chain_list[current_child]);
 resume_7:
             while (!cb(result))
@@ -209,9 +229,12 @@ resume_7:
             cb = NULL;
             if (result.err)
             {
+                result.data = my_result(result.data);
                 state = 100;
                 return;
             }
+            else if (parent->progress)
+                printf("%s\n", result.text.c_str());
             delete_inode_config(chain_list[current_child]);
             if (state == 100)
                 return;
@@ -221,11 +244,26 @@ resume_8:
                 return;
         }
         state = 100;
+        result = (cli_result_t){
+            .text = "",
+            .data = my_result(result.data),
+        };
 resume_100:
         // Done
         return;
     }
 
+    json11::Json my_result(json11::Json src)
+    {
+        auto obj = src.object_items();
+        obj["deleted_ids"] = deleted_ids;
+        obj["deleted_images"] = deleted_images;
+        obj["rebased_images"] = rebased_images;
+        obj["renamed_from"] = inverse_parent_name;
+        obj["renamed_to"] = inverse_child_name;
+        return obj;
+    }
+
     void get_merge_children()
     {
         // Get all children of from..to
@@ -338,7 +376,11 @@ resume_100:
             }
             for (auto inode_result: data["responses"].array_items())
             {
-                auto kv = parent->cli->st_cli.parse_etcd_kv(inode_result["kvs"][0]);
+                if (inode_result["response_range"]["kvs"].array_items().size() == 0)
+                {
+                    continue;
+                }
+                auto kv = parent->cli->st_cli.parse_etcd_kv(inode_result["response_range"]["kvs"][0]);
                 pool_id_t pool_id = 0;
                 inode_t inode = 0;
                 char null_byte = 0;
@@ -377,7 +419,7 @@ resume_100:
             inode_t child = cp.first;
             uint64_t child_used = inode_used[child];
             int rank = cp.second;
-            for (int i = chain_list.size()-rank; i < chain_list.size(); i++)
+            for (int i = chain_list.size()-1-rank; i < chain_list.size(); i++)
             {
                 inode_t parent = chain_list[i];
                 uint64_t parent_used = inode_used[parent];
@@ -413,8 +455,8 @@ resume_100:
         }
         inode_config_t *child_cfg = &child_it->second;
         inode_config_t *target_cfg = &target_it->second;
-        std::string child_name = child_cfg->name;
-        std::string target_name = target_cfg->name;
+        inverse_child_name = child_cfg->name;
+        inverse_parent_name = target_cfg->name;
         std::string child_cfg_key = base64_encode(
             parent->cli->st_cli.etcd_prefix+
             "/config/inode/"+std::to_string(INODE_POOL(inverse_child))+
@@ -425,6 +467,9 @@ resume_100:
             "/config/inode/"+std::to_string(INODE_POOL(inverse_parent))+
             "/"+std::to_string(INODE_NO_POOL(inverse_parent))
         );
+        std::string target_idx_key = base64_encode(
+            parent->cli->st_cli.etcd_prefix+"/index/image/"+inverse_parent_name
+        );
         // Fill new configuration
         inode_config_t new_cfg = *child_cfg;
         new_cfg.num = target_cfg->num;
@@ -449,6 +494,11 @@ resume_100:
                     { "key", child_cfg_key },
                 } },
             },
+            json11::Json::object {
+                { "request_delete_range", json11::Json::object {
+                    { "key", target_idx_key },
+                } },
+            },
             json11::Json::object {
                 { "request_put", json11::Json::object {
                     { "key", target_cfg_key },
@@ -495,12 +545,12 @@ resume_100:
         parent->cli->st_cli.etcd_txn_slow(json11::Json::object {
             { "compare", cmp },
             { "success", txn },
-        }, [this, target_name, child_name](std::string err, json11::Json res)
+        }, [this](std::string err, json11::Json res)
         {
             parent->waiting--;
             if (err != "")
             {
-                result = (cli_result_t){ .err = EIO, .text = "Error renaming "+target_name+" to "+child_name+": "+err };
+                result = (cli_result_t){ .err = EIO, .text = "Error renaming "+inverse_parent_name+" to "+inverse_child_name+": "+err };
                 state = 100;
                 return;
             }
@@ -508,14 +558,14 @@ resume_100:
             {
                 result = (cli_result_t){
                     .err = EAGAIN,
-                    .text = "Parent ("+target_name+"), child ("+child_name+"), or one of its children"
+                    .text = "Parent ("+inverse_parent_name+"), child ("+inverse_child_name+"), or one of its children"
                         " configuration was modified during rename",
                 };
                 state = 100;
                 return;
             }
             if (parent->progress)
-                printf("Layer %s renamed to %s\n", target_name.c_str(), child_name.c_str());
+                printf("Layer %s renamed to %s\n", inverse_parent_name.c_str(), inverse_child_name.c_str());
             parent->ringloop->wakeup();
         });
     }

