Add Hugo-based (https://gohugo.io) documentation

Return results and errors in a variable instead of just printing them,
separate vitastor-cli main() from cli_tool_t, move positional argument
parsing to CLI main from command implementations.

CMakeLists.txt

@@ -2,6 +2,6 @@ cmake_minimum_required(VERSION 2.8)
 
 project(vitastor)
 
-set(VERSION "0.6.10")
+set(VERSION "0.6.17")
 
 add_subdirectory(src)

README-ru.md

@@ -52,10 +52,10 @@ Vitastor на данный момент находится в статусе п
 - Слияние снапшотов (vitastor-cli {snap-rm,flatten,merge})
 - Консольный интерфейс для управления образами (vitastor-cli {ls,create,modify})
 - Плагин для Proxmox
+- Упрощённая NFS-прокси для эмуляции файлового доступа к образам (подходит для VMWare)
 
 ## Планы развития
 
-- Поддержка удаления снапшотов (слияния слоёв)
 - Более корректные скрипты разметки дисков и автоматического запуска OSD
 - Другие инструменты администрирования
 - Плагины для OpenNebula и других облачных систем
@@ -407,6 +407,7 @@ Vitastor с однопоточной NBD прокси на том же стен
 - На хостах мониторов:
   - Пропишите нужные вам значения в файле `/usr/lib/vitastor/mon/make-units.sh`
   - Создайте юниты systemd для etcd и мониторов: `/usr/lib/vitastor/mon/make-units.sh`
+- Запустите etcd и мониторы: `systemctl start etcd vitastor-mon`
 - Пропишите etcd_address и osd_network в `/etc/vitastor/vitastor.conf`. Например:
   ```
   {
@@ -414,7 +415,14 @@ Vitastor с однопоточной NBD прокси на том же стен
     "osd_network": "10.200.1.0/24"
   }
   ```
-- Создайте юниты systemd для OSD: `/usr/lib/vitastor/make-osd.sh /dev/disk/by-partuuid/XXX [/dev/disk/by-partuuid/YYY ...]`
+- Инициализуйте OSD:
+  - SSD: `/usr/lib/vitastor/make-osd.sh /dev/disk/by-partuuid/XXX [/dev/disk/by-partuuid/YYY ...]`
+  - Гибридные, HDD+SSD: `/usr/lib/vitastor/mon/make-osd-hybrid.js /dev/sda /dev/sdb ...` - передайте
+    все ваши SSD и HDD скрипту в командной строке подряд, скрипт автоматически выделит разделы под
+    журналы на SSD и данные на HDD. Скрипт пропускает HDD, на которых уже есть разделы
+    или вообще какие-то данные, поэтому если диски непустые, сначала очистите их с помощью
+    `wipefs -a`. SSD с таблицей разделов не пропускаются, но так как скрипт создаёт новые разделы
+    для журналов, на SSD должно быть доступно свободное нераспределённое место.
 - Вы можете менять параметры OSD в юнитах systemd или в `vitastor.conf`. Смысл некоторых параметров:
   - `disable_data_fsync 1` - отключает fsync, используется с SSD с конденсаторами.
   - `immediate_commit all` - используется с SSD с конденсаторами.
@@ -430,7 +438,6 @@ Vitastor с однопоточной NBD прокси на том же стен
     диски, используемые на одном из тестовых стендов - Intel D3-S4510 - очень сильно не любят такую
     перезапись, и для них была добавлена эта опция. Когда данный режим включён, также нужно поднимать
     значение `journal_sector_buffer_count`, так как иначе Vitastor не хватит буферов для записи в журнал.
-- Запустите все etcd: `systemctl start etcd`
 - Создайте глобальную конфигурацию в etcd: `etcdctl --endpoints=... put /vitastor/config/global '{"immediate_commit":"all"}'`
   (если все ваши диски - серверные с конденсаторами).
 - Создайте пулы: `etcdctl --endpoints=... put /vitastor/config/pools '{"1":{"name":"testpool","scheme":"replicated","pg_size":2,"pg_minsize":1,"pg_count":256,"failure_domain":"host"}}'`.
@@ -523,9 +530,48 @@ 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`.
 
+### NFS
+
+В Vitastor реализована упрощённая NFS 3.0 прокси для эмуляции файлового доступа к образам.
+Это не полноценная файловая система, т.к. метаданные всех файлов (образов) сохраняются
+в etcd и всё время хранятся в оперативной памяти - то есть, положить туда много файлов
+не получится.
+
+Однако в качестве способа доступа к образам виртуальных машин NFS прокси прекрасно подходит
+и позволяет подключить Vitastor, например, к VMWare.
+
+При этом, если вы используете режим immediate_commit=all (для SSD с конденсаторами или HDD
+с отключённым кэшем), то NFS-сервер не имеет состояния и вы можете свободно поднять
+его в нескольких экземплярах и использовать поверх них сетевой балансировщик нагрузки или
+схему с отказоустойчивостью.
+
+Использование vitastor-nfs:
+
+```
+vitastor-nfs [--etcd_address ADDR] [ДРУГИЕ ОПЦИИ]
+
+--subdir <DIR>    экспортировать "поддиректорию" - образы с префиксом имени <DIR>/ (по умолчанию пусто - экспортировать все образы)
+--portmap 0       отключить сервис portmap/rpcbind на порту 111 (по умолчанию включён и требует root привилегий)
+--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)
+--foreground 1    не уходить в фон после запуска
+```
+
+Пример монтирования Vitastor через NFS:
+
+```
+vitastor-nfs --etcd_address 192.168.5.10:2379 --portmap 0 --port 2050 --pool testpool
+```
+
+```
+mount localhost:/ /mnt/ -o port=2050,mountport=2050,nfsvers=3,soft,nolock,tcp
+```
+
 ### Kubernetes
 
-У Vitastor есть CSI-плагин для Kubernetes, поддерживающий RWO-тома.
+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](001-csi-config-map.yaml),

README.md

@@ -46,13 +46,13 @@ breaking changes in the future. However, the following is implemented:
 - Snapshot merge tool (vitastor-cli {snap-rm,flatten,merge})
 - Image management CLI (vitastor-cli {ls,create,modify})
 - Proxmox storage plugin
+- Simplified NFS proxy for file-based image access emulation (suitable for VMWare)
 