src/cli_rm_data.cpp

@@ -28,6 +28,7 @@ struct rm_inode_t
     cli_tool_t *parent = NULL;
     inode_list_t *lister = NULL;
     std::vector<rm_pg_t*> lists;
+    std::vector<osd_num_t> inactive_osds;
     uint64_t total_count = 0, total_done = 0, total_prev_pct = 0;
     uint64_t pgs_to_list = 0;
     bool lists_done = false;
@@ -86,6 +87,16 @@ struct rm_inode_t
             state = 100;
             return;
         }
+        inactive_osds = parent->cli->list_inode_get_inactive_osds(lister);
+        if (inactive_osds.size() && !parent->json_output)
+        {
+            fprintf(stderr, "Some data may remain after delete on OSDs which are currently down: ");
+            for (int i = 0; i < inactive_osds.size(); i++)
+            {
+                fprintf(stderr, i > 0 ? ", %lu" : "%lu", inactive_osds[i]);
+            }
+            fprintf(stderr, "\n");
+        }
         pgs_to_list = parent->cli->list_pg_count(lister);
         parent->cli->list_inode_next(lister, parent->parallel_osds);
     }
@@ -167,16 +178,33 @@ struct rm_inode_t
         }
         if (parent->progress && total_count > 0 && total_done*1000/total_count != total_prev_pct)
         {
-            printf("\rRemoved %lu/%lu objects, %lu more PGs to list...", total_done, total_count, pgs_to_list);
+            fprintf(stderr, "\rRemoved %lu/%lu objects, %lu more PGs to list...", total_done, total_count, pgs_to_list);
             total_prev_pct = total_done*1000/total_count;
         }
         if (lists_done && !lists.size())
         {
+            if (parent->progress && total_count > 0)
+            {
+                fprintf(stderr, "\n");
+            }
+            if (parent->progress && (total_done < total_count || inactive_osds.size() > 0))
+            {
+                fprintf(
+                    stderr, "Warning: Pool:%u,ID:%lu inode data may not have been fully removed.\n"
+                    " Use `vitastor-cli rm-data --pool %u --inode %lu` if you encounter it in listings.\n",
+                    pool_id, INODE_NO_POOL(inode), pool_id, INODE_NO_POOL(inode)
+                );
+            }
             result = (cli_result_t){
                 .err = error_count > 0 ? EIO : 0,
                 .text = error_count > 0 ? "Some blocks were not removed" : (
                     "Done, inode "+std::to_string(INODE_NO_POOL(inode))+" from pool "+
                     std::to_string(pool_id)+" removed"),
+                .data = json11::Json::object {
+                    { "removed_objects", total_done },
+                    { "total_objects", total_count },
+                    { "inactive_osds", inactive_osds },
+                },
             };
             state = 100;
         }

src/cli_rm_osd.cpp

@@ -410,14 +410,17 @@ struct rm_osd_t
                         parent->cli->st_cli.etcd_prefix+"/pg/history/"+
                         std::to_string(pool_cfg.id)+"/"+std::to_string(pg_num)
                     );
+                    auto hist = json11::Json::object {
+                        { "epoch", pg_cfg.epoch },
+                        { "all_peers", pg_cfg.all_peers },
+                        { "osd_sets", pg_cfg.target_history },
+                    };
+                    if (pg_cfg.next_scrub)
+                        hist["next_scrub"] = pg_cfg.next_scrub;
                     history_updates.push_back(json11::Json::object {
                         { "request_put", json11::Json::object {
                             { "key", history_key },
-                            { "value", base64_encode(json11::Json(json11::Json::object {
-                                { "epoch", pg_cfg.epoch },
-                                { "all_peers", pg_cfg.all_peers },
-                                { "osd_sets", pg_cfg.target_history },
-                            }).dump()) },
+                            { "value", base64_encode(json11::Json(hist).dump()) },
                         } },
                     });
                     history_checks.push_back(json11::Json::object {

src/cli_status.cpp

@@ -12,7 +12,7 @@ static const char *obj_states[] = { "clean", "misplaced", "degraded", "incomplet
 // Print cluster status:
 // etcd, mon, osd states
 // raw/used space, object states, pool states, pg states
-// client io, recovery io, rebalance io
+// client io, recovery io, rebalance io, scrub io
 struct status_printer_t
 {
     cli_tool_t *parent;
@@ -201,6 +201,7 @@ resume_2:
         bool readonly = json_is_true(parent->cli->config["readonly"]);
         bool no_recovery = json_is_true(parent->cli->config["no_recovery"]);
         bool no_rebalance = json_is_true(parent->cli->config["no_rebalance"]);
+        bool no_scrub = json_is_true(parent->cli->config["no_scrub"]);
         if (parent->json_output)
         {
             // JSON output
@@ -219,6 +220,7 @@ resume_2:
                 { "readonly", readonly },
                 { "no_recovery", no_recovery },
                 { "no_rebalance", no_rebalance },
+                { "no_scrub", no_scrub },
                 { "pool_count", pool_count },
                 { "active_pool_count", pools_active },
                 { "pg_states", pgs_by_state },
@@ -250,18 +252,25 @@ resume_2:
         }
         more_states.resize(more_states.size()-2);
         std::string recovery_io;
+        int io_indent = 0;
         {
             uint64_t deg_bps = agg_stats["recovery_stats"]["degraded"]["bps"].uint64_value();
             uint64_t deg_iops = agg_stats["recovery_stats"]["degraded"]["iops"].uint64_value();
             uint64_t misp_bps = agg_stats["recovery_stats"]["misplaced"]["bps"].uint64_value();
             uint64_t misp_iops = agg_stats["recovery_stats"]["misplaced"]["iops"].uint64_value();
+            uint64_t scrub_bps = agg_stats["op_stats"]["scrub"]["bps"].uint64_value();
+            uint64_t scrub_iops = agg_stats["op_stats"]["scrub"]["iops"].uint64_value();
+            if (misp_iops > 0 || misp_bps > 0 || no_rebalance)
+                io_indent = 3;
+            else if (deg_iops > 0 || deg_bps > 0 || no_recovery)
+                io_indent = 2;
             if (deg_iops > 0 || deg_bps > 0)
             {
-                recovery_io += "    recovery:  "+std::string(no_recovery ? "disabled, " : "")+
+                recovery_io += "    recovery: "+str_repeat(" ", io_indent-2)+std::string(no_recovery ? "disabled, " : "")+
                     format_size(deg_bps)+"/s, "+format_size(deg_iops, true)+" op/s\n";
             }
             else if (no_recovery)
-                recovery_io += "    recovery:  disabled\n";
+                recovery_io += "    recovery: disabled\n";
             if (misp_iops > 0 || misp_bps > 0)
             {
                 recovery_io += "    rebalance: "+std::string(no_rebalance ? "disabled, " : "")+
@@ -269,6 +278,13 @@ resume_2:
             }
             else if (no_rebalance)
                 recovery_io += "    rebalance: disabled\n";
+            if (scrub_iops > 0 || scrub_bps > 0)
+            {
+                recovery_io += "    scrub: "+str_repeat(" ", io_indent+1)+std::string(no_scrub ? "disabled, " : "")+
+                    format_size(scrub_bps)+"/s, "+format_size(scrub_iops, true)+" op/s\n";
+            }
+            else if (no_scrub)
+                recovery_io += "    scrub: "+str_repeat(" ", io_indent+1)+"disabled\n";
         }
         printf(
             "  cluster:\n"
@@ -296,7 +312,7 @@ resume_2:
             pools_active, pool_count,
             pgs_by_state_str.c_str(),
             readonly ? " (read-only mode)" : "",
-            recovery_io.size() > 0 ? "   " : "",
+            str_repeat(" ", io_indent).c_str(),
             format_size(agg_stats["op_stats"]["primary_read"]["bps"].uint64_value()).c_str(),
             format_size(agg_stats["op_stats"]["primary_read"]["iops"].uint64_value(), true).c_str(),
             format_size(agg_stats["op_stats"]["primary_write"]["bps"].uint64_value()).c_str(),

src/cluster_client.cpp

@@ -35,6 +35,7 @@ cluster_client_t::cluster_client_t(ring_loop_t *ringloop, timerfd_manager_t *tfd
             // peer_osd just connected
             continue_ops();
             continue_lists();
+            continue_raw_ops(peer_osd);
         }
         else if (dirty_buffers.size())
         {
@@ -104,6 +105,19 @@ cluster_op_t::~cluster_op_t()
     }
 }
 
+void cluster_client_t::continue_raw_ops(osd_num_t peer_osd)
+{
+    auto it = raw_ops.find(peer_osd);
+    while (it != raw_ops.end() && it->first == peer_osd)
+    {
+        auto op = it->second;
+        op->op_type = OSD_OP_OUT;
+        op->peer_fd = msgr.osd_peer_fds.at(peer_osd);
+        msgr.outbox_push(op);
+        raw_ops.erase(it++);
+    }
+}
+
 void cluster_client_t::init_msgr()
 {
     if (msgr_initialized)
@@ -512,6 +526,23 @@ void cluster_client_t::execute(cluster_op_t *op)
     }
 }
 
+void cluster_client_t::execute_raw(osd_num_t osd_num, osd_op_t *op)
+{
+    auto fd_it = msgr.osd_peer_fds.find(osd_num);
+    if (fd_it != msgr.osd_peer_fds.end())
+    {
+        op->op_type = OSD_OP_OUT;
+        op->peer_fd = fd_it->second;
+        msgr.outbox_push(op);
+    }
+    else
+    {
+        if (msgr.wanted_peers.find(osd_num) == msgr.wanted_peers.end())
+            msgr.connect_peer(osd_num, st_cli.peer_states[osd_num]);
+        raw_ops.emplace(osd_num, op);
+    }
+}
+
 void cluster_client_t::copy_write(cluster_op_t *op, std::map<object_id, cluster_buffer_t> & dirty_buffers)
 {
     // Save operation for replay when one of PGs goes out of sync
@@ -1178,6 +1209,10 @@ void cluster_client_t::handle_op_part(cluster_op_part_t *part)
             copy_part_bitmap(op, part);
             op->version = op->parts.size() == 1 ? part->op.reply.rw.version : 0;
         }
+        else if (op->opcode == OSD_OP_WRITE)
+        {
+            op->version = op->parts.size() == 1 ? part->op.reply.rw.version : 0;
+        }
         if (op->inflight_count == 0)
         {
             if (op->opcode == OSD_OP_SYNC)

src/cluster_client.h

@@ -103,6 +103,7 @@ class cluster_client_t
     ring_consumer_t consumer;
     std::vector<std::function<void(void)>> on_ready_hooks;
     std::vector<inode_list_t*> lists;
+    std::multimap<osd_num_t, osd_op_t*> raw_ops;
     int continuing_ops = 0;
     bool msgr_initialized = false;
 
@@ -118,6 +119,7 @@ public:
     cluster_client_t(ring_loop_t *ringloop, timerfd_manager_t *tfd, json11::Json & config);
     ~cluster_client_t();
     void execute(cluster_op_t *op);
+    void execute_raw(osd_num_t osd_num, osd_op_t *op);
     bool is_ready();
     void on_ready(std::function<void(void)> fn);
 
@@ -128,6 +130,7 @@ public:
     inode_list_t *list_inode_start(inode_t inode,
         std::function<void(inode_list_t* lst, std::set<object_id>&& objects, pg_num_t pg_num, osd_num_t primary_osd, int status)> callback);
     int list_pg_count(inode_list_t *lst);
+    const std::vector<osd_num_t> & list_inode_get_inactive_osds(inode_list_t *lst);
     void list_inode_next(inode_list_t *lst, int next_pgs);
     //inline uint32_t get_bs_bitmap_granularity() { return st_cli.global_bitmap_granularity; }
     //inline uint64_t get_bs_block_size() { return st_cli.global_block_size; }
@@ -153,4 +156,5 @@ protected:
     void continue_lists();
     void continue_listing(inode_list_t *lst);
     void send_list(inode_list_osd_t *cur_list);
+    void continue_raw_ops(osd_num_t peer_osd);
 };