 ## Roadmap
 
-- Snapshot deletion (layer merge) support
 - Better OSD creation and auto-start tools
 - Other administrative tools
-- Plugins for OpenNebula, Proxmox and other cloud systems
+- Plugins for OpenNebula and other cloud systems
 - iSCSI proxy
 - Faster failover
 - Scrubbing without checksums (verification of replicas)
@@ -360,6 +360,7 @@ and calculate disk offsets almost by hand. This will be fixed in near future.
 - On the monitor hosts:
   - Edit variables at the top of `/usr/lib/vitastor/mon/make-units.sh` to desired values.
   - Create systemd units for the monitor and etcd: `/usr/lib/vitastor/mon/make-units.sh`
+- Start etcd and monitors: `systemctl start etcd vitastor-mon`
 - Put etcd_address and osd_network into `/etc/vitastor/vitastor.conf`. Example:
   ```
   {
@@ -367,7 +368,13 @@ and calculate disk offsets almost by hand. This will be fixed in near future.
     "osd_network": "10.200.1.0/24"
   }
   ```
-- Create systemd units for your OSDs: `/usr/lib/vitastor/mon/make-osd.sh /dev/disk/by-partuuid/XXX [/dev/disk/by-partuuid/YYY ...]`
+- Initialize OSDs:
+  - Simplest, SSD-only: `/usr/lib/vitastor/mon/make-osd.sh /dev/disk/by-partuuid/XXX [/dev/disk/by-partuuid/YYY ...]`
+  - Hybrid, HDD+SSD: `/usr/lib/vitastor/mon/make-osd-hybrid.js /dev/sda /dev/sdb ...` - pass all your
+    devices (HDD and SSD) to this script - it will partition disks and initialize journals on its own.
+    This script skips HDDs which are already partitioned so if you want to use non-empty disks for
+    Vitastor you should first wipe them with `wipefs -a`. SSDs with GPT partition table are not skipped,
+    but some free unpartitioned space must be available because the script creates new partitions for journals.
 - You can change OSD configuration in units or in `vitastor.conf`. Notable configuration variables:
   - `disable_data_fsync 1` - only safe with server-grade drives with capacitors.
   - `immediate_commit all` - use this if all your drives are server-grade.
@@ -472,9 +479,49 @@ It will output the device name, like /dev/nbd0 which you can then format and mou
 
 Again, you can use `--pool <POOL> --inode <INODE> --size <SIZE>` insteaf of `--image <IMAGE>` if you want.
 
+### NFS
+
+Vitastor has a simplified NFS 3.0 proxy for file-based image access emulation. It's not
+suitable as a full-featured file system, at least because all file/image metadata is stored
+in etcd and kept in memory all the time - thus you can't put a lot of files in it.
+
+However, NFS proxy is totally fine as a method to provide VM image access and allows to
+plug Vitastor into, for example, VMWare. It's important to note that for VMWare it's a much
+better access method than iSCSI, because with iSCSI we'd have to put all VM images into one
+Vitastor image exported as a LUN to VMWare and formatted with VMFS. VMWare doesn't use VMFS
+over NFS.
+
+NFS proxy is stateless if you use immediate_commit=all mode (for SSD with capacitors or
+HDDs with disabled cache), so you can run multiple NFS proxies and use a network load
+balancer or any failover method you want to in that case.
+
+vitastor-nfs usage:
+
+```
+vitastor-nfs [--etcd_address ADDR] [OTHER OPTIONS]
+
+--subdir <DIR>    export images prefixed <DIR>/ (default empty - export all images)
+--portmap 0       do not listen on port 111 (portmap/rpcbind, requires root)
+--bind <IP>       bind service to <IP> address (default 0.0.0.0)
+--nfspath <PATH>  set NFS export path to <PATH> (default is /)
+--port <PORT>     use port <PORT> for NFS services (default is 2049)
+--pool <POOL>     use <POOL> as default pool for new files (images)
+--foreground 1    stay in foreground, do not daemonize
+```
+
+Example start and mount commands:
+
+```
+vitastor-nfs --etcd_address 192.168.5.10:2379 --portmap 0 --port 2050 --pool testpool
+```
+
+```
+mount localhost:/ /mnt/ -o port=2050,mountport=2050,nfsvers=3,soft,nolock,tcp
+```
+
 ### Kubernetes
 
-Vitastor has a CSI plugin for Kubernetes which supports RWO volumes.
+Vitastor has a CSI plugin for Kubernetes which supports RWO (and block RWX) volumes.
 
 To deploy it, take manifests from [csi/deploy/](csi/deploy/) directory, put your
 Vitastor configuration in [csi/deploy/001-csi-config-map.yaml](001-csi-config-map.yaml),

csi/Makefile

@@ -1,4 +1,4 @@
-VERSION ?= v0.6.10
+VERSION ?= v0.6.17
 
 all: build push
 

csi/deploy/004-csi-nodeplugin.yaml

@@ -49,7 +49,7 @@ spec:
             capabilities:
               add: ["SYS_ADMIN"]
             allowPrivilegeEscalation: true
-          image: vitalif/vitastor-csi:v0.6.10
+          image: vitalif/vitastor-csi:v0.6.17
           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:v0.6.10
+          image: vitalif/vitastor-csi:v0.6.17
           args:
             - "--node=$(NODE_ID)"
             - "--endpoint=$(CSI_ENDPOINT)"

csi/deploy/example-pvc-block.yaml

@@ -0,0 +1,13 @@
+---
+apiVersion: v1
+kind: PersistentVolumeClaim
+metadata:
+  name: test-vitastor-pvc-block
+spec:
+  storageClassName: vitastor
+  volumeMode: Block
+  accessModes:
+    - ReadWriteMany
+  resources:
+    requests:
+      storage: 10Gi

csi/deploy/example-test-pod-block.yaml

@@ -0,0 +1,17 @@
+apiVersion: v1
+kind: Pod
+metadata:
+  name: vitastor-test-block-pvc
+  namespace: default
+spec:
+  containers:
+  - name: vitastor-test-block-pvc
+    image: nginx
+    volumeDevices:
+      - name: data
+        devicePath: /dev/xvda
+  volumes:
+  - name: data
+    persistentVolumeClaim:
+      claimName: test-vitastor-pvc-block
+      readOnly: false

csi/deploy/example-test-pod.yaml

@@ -0,0 +1,17 @@
+apiVersion: v1
+kind: Pod
+metadata:
+  name: vitastor-test-nginx
+  namespace: default
+spec:
+  containers:
+   - name: vitastor-test-nginx
+     image: nginx
+     volumeMounts:
+       - mountPath: /usr/share/nginx/html/s3
+         name: data
+  volumes:
+   - name: data
+     persistentVolumeClaim:
+       claimName: test-vitastor-pvc
+       readOnly: false

csi/src/config.go

@@ -5,7 +5,7 @@ package vitastor
 
 const (
     vitastorCSIDriverName    = "csi.vitastor.io"
-    vitastorCSIDriverVersion = "0.6.10"
+    vitastorCSIDriverVersion = "0.6.17"
 )
 
 // Config struct fills the parameters of request or user input

csi/src/nodeserver.go

@@ -67,29 +67,44 @@ func (ns *NodeServer) NodePublishVolume(ctx context.Context, req *csi.NodePublis
     klog.Infof("received node publish volume request %+v", protosanitizer.StripSecrets(req))
 
     targetPath := req.GetTargetPath()
+    isBlock := req.GetVolumeCapability().GetBlock() != nil
 