src/cluster_client_list.cpp

@@ -36,6 +36,7 @@ struct inode_list_t
     inode_t inode = 0;
     int done_pgs = 0;
     int want = 0;
+    std::vector<osd_num_t> inactive_osds;
     std::vector<inode_list_pg_t*> pgs;
     std::function<void(inode_list_t* lst, std::set<object_id>&& objects, pg_num_t pg_num, osd_num_t primary_osd, int status)> callback;
 };
@@ -60,6 +61,7 @@ inode_list_t* cluster_client_t::list_inode_start(inode_t inode,
     lst->inode = inode;
     lst->callback = callback;
     auto pool_cfg = st_cli.pool_config[pool_id];
+    std::set<osd_num_t> inactive_osd_set;
     for (auto & pg_item: pool_cfg.pg_config)
     {
         auto & pg = pg_item.second;
@@ -106,11 +108,18 @@ inode_list_t* cluster_client_t::list_inode_start(inode_t inode,
             }
             for (osd_num_t peer_osd: all_peers)
             {
-                r->list_osds.push_back((inode_list_osd_t){
-                    .pg = r,
-                    .osd_num = peer_osd,
-                    .sent = false,
-                });
+                if (st_cli.peer_states.find(peer_osd) != st_cli.peer_states.end())
+                {
+                    r->list_osds.push_back((inode_list_osd_t){
+                        .pg = r,
+                        .osd_num = peer_osd,
+                        .sent = false,
+                    });
+                }
+                else
+                {
+                    inactive_osd_set.insert(peer_osd);
+                }
             }
         }
         else
@@ -132,6 +141,7 @@ inode_list_t* cluster_client_t::list_inode_start(inode_t inode,
     {
         lst->pgs[i]->pos = i;
     }
+    lst->inactive_osds.insert(lst->inactive_osds.end(), inactive_osd_set.begin(), inactive_osd_set.end());
     lists.push_back(lst);
     return lst;
 }
@@ -141,6 +151,11 @@ int cluster_client_t::list_pg_count(inode_list_t *lst)
     return lst->pgs.size();
 }
 
+const std::vector<osd_num_t> & cluster_client_t::list_inode_get_inactive_osds(inode_list_t *lst)
+{
+    return lst->inactive_osds;
+}
+
 void cluster_client_t::list_inode_next(inode_list_t *lst, int next_pgs)
 {
     if (next_pgs >= 0)

src/disk_tool_utils.cpp

@@ -55,23 +55,6 @@ std::string realpath_str(std::string path, bool nofail)
     return rp;
 }
 
-std::string read_all_fd(int fd)
-{
-    int res_size = 0;
-    std::string res;
-    while (1)
-    {
-        res.resize(res_size+1024);
-        int r = read(fd, (char*)res.data()+res_size, res.size()-res_size);
-        if (r > 0)
-            res_size += r;
-        else if (!r || errno != EAGAIN && errno != EINTR)
-            break;
-    }
-    res.resize(res_size);
-    return res;
-}
-
 std::string read_file(std::string file, bool allow_enoent)
 {
     std::string res;

src/etcd_state_client.cpp

@@ -7,8 +7,8 @@
 #ifndef __MOCK__
 #include "addr_util.h"
 #include "http_client.h"
-#include "str_util.h"
 #endif
+#include "str_util.h"
 
 etcd_state_client_t::~etcd_state_client_t()
 {
@@ -777,6 +777,10 @@ void etcd_state_client_t::parse_state(const etcd_kv_t & kv)
                 fprintf(stderr, "Pool %u has invalid bitmap_granularity (must divide block_size), skipping pool\n", pool_id);
                 continue;
             }
+            // Scrub Interval
+            pc.scrub_interval = parse_time(pool_item.second["scrub_interval"].string_value());
+            if (!pc.scrub_interval)
+                pc.scrub_interval = 0;
             // Immediate Commit Mode
             pc.immediate_commit = pool_item.second["immediate_commit"].is_string()
                 ? (pool_item.second["immediate_commit"].string_value() == "all"
@@ -919,6 +923,8 @@ void etcd_state_client_t::parse_state(const etcd_kv_t & kv)
             }
             // Read epoch
             pg_cfg.epoch = value["epoch"].uint64_value();
+            // Next scrub timestamp (0 or empty = scrub is not needed)
+            pg_cfg.next_scrub = value["next_scrub"].uint64_value();
             if (on_change_pg_history_hook != NULL)
             {
                 on_change_pg_history_hook(pool_id, pg_num);

src/etcd_state_client.h

@@ -39,6 +39,7 @@ struct pg_config_t
     osd_num_t cur_primary;
     int cur_state;
     uint64_t epoch;
+    uint64_t next_scrub;
 };
 
 struct pool_config_t
@@ -55,6 +56,7 @@ struct pool_config_t
     uint64_t max_osd_combinations;
     uint64_t pg_stripe_size;
     std::map<pg_num_t, pg_config_t> pg_config;
+    uint64_t scrub_interval;
 };
 
 struct inode_config_t

src/messenger.cpp

@@ -251,6 +251,10 @@ void osd_messenger_t::try_connect_peer_addr(osd_num_t peer_osd, const char *peer
         return;
     }
     clients[peer_fd] = new osd_client_t();
+    if (log_level > 0)
+    {
+        fprintf(stderr, "Connecting to OSD %lu at %s:%d (client %d)\n", peer_osd, peer_host, peer_port, peer_fd);
+    }
     clients[peer_fd]->peer_addr = addr;
     clients[peer_fd]->peer_port = peer_port;
     clients[peer_fd]->peer_fd = peer_fd;
@@ -313,7 +317,10 @@ void osd_messenger_t::handle_peer_epoll(int peer_fd, int epoll_events)
     if (epoll_events & EPOLLRDHUP)
     {
         // Stop client
-        fprintf(stderr, "[OSD %lu] client %d disconnected\n", this->osd_num, peer_fd);
+        if (log_level > 0)
+        {
+            fprintf(stderr, "[OSD %lu] client %d disconnected\n", this->osd_num, peer_fd);
+        }
         stop_client(peer_fd, true);
     }
     else if (epoll_events & EPOLLIN)

src/messenger.h

@@ -50,7 +50,7 @@ struct osd_client_t
 
     sockaddr_storage peer_addr;
     int peer_port;
-    int peer_fd;
+    int peer_fd = -1;
     int peer_state;
     int connect_timeout_id = -1;
     int ping_time_remaining = 0;
@@ -87,11 +87,7 @@ struct osd_client_t
     std::vector<iovec> send_list, next_send_list;
     std::vector<msgr_sendp_t> outbox, next_outbox;
 
-    ~osd_client_t()
-    {
-        free(in_buf);
-        in_buf = NULL;
-    }
+    ~osd_client_t();
 };
 
 struct osd_wanted_peer_t
@@ -136,6 +132,7 @@ protected:
     uint64_t rdma_max_msg = 0;
 #endif
 
+    bool has_send_loop = false;
     std::vector<int> read_ready_clients;
     std::vector<int> write_ready_clients;
     // We don't use ringloop->set_immediate here because we may have no ringloop in client :)
@@ -163,6 +160,7 @@ public:
     std::function<bool(osd_client_t*, json11::Json)> check_config_hook;
     void read_requests();
     void send_replies();
+    bool can_send();
     void accept_connections(int listen_fd);
     ~osd_messenger_t();
 
@@ -177,6 +175,8 @@ public:
     bool connect_rdma(int peer_fd, std::string rdma_address, uint64_t client_max_msg);
 #endif
 
+    void measure_exec(osd_op_t *cur_op);
+
 protected:
     void try_connect_peer(uint64_t osd_num);
     void try_connect_peer_addr(osd_num_t peer_osd, const char *peer_host, int peer_port);
@@ -188,7 +188,6 @@ protected:
     void cancel_op(osd_op_t *op);
 
     bool try_send(osd_client_t *cl);
-    void measure_exec(osd_op_t *cur_op);
     void handle_send(int result, osd_client_t *cl);
 
     bool handle_read(int result, osd_client_t *cl);

src/msgr_op.cpp

@@ -9,6 +9,10 @@ osd_op_t::~osd_op_t()
 {
     assert(!bs_op);
     assert(!op_data);
+    if (bitmap_buf)
+    {
+        free(bitmap_buf);
+    }
     if (rmw_buf)
     {
         free(rmw_buf);

src/msgr_op.h

@@ -165,6 +165,7 @@ struct osd_op_t
     void *bitmap = NULL;
     unsigned bitmap_len = 0;
     unsigned bmp_data = 0;
+    void *bitmap_buf = NULL;
     void *rmw_buf = NULL;
     osd_primary_op_data_t* op_data = NULL;
     std::function<void(osd_op_t*)> callback;

src/msgr_receive.cpp

@@ -251,10 +251,6 @@ void osd_messenger_t::handle_op_hdr(osd_client_t *cl)
         }
         cl->read_remaining = cur_op->req.sec_read_bmp.len;
     }
-    else if (cur_op->req.hdr.opcode == OSD_OP_READ)
-    {
-        cl->read_remaining = 0;
-    }
     else if (cur_op->req.hdr.opcode == OSD_OP_WRITE)
     {
         if (cur_op->req.rw.len > 0)
@@ -274,6 +270,12 @@ void osd_messenger_t::handle_op_hdr(osd_client_t *cl)
         }
         cl->read_remaining = cur_op->req.show_conf.json_len;
     }
+    /*else if (cur_op->req.hdr.opcode == OSD_OP_READ ||
+        cur_op->req.hdr.opcode == OSD_OP_SCRUB ||
+        cur_op->req.hdr.opcode == OSD_OP_DESCRIBE)
+    {
+        cl->read_remaining = 0;
+    }*/
     if (cl->read_remaining > 0)
     {
         // Read data
@@ -367,6 +369,16 @@ bool osd_messenger_t::handle_reply_hdr(osd_client_t *cl)
         op->buf = malloc_or_die(op->reply.hdr.retval);
         cl->recv_list.push_back(op->buf, op->reply.hdr.retval);
     }
+    else if (op->reply.hdr.opcode == OSD_OP_DESCRIBE && op->reply.describe.result_bytes > 0)
+    {
+        delete cl->read_op;
+        cl->read_op = op;
+        cl->read_state = CL_READ_REPLY_DATA;
+        cl->read_remaining = op->reply.describe.result_bytes;
+        free(op->buf);
+        op->buf = malloc_or_die(op->reply.describe.result_bytes);
+        cl->recv_list.push_back(op->buf, op->reply.describe.result_bytes);
+    }
     else
     {
 reuse:

src/msgr_send.cpp

@@ -73,7 +73,8 @@ void osd_messenger_t::outbox_push(osd_op_t *cur_op)
         ? (cur_op->req.hdr.opcode == OSD_OP_READ ||
         cur_op->req.hdr.opcode == OSD_OP_SEC_READ ||
         cur_op->req.hdr.opcode == OSD_OP_SEC_LIST ||
-        cur_op->req.hdr.opcode == OSD_OP_SHOW_CONFIG)
+        cur_op->req.hdr.opcode == OSD_OP_SHOW_CONFIG ||
+        cur_op->req.hdr.opcode == OSD_OP_DESCRIBE)
         : (cur_op->req.hdr.opcode == OSD_OP_WRITE ||
         cur_op->req.hdr.opcode == OSD_OP_SEC_WRITE ||
         cur_op->req.hdr.opcode == OSD_OP_SEC_WRITE_STABLE ||
@@ -83,9 +84,12 @@ void osd_messenger_t::outbox_push(osd_op_t *cur_op)
     {
         for (int i = 0; i < cur_op->iov.count; i++)
         {
-            assert(cur_op->iov.buf[i].iov_base);
-            to_send_list.push_back(cur_op->iov.buf[i]);
-            to_outbox.push_back((msgr_sendp_t){ .op = cur_op, .flags = 0 });
+            if (cur_op->iov.buf[i].iov_len > 0)
+            {
+                assert(cur_op->iov.buf[i].iov_base);
+                to_send_list.push_back(cur_op->iov.buf[i]);
+                to_outbox.push_back((msgr_sendp_t){ .op = cur_op, .flags = 0 });
+            }
         }
     }
     if (cur_op->req.hdr.opcode == OSD_OP_SEC_READ_BMP)
@@ -107,22 +111,28 @@ void osd_messenger_t::outbox_push(osd_op_t *cur_op)
         return;
     }
 #endif
-    if (!ringloop)
+    if (!ringloop && !has_send_loop)
     {
-        // FIXME: It's worse because it doesn't allow batching
+        // "Send loop" should be used, like in QEMU driver, or performance will suffer
+        // due to a large number of sendmsg() syscalls where they can be avoided.
+        // "Send loop" increases 4k T1Q128 randread from ~55k iops to ~90k on a 1 OSD.
+        // The difference is smaller when there are more OSDs though...
         while (cl->outbox.size())
         {
             try_send(cl);
         }
     }
-    else if (cl->write_msg.msg_iovlen > 0 || !try_send(cl))
+    else if (cl->write_msg.msg_iovlen > 0 || has_send_loop || !try_send(cl))
     {
         if (cl->write_state == 0)
         {
             cl->write_state = CL_WRITE_READY;
             write_ready_clients.push_back(cur_op->peer_fd);
         }
-        ringloop->wakeup();
+        if (ringloop)
+        {
+            ringloop->wakeup();
+        }
     }
 }
 
@@ -149,8 +159,10 @@ void osd_messenger_t::measure_exec(osd_op_t *cur_op)
         (cur_op->tv_end.tv_nsec - cur_op->tv_begin.tv_nsec)/1000
     );
     if (cur_op->req.hdr.opcode == OSD_OP_READ ||
-        cur_op->req.hdr.opcode == OSD_OP_WRITE)
+        cur_op->req.hdr.opcode == OSD_OP_WRITE ||
+        cur_op->req.hdr.opcode == OSD_OP_SCRUB)
     {
+        // req.rw.len is internally set to the full object size for scrubs
         stats.op_stat_bytes[cur_op->req.hdr.opcode] += cur_op->req.rw.len;
     }
     else if (cur_op->req.hdr.opcode == OSD_OP_SEC_READ ||
@@ -212,6 +224,12 @@ void osd_messenger_t::send_replies()
     write_ready_clients.clear();
 }
 
+bool osd_messenger_t::can_send()
+{
+    has_send_loop = true;
+    return write_ready_clients.size() > 0;
+}
+
 void osd_messenger_t::handle_send(int result, osd_client_t *cl)
 {
     cl->write_msg.msg_iovlen = 0;

src/msgr_stop.cpp

@@ -122,17 +122,6 @@ void osd_messenger_t::stop_client(int peer_fd, bool force, bool force_delete)
         // Cancel outbound operations
         cancel_osd_ops(cl);
     }
-#ifndef __MOCK__
-    // And close the FD only when everything is done
-    // ...because peer_fd number can get reused after close()
-    close(peer_fd);
-#ifdef WITH_RDMA
-    if (cl->rdma_conn)
-    {
-        delete cl->rdma_conn;
-    }
-#endif
-#endif
     // Find the item again because it can be invalidated at this point
     it = clients.find(peer_fd);
     if (it != clients.end())
@@ -145,3 +134,25 @@ void osd_messenger_t::stop_client(int peer_fd, bool force, bool force_delete)
         delete cl;
     }
 }
+
+osd_client_t::~osd_client_t()
+{
+    free(in_buf);
+    in_buf = NULL;
+    if (peer_fd >= 0)
+    {
+        // Close the FD only when the client is actually destroyed
+        // Which only happens when all references are cleared
+        close(peer_fd);
+        peer_fd = -1;
+    }
+#ifndef __MOCK__
+#ifdef WITH_RDMA
+    if (rdma_conn)
+    {
+        delete rdma_conn;
+        rdma_conn = NULL;
+    }
+#endif
+#endif
+}

src/nbd_proxy.cpp

@@ -137,12 +137,19 @@ public:
             "OPTIONS:\n"
             "  All usual Vitastor config options like --etcd_address <etcd_address> plus NBD-specific:\n"
             "  --nbd_timeout 30\n"
-            "    timeout in seconds after which the kernel will stop the device\n"
-            "    you can set it to 0, but beware that you won't be able to stop the device at all\n"
-            "    if vitastor-nbd process dies\n"
+            "    Timeout for I/O operations in seconds after exceeding which the kernel stops\n"
+            "    the device. You can set it to 0 to disable the timeout, but beware that you\n"
+            "    won't be able to stop the device at all if vitastor-nbd process dies.\n"
             "  --nbd_max_devices 64 --nbd_max_part 3\n"
-            "    options for the \"nbd\" kernel module when modprobing it (nbds_max and max_part).\n"
-            "    note that maximum allowed (nbds_max)*(1+max_part) is 256.\n",
+            "    Options for the \"nbd\" kernel module when modprobing it (nbds_max and max_part).\n"
+            "    note that maximum allowed (nbds_max)*(1+max_part) is 256.\n"
+            "  --logfile /path/to/log/file.txt\n"
+            "    Wite log messages to the specified file instead of dropping them (in background mode)\n"
+            "    or printing them to the standard output (in foreground mode).\n"
+            "  --dev_num N\n"
+            "    Use the specified device /dev/nbdN instead of automatic selection.\n"
+            "  --foreground 1\n"
+            "    Stay in foreground, do not daemonize.\n",
             exe_name, exe_name, exe_name
         );
         exit(0);