     // Check that it's not already mounted
-    free, error := mount.IsNotMountPoint(ns.mounter, targetPath)
+    _, error := mount.IsNotMountPoint(ns.mounter, targetPath)
     if (error != nil)
     {
         if (os.IsNotExist(error))
         {
-            error := os.MkdirAll(targetPath, 0777)
+            if (isBlock)
-            if (error != nil)
             {
-                return nil, status.Error(codes.Internal, error.Error())
+                pathFile, err := os.OpenFile(targetPath, os.O_CREATE|os.O_RDWR, 0o600)
+                if (err != nil)
+                {
+                    klog.Errorf("failed to create block device mount target %s with error: %v", targetPath, err)
+                    return nil, status.Error(codes.Internal, err.Error())
+                }
+                err = pathFile.Close()
+                if (err != nil)
+                {
+                    klog.Errorf("failed to close %s with error: %v", targetPath, err)
+                    return nil, status.Error(codes.Internal, err.Error())
+                }
+            }
+            else
+            {
+                err := os.MkdirAll(targetPath, 0777)
+                if (err != nil)
+                {
+                    klog.Errorf("failed to create fs mount target %s with error: %v", targetPath, err)
+                    return nil, status.Error(codes.Internal, err.Error())
+                }
             }
-            free = true
         }
         else
         {
             return nil, status.Error(codes.Internal, error.Error())
         }
     }
-    if (!free)
-    {
-        return &csi.NodePublishVolumeResponse{}, nil
-    }
 
     ctxVars := make(map[string]string)
     err := json.Unmarshal([]byte(req.VolumeId), &ctxVars)
@@ -149,7 +164,6 @@ func (ns *NodeServer) NodePublishVolume(ctx context.Context, req *csi.NodePublis
 
     // Format the device (ext4 or xfs)
     fsType := req.GetVolumeCapability().GetMount().GetFsType()
-    isBlock := req.GetVolumeCapability().GetBlock() != nil
     opt := req.GetVolumeCapability().GetMount().GetMountFlags()
     opt = append(opt, "_netdev")
     if ((req.VolumeCapability.AccessMode.Mode == csi.VolumeCapability_AccessMode_MULTI_NODE_READER_ONLY ||

debian/changelog

@@ -1,4 +1,4 @@
-vitastor (0.6.10-1) unstable; urgency=medium
+vitastor (0.6.17-1) unstable; urgency=medium
 
   * RDMA support
   * Bugfixes

debian/vitastor-client.install

@@ -2,5 +2,6 @@ usr/bin/vita
 usr/bin/vitastor-cli
 usr/bin/vitastor-rm
 usr/bin/vitastor-nbd
+usr/bin/vitastor-nfs
 usr/lib/*/libvitastor*.so*
 mon/make-osd.sh /usr/lib/vitastor

debian/vitastor.Dockerfile

@@ -33,8 +33,8 @@ RUN set -e -x; \
     mkdir -p /root/packages/vitastor-$REL; \
     rm -rf /root/packages/vitastor-$REL/*; \
     cd /root/packages/vitastor-$REL; \
-    cp -r /root/vitastor vitastor-0.6.10; \
+    cp -r /root/vitastor vitastor-0.6.17; \
-    cd vitastor-0.6.10; \
+    cd vitastor-0.6.17; \
     ln -s /root/fio-build/fio-*/ ./fio; \
     FIO=$(head -n1 fio/debian/changelog | perl -pe 's/^.*\((.*?)\).*$/$1/'); \
     ls /usr/include/linux/raw.h || cp ./debian/raw.h /usr/include/linux/raw.h; \
@@ -47,8 +47,8 @@ RUN set -e -x; \
     rm -rf a b; \
     echo "dep:fio=$FIO" > debian/fio_version; \
     cd /root/packages/vitastor-$REL; \
-    tar --sort=name --mtime='2020-01-01' --owner=0 --group=0 --exclude=debian -cJf vitastor_0.6.10.orig.tar.xz vitastor-0.6.10; \
+    tar --sort=name --mtime='2020-01-01' --owner=0 --group=0 --exclude=debian -cJf vitastor_0.6.17.orig.tar.xz vitastor-0.6.17; \
-    cd vitastor-0.6.10; \
+    cd vitastor-0.6.17; \
     V=$(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/gen-docs.js

@@ -0,0 +1,55 @@
+#!/usr/bin/nodejs
+
+const fs = require('fs');
+const yaml = require('yaml');
+
+const L = {
+    en: {},
+    ru: {
+        Type: 'Тип',
+        Default: 'Значение по умолчанию',
+        Minimum: 'Минимальное значение',
+    },
+};
+const types = {
+    en: {
+        string: 'string',
+        bool: 'boolean',
+        int: 'integer',
+        sec: 'seconds',
+        ms: 'milliseconds',
+        us: 'microseconds',
+    },
+    ru: {
+        string: 'строка',
+        bool: 'булево (да/нет)',
+        int: 'целое число',
+        sec: 'секунды',
+        ms: 'миллисекунды',
+        us: 'микросекунды',
+    },
+};
+const params_files = fs.readdirSync(__dirname+'/params')
+    .filter(f => f.substr(-4) == '.yml')
+    .map(f => f.substr(0, f.length-4));
+
+for (const file of params_files)
+{
+    const cfg = yaml.parse(fs.readFileSync(__dirname+'/params/'+file+'.yml', { encoding: 'utf-8' }));
+    for (const lang in types)
+    {
+        let out = '\n\n{{< toc >}}';
+        for (const c of cfg)
+        {
+            out += `\n\n## ${c.name}\n\n`;
+            out += `- ${L[lang]['Type'] || 'Type'}: ${c["type_"+lang] || types[lang][c.type] || c.type}\n`;
+            if (c.default !== undefined)
+                out += `- ${L[lang]['Default'] || 'Default'}: ${c.default}\n`;
+            if (c.min !== undefined)
+                out += `- ${L[lang]['Minimum'] || 'Minimum'}: ${c.min}\n`;
+            out += `\n`+(c["info_"+lang] || c["info"]).replace(/\s+$/, '');
+        }
+        const head = fs.readFileSync(__dirname+'/params/head/'+file+'.'+lang+'.md', { encoding: 'utf-8' });
+        fs.writeFileSync(__dirname+'/hugo/content/config/'+file+'.'+lang+'.md', head.replace(/\s+$/, '')+out+"\n");
+    }
+}

docs/hugo/archetypes/default.md

@@ -0,0 +1,6 @@
+---
+title: "{{ replace .Name "-" " " | title }}"
+date: {{ .Date }}
+draft: true
+---
+

docs/hugo/config.yaml

@@ -0,0 +1,35 @@
+baseURL: http://localhost
+title: Vitastor
+theme: hugo-geekdoc
+#languageCode: en-us
+
+pluralizeListTitles: false
+
+# Geekdoc required configuration
+pygmentsUseClasses: true
+pygmentsCodeFences: true
+disablePathToLower: true
+
+# Required if you want to render robots.txt template
+enableRobotsTXT: true
+
+defaultContentLanguage: en
+languages:
+  en:
+    weight: 1
+    languageName: English
+  ru:
+    weight: 1
+    languageName: Русский
+
+markup:
+  goldmark:
+    renderer:
+      # Needed for mermaid shortcode
+      unsafe: true
+  tableOfContents:
+    startLevel: 1
+    endLevel: 9
+
+taxonomies:
+  tag: tags

docs/hugo/content/_index.md

@@ -0,0 +1,6 @@
+## The Idea
+
+Vitastor is a small, simple and fast clustered block storage (storage for VM drives),
+architecturally similar to Ceph which means strong consistency, primary-replication,
+symmetric clustering and automatic data distribution over any number of drives
+of any size with configurable redundancy (replication or erasure codes/XOR).

docs/hugo/content/config/_index.en.md

@@ -0,0 +1,61 @@
+---
+title: Parameter Reference
+weight: 1
+---
+
+Vitastor configuration consists of:
+- Configuration parameters (key-value), described here
+- [Pool configuration]({{< ref "config/pool" >}})
+- OSD placement tree configuration
+- Inode configuration i.e. image metadata like name, size and parent reference
+
+Configuration parameters can be set in 3 places:
+- Configuration file (`/etc/vitastor/vitastor.conf` or other path)
+- etcd key `/vitastor/config/global`. Most variables can be set there, but etcd
+  connection parameters should obviously be set in the configuration file.
+- Command line of Vitastor components: OSD, 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.
+
+In the future, additional configuration methods may be added:
+- OSD superblock which will, by design, contain parameters related to the disk
+  layout and to one specific OSD.
+- OSD-specific keys in etcd like `/vitastor/config/osd/<number>`.
+
+## Common Parameters
+
+These are the most common parameters which apply to all components of Vitastor.
+
+[See the list]({{< ref "common" >}})
+
+## Cluster-Wide Disk Layout Parameters
+
+These parameters apply to clients and OSDs and can't be changed after OSD
+initialization.
+
+[See the list]({{< ref "layout-cluster" >}})
+
+## OSD Disk Layout Parameters
+
+These parameters apply to OSDs and can't be changed after OSD initialization.
+
+[See the list]({{< ref "layout-osd" >}})
+
+## Network Protocol Parameters
+
+These parameters apply to clients and OSDs and can be changed with a restart.
+
+[See the list]({{< ref "network" >}})
+
+## Runtime OSD Parameters
+
+These parameters apply to OSDs and can be changed with an OSD restart.
+
+[See the list]({{< ref "osd" >}})
+
+## Monitor Parameters
+
+These parameters only apply to Monitors.
+
+[See the list]({{< ref "monitor" >}})

docs/hugo/content/config/_index.ru.md

@@ -0,0 +1,63 @@
+---
+title: Перечень настроек
+weight: 1
+---
+
+Конфигурация Vitastor состоит из:
+- Параметров (ключ-значение), описанных на данной странице
+- Настроек пулов
+- Настроек дерева OSD
+- Настроек инодов, т.е. метаданных образов, таких, как имя, размер и ссылки на
+  родительский образ
+
+Параметры конфигурации могут задаваться в 3 местах:
+- Файле конфигурации (`/etc/vitastor/vitastor.conf` или по другому пути)
+- Ключе в etcd `/vitastor/config/global`. Большая часть параметров может
+  задаваться там, кроме, естественно, самих параметров соединения с etcd,
+  которые должны задаваться в файле конфигурации
+- В командной строке компонентов Vitastor: OSD, монитора, опциях fio и QEMU,
+  настроек OpenStack, Proxmox и т.п. Последние, как правило, не включают полный
+  набор параметров напрямую, но разрешают определить путь к файлу конфигурации
+  и задать любые параметры в нём.
+
+В будущем также могут быть добавлены другие способы конфигурации:
+- Суперблок OSD, в котором будут храниться параметры OSD, связанные с дисковым
+  форматом и с этим конкретным OSD.
+- OSD-специфичные ключи в etcd типа `/vitastor/config/osd/<номер>`.
+
+## Общие параметры
+
+Это наиболее общие параметры, используемые всеми компонентами Vitastor.
+
+[Посмотреть список]({{< ref "common" >}})
+
+## Дисковые параметры уровня кластера
+
+Эти параметры используются клиентами и OSD и не могут быть изменены после
+инициализации OSD.
+
+[Посмотреть список]({{< ref "layout-cluster" >}})
+
+## Дисковые параметры OSD
+
+Эти параметры используются OSD и не могут быть изменены после инициализации OSD.
+
+[Посмотреть список]({{< ref "layout-osd" >}})
+
+## Параметры сетевого протокола
+
+Эти параметры используются клиентами и OSD и могут быть изменены с перезапуском.
+
+[Посмотреть список]({{< ref "network" >}})
+
+## Изменяемые параметры OSD
+
+Эти параметры используются OSD и могут быть изменены с перезапуском.
+
+[Посмотреть список]({{< ref "osd" >}})
+
+## Параметры мониторов
+
+Данные параметры используются только мониторами Vitastor.
+
+[Посмотреть список]({{< ref "monitor" >}})