src/osd.cpp

@@ -13,6 +13,7 @@
 #include "osd_primary.h"
 #include "osd.h"
 #include "http_client.h"
+#include "str_util.h"
 
 static blockstore_config_t json_to_bs(const json11::Json::object & config)
 {
@@ -168,6 +169,8 @@ void osd_t::parse_config(bool init)
     no_rebalance = json_is_true(config["no_rebalance"]);
     auto old_no_recovery = no_recovery;
     no_recovery = json_is_true(config["no_recovery"]);
+    auto old_no_scrub = no_scrub;
+    no_scrub = json_is_true(config["no_scrub"]);
     auto old_autosync_interval = autosync_interval;
     if (!config["autosync_interval"].is_null())
     {
@@ -207,6 +210,38 @@ void osd_t::parse_config(bool init)
     inode_vanish_time = config["inode_vanish_time"].uint64_value();
     if (!inode_vanish_time)
         inode_vanish_time = 60;
+    auto old_auto_scrub = auto_scrub;
+    auto_scrub = json_is_true(config["auto_scrub"]);
+    global_scrub_interval = parse_time(config["scrub_interval"].string_value());
+    if (!global_scrub_interval)
+        global_scrub_interval = 30*86400;
+    scrub_queue_depth = config["scrub_queue_depth"].uint64_value();
+    if (scrub_queue_depth < 1 || scrub_queue_depth > MAX_RECOVERY_QUEUE)
+        scrub_queue_depth = 1;
+    scrub_find_best = !json_is_false(config["scrub_find_best"]);
+    scrub_ec_max_bruteforce = config["scrub_ec_max_bruteforce"].uint64_value();
+    if (scrub_ec_max_bruteforce < 1)
+        scrub_ec_max_bruteforce = 100;
+    scrub_sleep_ms = config["scrub_sleep"].uint64_value();
+    scrub_list_limit = config["scrub_list_limit"].uint64_value();
+    if (!scrub_list_limit)
+        scrub_list_limit = 1000;
+    if (!old_auto_scrub && auto_scrub)
+    {
+        // Schedule scrubbing
+        for (auto & pgp: pgs)
+        {
+            plan_scrub(pgp.second);
+        }
+    }
+    if (old_no_scrub && !no_scrub)
+    {
+        // Wakeup scrubbing
+        for (auto & pgp: pgs)
+        {
+            schedule_scrub(pgp.second);
+        }
+    }
     if ((old_no_rebalance && !no_rebalance || old_no_recovery && !no_recovery) &&
         !(peering_state & (OSD_RECOVERING | OSD_FLUSHING_PGS)))
     {
@@ -337,6 +372,8 @@ void osd_t::exec_op(osd_op_t *cur_op)
         cur_op->req.hdr.opcode != OSD_OP_SEC_LIST &&
         cur_op->req.hdr.opcode != OSD_OP_READ &&
         cur_op->req.hdr.opcode != OSD_OP_SEC_READ_BMP &&
+        cur_op->req.hdr.opcode != OSD_OP_SCRUB &&
+        cur_op->req.hdr.opcode != OSD_OP_DESCRIBE &&
         cur_op->req.hdr.opcode != OSD_OP_SHOW_CONFIG)
     {
         // Readonly mode
@@ -367,6 +404,14 @@ void osd_t::exec_op(osd_op_t *cur_op)
     {
         continue_primary_del(cur_op);
     }
+    else if (cur_op->req.hdr.opcode == OSD_OP_SCRUB)
+    {
+        continue_primary_scrub(cur_op);
+    }
+    else if (cur_op->req.hdr.opcode == OSD_OP_DESCRIBE)
+    {
+        continue_primary_describe(cur_op);
+    }
     else
     {
         exec_secondary(cur_op);
@@ -431,6 +476,10 @@ void osd_t::print_stats()
             recovery_stat_bytes[1][i] = recovery_stat_bytes[0][i];
         }
     }
+    if (corrupted_objects > 0)
+    {
+        printf("[OSD %lu] %lu object(s) corrupted\n", osd_num, corrupted_objects);
+    }
     if (incomplete_objects > 0)
     {
         printf("[OSD %lu] %lu object(s) incomplete\n", osd_num, incomplete_objects);
@@ -498,10 +547,11 @@ void osd_t::print_slow()
                 else if (op->req.hdr.opcode == OSD_OP_SEC_LIST)
                 {
                     bufprintf(
-                        " inode=%lx-%lx pg=%u/%u, stripe=%lu",
-                        op->req.sec_list.min_inode, op->req.sec_list.max_inode,
+                        " oid=%lx/%lx-%lx/%lx pg=%u/%u, stripe=%lu, limit=%u",
+                        op->req.sec_list.min_inode, op->req.sec_list.min_stripe,
+                        op->req.sec_list.max_inode, op->req.sec_list.max_stripe,
                         op->req.sec_list.list_pg, op->req.sec_list.pg_count,
-                        op->req.sec_list.pg_stripe_size
+                        op->req.sec_list.pg_stripe_size, op->req.sec_list.stable_limit
                     );
                 }
                 else if (op->req.hdr.opcode == OSD_OP_READ || op->req.hdr.opcode == OSD_OP_WRITE ||

src/osd.h

@@ -28,6 +28,7 @@
 #define OSD_PEERING_PGS 0x04
 #define OSD_FLUSHING_PGS 0x08
 #define OSD_RECOVERING 0x10
+#define OSD_SCRUBBING 0x20
 