docs/hugo/content/config/pool.en.md

@@ -0,0 +1,178 @@
+---
+title: Pool configuration
+weight: 100
+---
+
+Pool configuration is set in etcd key `/vitastor/config/pools` in the following
+JSON format:
+
+```
+{
+  "<Numeric ID>": {
+    "name": "<name>",
+    ...other parameters...
+  }
+}
+```
+
+{{< toc >}}
+
+# Parameters
+
+## name
+
+- Type: string
+- Required
+
+Pool name.
+
+## scheme
+
+- Type: string
+- Required
+- One of: "replicated", "xor" or "jerasure"
+
+Redundancy scheme used for data in this pool.
+
+## pg_size
+
+- Type: integer
+- Required
+
+Total number of disks for PGs of this pool - i.e., number of replicas for
+replicated pools and number of data plus parity disks for EC/XOR pools.
+
+## parity_chunks
+
+- Type: integer
+
+Number of parity chunks for EC/XOR pools. For such pools, data will be lost
+if you lose more than parity_chunks disks at once, so this parameter can be
+equally described as FTT (number of failures to tolerate).
+
+Required for EC/XOR pools, ignored for replicated pools.
+
+## pg_minsize
+
+- Type: integer
+- Required
+
+Number of available live disks for PGs of this pool to remain active.
+That is, if it becomes impossible to place PG data on at least (pg_minsize)
+OSDs, PG is deactivated for both read and write. So you know that a fresh
+write always goes to at least (pg_minsize) OSDs (disks).
+
+FIXME: pg_minsize behaviour may be changed in the future to only make PGs
+read-only instead of deactivating them.
+
+## pg_count
+
+- Type: integer
+- Required
+
+Number of PGs for this pool. The value should be big enough for the monitor /
+LP solver to be able to optimize data placement.
+
+"Enough" is usually around 64-128 PGs per OSD, i.e. you set pg_count for pool
+to (total OSD count * 100 / pg_size). You can round it to the closest power of 2,
+because it makes it easier to reduce or increase PG count later by dividing or
+multiplying it by 2.
+
+In Vitastor, PGs are ephemeral, so you can change pool PG count anytime just
+by overwriting pool configuration in etcd. Amount of the data affected by
+rebalance will be smaller if the new PG count is a multiple of the old PG count
+or vice versa.
+
+## failure_domain
+
+- Type: string
+- Default: host
+
+Failure domain specification. Must be "host" or "osd" or refer to one of the
+placement tree levels, defined in [placement_levels]({{< ref "config/monitor#placement_levels" >}}).
+
+Two replicas, or two parts in case of EC/XOR, of the same block of data are
+never put on OSDs in the same failure domain (for example, on the same host).
+So failure domain specifies the unit which failure you are protecting yourself
+from.
+
+## max_osd_combinations
+
+- Type: integer
+- Default: 10000
+
+Vitastor data placement algorithm is based on the LP solver and OSD combinations
+which are fed to it are generated ramdonly. This parameter specifies the maximum
+number of combinations to generate when optimising PG placement.
+
+This parameter usually doesn't require to be changed.
+
+## pg_stripe_size
+
+- Type: integer
+- Default: 0
+
+Specifies the stripe size for this pool according to which images are split into
+different PGs. Stripe size can't be smaller than [block_size]({{< ref "config/layout-cluster#block_size" >}})
+multiplied by (pg_size - parity_chunks) for EC/XOR pools, or 1 for replicated pools,
+and the same value is used by default.
+
+This means first `pg_stripe_size = (block_size * (pg_size-parity_chunks))` bytes
+of an image go to one PG, next `pg_stripe_size` bytes go to another PG and so on.
+
+Usually doesn't require to be changed separately from the block size.
+
+## root_node
+
+- Type: string
+
+Specifies the root node of the OSD tree to restrict this pool OSDs to.
+Referenced root node must exist in /vitastor/config/node_placement.
+
+## osd_tags
+
+- Type: string or array of strings
+
+Specifies OSD tags to restrict this pool to. If multiple tags are specified,
+only OSDs having all of these tags will be used for this pool.
+
+## primary_affinity_tags
+
+- Type: string or array of strings
+
+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.
+
+# Examples
+
+## Replicated pool
+
+```
+{
+  "1": {
+    "name":"testpool",
+    "scheme":"replicated",
+    "pg_size":2,
+    "pg_minsize":1,
+    "pg_count":256,
+    "failure_domain":"host"
+  }
+}
+```
+
+## Erasure-coded pool
+
+```
+{
+  "2": {
+    "name":"ecpool",
+    "scheme":"jerasure",
+    "pg_size":3,
+    "parity_chunks":1,
+    "pg_minsize":2,
+    "pg_count":256,
+    "failure_domain":"host"
+  }
+}
+```

docs/hugo/content/installation/packages.md

@@ -0,0 +1,41 @@
+---
+title: Packages
+weight: 2
+---
+
+## Debian
+
+- Trust Vitastor package signing key:
+  `wget -q -O - https://vitastor.io/debian/pubkey | sudo apt-key add -`
+- Add Vitastor package repository to your /etc/apt/sources.list:
+  - Debian 11 (Bullseye/Sid): `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`
+- Install packages: `apt update; apt install vitastor lp-solve etcd linux-image-amd64 qemu`
+
+## CentOS
+
+- Add Vitastor package repository:
+  - CentOS 7: `yum install https://vitastor.io/rpms/centos/7/vitastor-release-1.0-1.el7.noarch.rpm`
+  - CentOS 8: `dnf install https://vitastor.io/rpms/centos/8/vitastor-release-1.0-1.el8.noarch.rpm`
+- Enable EPEL: `yum/dnf install epel-release`
+- Enable additional CentOS repositories:
+  - CentOS 7: `yum install centos-release-scl`
+  - CentOS 8: `dnf install centos-release-advanced-virtualization`
+- Enable elrepo-kernel:
+  - CentOS 7: `yum install https://www.elrepo.org/elrepo-release-7.el7.elrepo.noarch.rpm`
+  - CentOS 8: `dnf install https://www.elrepo.org/elrepo-release-8.el8.elrepo.noarch.rpm`
+- Install packages: `yum/dnf install vitastor lpsolve etcd kernel-ml qemu-kvm`
+
+## Installation requirements
+
+- Linux kernel 5.4 or newer, for io_uring support. 5.8 or later is highly
+  recommended because io_uring is a relatively new technology and there is
+  at least one bug which reproduces with io_uring and HP SmartArray
+  controllers in 5.4
+- liburing 0.4 or newer
+- lp_solve
+- 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

docs/hugo/content/installation/quickstart.md

@@ -0,0 +1,72 @@
+---
+title: Quick Start
+weight: 1
+---
+
+Prepare:
+
+- Get some SATA or NVMe SSDs with capacitors (server-grade drives). You can use desktop SSDs
+  with lazy fsync, but prepare for inferior single-thread latency. Read more about capacitors
+  [here]({{< ref "config/layout-cluster#immediate_commit" >}}).
+- Get a fast network (at least 10 Gbit/s). Something like Mellanox ConnectX-4 with RoCEv2 is ideal.
+- Disable CPU powersaving: `cpupower idle-set -D 0 && cpupower frequency-set -g performance`.
+- [Install Vitastor packages]({{< ref "installation/packages" >}}).
+
+## Configure monitors
+
+On the monitor hosts:
+- Edit variables at the top of `/usr/lib/vitastor/mon/make-units.sh` to desired values.
+- Create systemd units for the monitor and etcd: `/usr/lib/vitastor/mon/make-units.sh`
+- Start etcd and monitors: `systemctl start etcd vitastor-mon`
+
+## Configure OSDs
+
+- Put etcd_address and osd_network into `/etc/vitastor/vitastor.conf`. Example:
+  ```
+  {
+    "etcd_address": ["10.200.1.10:2379","10.200.1.11:2379","10.200.1.12:2379"],
+    "osd_network": "10.200.1.0/24"
+  }
+  ```
+- Initialize OSDs:
+  - Simplest, SSD-only: `/usr/lib/vitastor/mon/make-osd.sh /dev/disk/by-partuuid/XXX [/dev/disk/by-partuuid/YYY ...]`
+  - Hybrid, HDD+SSD: `/usr/lib/vitastor/mon/make-osd-hybrid.js /dev/sda /dev/sdb ...` &mdash; pass all your
+    devices (HDD and SSD) to this script &mdash; it will partition disks and initialize journals on its own.
+    This script skips HDDs which are already partitioned so if you want to use non-empty disks for
+    Vitastor you should first wipe them with `wipefs -a`. SSDs with GPT partition table are not skipped,
+    but some free unpartitioned space must be available because the script creates new partitions for journals.
+- You can change OSD configuration in units or in `vitastor.conf`.
+  Check [Configuration Reference]({{< ref "config" >}}) for parameter descriptions.
+- `systemctl start vitastor.target` everywhere.
+- If all your drives have capacitors, create global configuration in etcd: \
+  `etcdctl --endpoints=... put /vitastor/config/global '{"immediate_commit":"all"}'`
+
+## Create a pool
+
+Create pool configuration in etcd:
+
+```
+etcdctl --endpoints=... put /vitastor/config/pools '{"1":{"name":"testpool",
+  "scheme":"replicated","pg_size":2,"pg_minsize":1,"pg_count":256,"failure_domain":"host"}}'
+```
+
+For jerasure pools the configuration should look like the following:
+
+```
+etcdctl --endpoints=... put /vitastor/config/pools '{"2":{"name":"ecpool",
+  "scheme":"jerasure","pg_size":4,"parity_chunks":2,"pg_minsize":2,"pg_count":256,"failure_domain":"host"}`
+```
+
+After you do this, one of the monitors will configure PGs and OSDs will start them.
+
+You can check PG states with `etcdctl --endpoints=... get --prefix /vitastor/pg/state`. All PGs should become 'active'.
+
+## Create an image
+
+Use vitastor-cli ([read CLI documentation here]({{< ref "usage/cli" >}})):
+
+```
+vitastor-cli create -s 10G testimg
+```
+
+After that, you can run benchmarks or start QEMU manually with this image.

docs/hugo/content/installation/source.md

@@ -0,0 +1,54 @@
+---
+title: Building from Source
+weight: 3
+---
+
+## Requirements
+
+- gcc and g++ 8 or newer, clang 10 or newer, or other compiler with C++11 plus
+  designated initializers support from C++20
+- CMake
+- liburing, jerasure headers
+
+## Basic instructions
+
+Download source, for example using git: `git clone --recurse-submodules https://yourcmc.ru/git/vitalif/vitastor/`
+
+Get `fio` source and symlink it into `<vitastor>/fio`. If you don't want to build fio engine,
+you can disable it by passing `-DWITH_FIO=no` to cmake.
+
+Build and install Vitastor:
+
+```
+cd vitastor
+mkdir build
+cd build
+cmake .. && make -j8 install
+```
+
+## QEMU Driver
+
+It's recommended to build the QEMU driver (qemu_driver.c) in-tree, as a part of
+QEMU build process. To do that:
+- Install vitastor client library headers (from source or from vitastor-client-dev package)
+- Take a corresponding patch from `patches/qemu-*-vitastor.patch` and apply it to QEMU source
+- Copy `src/qemu_driver.c` to QEMU source directory as `block/block-vitastor.c`
+- Build QEMU as usual
+
+But it is also possible to build it out-of-tree. To do that:
+- Get QEMU source, begin to build it, stop the build and copy headers:
+   - `<qemu>/include` &rarr; `<vitastor>/qemu/include`
+   - Debian:
+      * Use qemu packages from the main repository
+      * `<qemu>/b/qemu/config-host.h` &rarr; `<vitastor>/qemu/b/qemu/config-host.h`
+      * `<qemu>/b/qemu/qapi` &rarr; `<vitastor>/qemu/b/qemu/qapi`
+   - CentOS 8:
+      * Use qemu packages from the Advanced-Virtualization repository. To enable it, run
+        `yum install centos-release-advanced-virtualization.noarch` and then `yum install qemu`
+      * `<qemu>/config-host.h` &rarr; `<vitastor>/qemu/b/qemu/config-host.h`
+      * For QEMU 3.0+: `<qemu>/qapi` &rarr; `<vitastor>/qemu/b/qemu/qapi`
+      * For QEMU 2.0+: `<qemu>/qapi-types.h` &rarr; `<vitastor>/qemu/b/qemu/qapi-types.h`
+   - `config-host.h` and `qapi` are required because they contain generated headers
+- Configure Vitastor with `WITH_QEMU=yes` and, if you're on RHEL, also with `QEMU_PLUGINDIR=qemu-kvm`:
+  `cmake .. -DWITH_QEMU=yes`.
+- After that, Vitastor will build `block-vitastor.so` during its build process.

docs/hugo/content/introduction/_index.md

@@ -0,0 +1,4 @@
+---
+title: Introduction
+weight: -1
+---

docs/hugo/content/introduction/architecture.md

@@ -0,0 +1,73 @@
+---
+title: Architecture
+weight: 3
+---
+
+For people familiar with Ceph, Vitastor is quite similar:
+
+- Vitastor also has Pools, PGs, OSDs, Monitors, Failure Domains, Placement Tree:
+  - OSD (Object Storage Daemon) is a process that stores data and serves read/write requests.
+  - PG (Placement Group) is a container for data that (normally) shares the same replicas.
+  - Pool is a container for data that has the same redundancy scheme and placement rules.
+  - Monitor is a separate daemon that watches cluster state and controls data distribution.
+  - Failure Domain is a group of OSDs that you allow to fail. It's "host" by default.
+  - Placement Tree groups OSDs in a hierarchy to later split them into Failure Domains.
+- Vitastor also distributes every image data across the whole cluster.
+- Vitastor is also transactional (every write to the cluster is atomic).
+- OSDs also have journal and metadata and they can also be put on separate drives.
+- Just like in Ceph, client library attempts to recover from any cluster failure so
+  you can basically reboot the whole cluster and only pause, but not crash, your clients
+  (please report a bug if the client crashes in that case).
+
+However, there are also differences:
+
+- Vitastor's main focus is on SSDs. Hybrid SSD+HDD setups are also possible.
+- Vitastor OSD is (and will always be) single-threaded. If you want to dedicate more than 1 core
+  per drive you should run multiple OSDs each on a different partition of the drive.
+  Vitastor isn't CPU-hungry though (as opposed to Ceph), so 1 core is sufficient in a lot of cases.
+- Metadata and journal are always kept in memory. Metadata size depends linearly on drive capacity
+  and data store block size which is 128 KB by default. With 128 KB blocks metadata should occupy
+  around 512 MB per 1 TB (which is still less than Ceph wants). Journal doesn't have to be big,
+  the example test below was conducted with only 16 MB journal. A big journal is probably even
+  harmful as dirty write metadata also take some memory.
+- 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.
+- The basic layer of Vitastor is block storage with fixed-size blocks, not object storage with
+  rich semantics like in Ceph (RADOS).
+- There's a "lazy fsync" mode which allows to batch writes before flushing them to the disk.
+  This allows to use Vitastor with desktop SSDs, but still lowers performance due to additional
+  network roundtrips, so use server SSDs with capacitor-based power loss protection
+  ("Advanced Power Loss Protection") for best performance.
+- PGs are ephemeral. This means that they aren't stored on data disks and only exist in memory
+  while OSDs are running.
+- Recovery process is per-object (per-block), not per-PG. Also there are no PGLOGs.
+- Monitors don't store data. Cluster configuration and state is stored in etcd in simple human-readable
+  JSON structures. Monitors only watch cluster state and handle data movement.
+  Thus Vitastor's Monitor isn't a critical component of the system and is more similar to Ceph's Manager.
+  Vitastor's Monitor is implemented in node.js.
+- PG distribution isn't based on consistent hashes. All PG mappings are stored in etcd.
+  Rebalancing PGs between OSDs is done by mathematical optimization - data distribution problem
+  is reduced to a linear programming problem and solved by lp_solve. This allows for almost
+  perfect (96-99% uniformity compared to Ceph's 80-90%) data distribution in most cases, ability
+  to map PGs by hand without breaking rebalancing logic, reduced OSD peer-to-peer communication
+  (on average, OSDs have fewer peers) and less data movement. It also probably has a drawback -
+  this method may fail in very large clusters, but up to several hundreds of OSDs it's perfectly fine.
+  It's also easy to add consistent hashes in the future if something proves their necessity.
+- There's no separate CRUSH layer. You select pool redundancy scheme, placement root, failure domain
+  and so on directly in pool configuration.
+- Images are global i.e. you can't create multiple images with the same name in different pools.
+
+## Implementation Principles
+
+- I like architecturally simple solutions. Vitastor is and will always be designed
+  exactly like that.
+- I also like reinventing the wheel to some extent, like writing my own HTTP client
+  for etcd interaction instead of using prebuilt libraries, because in this case
+  I'm confident about what my code does and what it doesn't do.
+- I don't care about C++ "best practices" like RAII or proper inheritance or usage of
+  smart pointers or whatever and I don't intend to change my mind, so if you're here
+  looking for ideal reference C++ code, this probably isn't the right place.
+- I like node.js better than any other dynamically-typed language interpreter
+  because it's faster than any other interpreter in the world, has neutral C-like
+  syntax and built-in event loop. That's why Monitor is implemented in node.js.