 #define MAX_AUTOSYNC_INTERVAL 3600
 #define DEFAULT_AUTOSYNC_INTERVAL 5
@@ -98,6 +99,7 @@ class osd_t
     bool run_primary = false;
     bool no_rebalance = false;
     bool no_recovery = false;
+    bool no_scrub = false;
     std::string bind_address;
     int bind_port, listen_backlog = 128;
     // FIXME: Implement client queue depth limit
@@ -113,6 +115,13 @@ class osd_t
     int recovery_sync_batch = DEFAULT_RECOVERY_BATCH;
     int inode_vanish_time = 60;
     int log_level = 0;
+    bool auto_scrub = false;
+    uint64_t global_scrub_interval = 30*86400;
+    uint64_t scrub_queue_depth = 1;
+    uint64_t scrub_sleep_ms = 0;
+    uint32_t scrub_list_limit = 1000;
+    bool scrub_find_best = true;
+    uint64_t scrub_ec_max_bruteforce = 100;
 
     // cluster state
 
@@ -135,15 +144,24 @@ class osd_t
     std::set<pool_pg_num_t> dirty_pgs;
     std::set<osd_num_t> dirty_osds;
     int copies_to_delete_after_sync_count = 0;
-    uint64_t misplaced_objects = 0, degraded_objects = 0, incomplete_objects = 0;
+    uint64_t misplaced_objects = 0, degraded_objects = 0, incomplete_objects = 0, inconsistent_objects = 0, corrupted_objects = 0;
     int peering_state = 0;
     std::map<object_id, osd_recovery_op_t> recovery_ops;
+    std::map<object_id, osd_op_t*> scrub_ops;
     bool recovery_last_degraded = true;
     pool_pg_num_t recovery_last_pg;
     object_id recovery_last_oid;
     int recovery_pg_done = 0, recovery_done = 0;
     osd_op_t *autosync_op = NULL;
 
+    // Scrubbing
+    uint64_t scrub_nearest_ts = 0;
+    int scrub_timer_id = -1;
+    pool_pg_num_t scrub_last_pg = {};
+    osd_op_t *scrub_list_op = NULL;
+    pg_list_result_t scrub_cur_list = {};
+    uint64_t scrub_list_pos = 0;
+
     // Unstable writes
     uint64_t unstable_write_count = 0;
     std::map<osd_object_id_t, uint64_t> unstable_writes;
@@ -221,6 +239,14 @@ class osd_t
     bool continue_recovery();
     pg_osd_set_state_t* change_osd_set(pg_osd_set_state_t *st, pg_t *pg);
 
+    // scrub
+    void scrub_list(pool_pg_num_t pg_id, osd_num_t role_osd, object_id min_oid);
+    int pick_next_scrub(object_id & next_oid);
+    void submit_scrub_op(object_id oid);
+    bool continue_scrub();
+    void plan_scrub(pg_t & pg, bool report_state = true);
+    void schedule_scrub(pg_t & pg);
+
     // op execution
     void exec_op(osd_op_t *cur_op);
     void finish_op(osd_op_t *cur_op, int retval);
@@ -235,13 +261,19 @@ class osd_t
     void autosync();
     bool prepare_primary_rw(osd_op_t *cur_op);
     void continue_primary_read(osd_op_t *cur_op);
+    void continue_primary_scrub(osd_op_t *cur_op);
+    void continue_primary_describe(osd_op_t *cur_op);
     void continue_primary_write(osd_op_t *cur_op);
     void cancel_primary_write(osd_op_t *cur_op);
     void continue_primary_sync(osd_op_t *cur_op);
     void continue_primary_del(osd_op_t *cur_op);
     bool check_write_queue(osd_op_t *cur_op, pg_t & pg);
-    void remove_object_from_state(object_id & oid, pg_osd_set_state_t *object_state, pg_t &pg);
-    void free_object_state(pg_t & pg, pg_osd_set_state_t **object_state);
+    pg_osd_set_state_t* add_object_to_set(pg_t & pg, const object_id oid, const pg_osd_set_t & osd_set,
+        uint64_t old_pg_state, int log_at_level);
+    void remove_object_from_state(object_id & oid, pg_osd_set_state_t **object_state, pg_t &pg, bool report = true);
+    pg_osd_set_state_t *mark_object_corrupted(pg_t & pg, object_id oid, pg_osd_set_state_t *prev_object_state,
+        osd_rmw_stripe_t *stripes, bool ref, bool inconsistent);
+    void deref_object_state(pg_t & pg, pg_osd_set_state_t **object_state, bool deref);
     bool remember_unstable_write(osd_op_t *cur_op, pg_t & pg, pg_osd_set_t & loc_set, int base_state);
     void handle_primary_subop(osd_op_t *subop, osd_op_t *cur_op);
     void handle_primary_bs_subop(osd_op_t *subop);
@@ -256,10 +288,11 @@ class osd_t
     int submit_primary_sync_subops(osd_op_t *cur_op);
     void submit_primary_stab_subops(osd_op_t *cur_op);
 
-    uint64_t* get_object_osd_set(pg_t &pg, object_id &oid, uint64_t *def, pg_osd_set_state_t **object_state);
+    uint64_t* get_object_osd_set(pg_t &pg, object_id &oid, pg_osd_set_state_t **object_state);
 
     void continue_chained_read(osd_op_t *cur_op);
     int submit_chained_read_requests(pg_t & pg, osd_op_t *cur_op);
+    void check_corrupted_chained(pg_t & pg, osd_op_t *cur_op);
     void send_chained_read_results(pg_t & pg, osd_op_t *cur_op);
     std::vector<osd_chain_read_t> collect_chained_read_requests(osd_op_t *cur_op);
     int collect_bitmap_requests(osd_op_t *cur_op, pg_t & pg, std::vector<bitmap_request_t> & bitmap_requests);