docs/hugo/content/introduction/author.md

@@ -0,0 +1,34 @@
+---
+title: Author and License
+weight: 3
+---
+
+Copyright (c) Vitaliy Filippov (vitalif [at] yourcmc.ru), 2019+
+
+Join Vitastor Telegram Chat: https://t.me/vitastor
+
+All server-side code (OSD, Monitor and so on) is licensed under the terms of
+Vitastor Network Public License 1.1 (VNPL 1.1), a copyleft license based on
+GNU GPLv3.0 with the additional "Network Interaction" clause which requires
+opensourcing all programs directly or indirectly interacting with Vitastor
+through a computer network and expressly designed to be used in conjunction
+with it ("Proxy Programs"). Proxy Programs may be made public not only under
+the terms of the same license, but also under the terms of any GPL-Compatible
+Free Software License, as listed by the Free Software Foundation.
+This is a stricter copyleft license than the Affero GPL.
+
+Please note that VNPL doesn't require you to open the code of proprietary
+software running inside a VM if it's not specially designed to be used with
+Vitastor.
+
+Basically, you can't use the software in a proprietary environment to provide
+its functionality to users without opensourcing all intermediary components
+standing between the user and Vitastor or purchasing a commercial license
+from the author 😀.
+
+Client libraries (cluster_client and so on) are dual-licensed under the same
+VNPL 1.1 and also GNU GPL 2.0 or later to allow for compatibility with GPLed
+software like QEMU and fio.
+
+You can find the full text of VNPL-1.1 in the file [VNPL-1.1.txt](VNPL-1.1.txt).
+GPL 2.0 is also included in this repository as [GPL-2.0.txt](GPL-2.0.txt).

docs/hugo/content/introduction/features.md

@@ -0,0 +1,60 @@
+---
+title: Features
+weight: 1
+---
+
+Vitastor is currently a pre-release and it still misses some important features.
+However, the following is implemented:
+
+- Basic part: highly-available block storage with symmetric clustering and no SPOF
+- Performance ;-D
+- Multiple redundancy schemes: Replication, XOR n+1, Reed-Solomon erasure codes
+  based on jerasure library with any number of data and parity drives in a group
+- Configuration via simple JSON data structures in etcd (parameters, pools and images)
+- Automatic data distribution over OSDs, with support for:
+  - Mathematical optimization for better uniformity and less data movement
+  - Multiple pools
+  - Placement tree, OSD selection by tags (device classes) and placement root
+  - Configurable failure domains
+- Recovery of degraded blocks
+- Rebalancing (data movement between OSDs)
+- Lazy fsync support
+- Per-OSD and per-image I/O and space usage statistics in etcd
+- Snapshots and copy-on-write image clones
+- Write throttling to smooth random write workloads in SSD+HDD configurations
+- RDMA/RoCEv2 support via libibverbs
+
+CLI (vitastor-cli):
+- Pool listing and space stats (df)
+- Image listing, space and I/O stats (ls)
+- Image and snapshot creation (create, modify)
+- Image removal and snapshot merge (rm, flatten, merge, rm-data)
+
+Plugins and packaging:
+- Debian and CentOS packages
+- Generic user-space client library
+- Native QEMU driver
+- Loadable fio engine for benchmarks
+- NBD proxy for kernel mounts
+- CSI plugin for Kubernetes
+- OpenStack support: Cinder driver, Nova and libvirt patches
+- Proxmox storage plugin and packages
+
+## Roadmap
+
+The following features are planned for the future:
+
+- Better OSD creation and auto-start tools
+- Other administrative tools
+- Web GUI
+- OpenNebula plugin
+- iSCSI proxy
+- Simplified NFS proxy
+- Multi-threaded client
+- Faster failover
+- Scrubbing without checksums (verification of replicas)
+- Checksums
+- Tiered storage (SSD caching)
+- NVDIMM support
+- Compression (possibly)
+- Read caching using system page cache (possibly)

docs/hugo/content/performance/comparison1.md

@@ -0,0 +1,93 @@
+---
+title: Example Comparison with Ceph
+weight: 4
+---
+
+Hardware configuration: 4 nodes, each with:
+- 6x SATA SSD Intel D3-S4510 3.84 TB
+- 2x Xeon Gold 6242 (16 cores @ 2.8 GHz)
+- 384 GB RAM
+- 1x 25 GbE network interface (Mellanox ConnectX-4 LX), connected to a Juniper QFX5200 switch
+
+CPU powersaving was disabled. Both Vitastor and Ceph were configured with 2 OSDs per 1 SSD.
+
+All of the results below apply to 4 KB blocks and random access (unless indicated otherwise).
+
+T8Q64 tests were conducted over 8 400GB RBD images from all hosts (every host was running 2 instances of fio).
+This is because Ceph has performance penalties related to running multiple clients over a single RBD image.
+
+cephx_sign_messages was set to false during tests, RocksDB and Bluestore settings were left at defaults.
+
+T8Q64 read test was conducted over 1 larger inode (3.2T) from all hosts (every host was running 2 instances of fio).
+Vitastor has no performance penalties related to running multiple clients over a single inode.
+If conducted from one node with all primary OSDs moved to other nodes the result was slightly lower (689000 iops),
+this is because all operations resulted in network roundtrips between the client and the primary OSD.
+When fio was colocated with OSDs (like in Ceph benchmarks above), 1/4 of the read workload actually
+used the loopback network.
+
+Vitastor was configured with: `--disable_data_fsync true --immediate_commit all --flusher_count 8
+  --disk_alignment 4096 --journal_block_size 4096 --meta_block_size 4096
+  --journal_no_same_sector_overwrites true --journal_sector_buffer_count 1024
+  --journal_size 16777216`.
+
+## Raw drive performance
+
+- T1Q1 write ~27000 iops (~0.037ms latency)
+- T1Q1 read ~9800 iops (~0.101ms latency)
+- T1Q32 write ~60000 iops
+- T1Q32 read ~81700 iops
+
+## 2 replicas
+
+### Ceph 15.2.4 (Bluestore)
+
+- T1Q1 write ~1000 iops (~1ms latency)
+- T1Q1 read ~1750 iops (~0.57ms latency)
+- T8Q64 write ~100000 iops, total CPU usage by OSDs about 40 virtual cores on each node
+- T8Q64 read ~480000 iops, total CPU usage by OSDs about 40 virtual cores on each node
+
+In fact, not that bad for Ceph. These servers are an example of well-balanced Ceph nodes.
+However, CPU usage and I/O latency were through the roof, as usual.
+
+### Vitastor 0.4.0 (native)
+
+- T1Q1 write: 7087 iops (0.14ms latency)
+- T1Q1 read: 6838 iops (0.145ms latency)
+- T2Q64 write: 162000 iops, total CPU usage by OSDs about 3 virtual cores on each node
+- T8Q64 read: 895000 iops, total CPU usage by OSDs about 4 virtual cores on each node
+- Linear write (4M T1Q32): 2800 MB/s
+- Linear read (4M T1Q32): 1500 MB/s
+
+### Vitastor 0.4.0 (NBD)
+
+NBD is currently required to mount Vitastor via kernel, but it imposes additional overhead
+due to additional copying between the kernel and userspace. This mostly hurts linear
+bandwidth, not iops.
+
+Vitastor with single-threaded NBD on the same hardware:
+- T1Q1 write: 6000 iops (0.166ms latency)
+- T1Q1 read: 5518 iops (0.18ms latency)
+- T1Q128 write: 94400 iops
+- T1Q128 read: 103000 iops
+- Linear write (4M T1Q128): 1266 MB/s (compared to 2800 MB/s via fio)
+- Linear read (4M T1Q128): 975 MB/s (compared to 1500 MB/s via fio)
+
+## EC/XOR 2+1
+
+### Ceph 15.2.4
+
+- T1Q1 write: 730 iops (~1.37ms latency)
+- T1Q1 read: 1500 iops with cold cache (~0.66ms latency), 2300 iops after 2 minute metadata cache warmup (~0.435ms latency)
+- T4Q128 write (4 RBD images): 45300 iops, total CPU usage by OSDs about 30 virtual cores on each node
+- T8Q64 read (4 RBD images): 278600 iops, total CPU usage by OSDs about 40 virtual cores on each node
+- Linear write (4M T1Q32): 1950 MB/s before preallocation, 2500 MB/s after preallocation
+- Linear read (4M T1Q32): 2400 MB/s
+
+### Vitastor 0.4.0
+
+- T1Q1 write: 2808 iops (~0.355ms latency)
+- T1Q1 read: 6190 iops (~0.16ms latency)
+- T2Q64 write: 85500 iops, total CPU usage by OSDs about 3.4 virtual cores on each node
+- T8Q64 read: 812000 iops, total CPU usage by OSDs about 4.7 virtual cores on each node
+- Linear write (4M T1Q32): 3200 MB/s
+- Linear read (4M T1Q32): 1800 MB/s

docs/hugo/content/performance/theoretical.md

@@ -0,0 +1,46 @@
+---
+title: Vitastor's Theoretical Maximum Performance
+weight: 3
+---
+
+Replicated setups:
+- Single-threaded (T1Q1) read latency: 1 network roundtrip + 1 disk read.
+- Single-threaded write+fsync latency:
+  - With immediate commit: 2 network roundtrips + 1 disk write.
+  - With lazy commit: 4 network roundtrips + 1 disk write + 1 disk flush.
+- Saturated parallel read iops: min(network bandwidth, sum(disk read iops)).
+- Saturated parallel write iops: min(network bandwidth, sum(disk write iops / number of replicas / write amplification)).
+
+EC/XOR setups:
+- Single-threaded (T1Q1) read latency: 1.5 network roundtrips + 1 disk read.
+- Single-threaded write+fsync latency:
+  - With immediate commit: 3.5 network roundtrips + 1 disk read + 2 disk writes.
+  - With lazy commit: 5.5 network roundtrips + 1 disk read + 2 disk writes + 2 disk fsyncs.
+  - 0.5 in actually (k-1)/k which means that an additional roundtrip doesn't happen when
+    the read sub-operation can be served locally.
+- Saturated parallel read iops: min(network bandwidth, sum(disk read iops)).
+- Saturated parallel write iops: min(network bandwidth, sum(disk write iops * number of data drives / (number of data + parity drives) / write amplification)).
+  In fact, you should put disk write iops under the condition of ~10% reads / ~90% writes in this formula.
+
+Write amplification for 4 KB blocks is usually 3-5 in Vitastor:
+1. Journal block write
+2. Journal data write
+3. Metadata block write
+4. Another journal block write for EC/XOR setups
+5. Data block write
+
+If you manage to get an SSD which handles 512 byte blocks well (Optane?) you may
+lower 1, 3 and 4 to 512 bytes (1/8 of data size) and get WA as low as 2.375.
+
+Lazy fsync also reduces WA for parallel workloads because journal blocks are only
+written when they fill up or fsync is requested.
+
+## In Practice
+
+In practice, using tests from [Understanding Performance]({{< ref "performance/understanding" >}})
+and good server-grade SSD/NVMe drives, you should head for:
+- At least 5000 T1Q1 replicated read and write iops (maximum 0.2ms latency)
+- At least ~80k parallel read iops or ~30k write iops per 1 core (1 OSD)
+- Disk-speed or wire-speed linear reads and writes, whichever is the bottleneck in your case
+
+If your results are lower, that may mean you have bad drives, bad network or some kind of misconfiguration.

docs/hugo/content/performance/tuning.md

@@ -0,0 +1,6 @@
+---
+title: Tuning
+weight: 2
+---
+
+- Disable CPU powersaving

docs/hugo/content/performance/understanding.md

@@ -0,0 +1,52 @@
+---
+title: Understanding Storage Performance
+weight: 1
+---
+
+The most important thing for fast storage is latency, not parallel iops.
+
+The best possible latency is achieved with one thread and queue depth of 1 which basically means
+"client load as low as possible". In this case IOPS = 1/latency, and this number doesn't
+scale with number of servers, drives, server processes or threads and so on.
+Single-threaded IOPS and latency numbers only depend on *how fast a single daemon is*.
+
+Why is it important? It's important because some of the applications *can't* use
+queue depth greater than 1 because their task isn't parallelizable. A notable example
+is any ACID DBMS because all of them write their WALs sequentially with fsync()s.
+
+fsync, by the way, is another important thing often missing in benchmarks. The point is
+that drives have cache buffers and don't guarantee that your data is actually persisted
+until you call fsync() which is translated to a FLUSH CACHE command by the OS.
+
+Desktop SSDs are very fast without fsync - NVMes, for example, can process ~80000 write
+operations per second with queue depth of 1 without fsync - but they're really slow with
+fsync because they have to actually write data to flash chips when you call fsync. Typical
+number is around 1000-2000 iops with fsync.
+
+Server SSDs often have supercapacitors that act as a built-in UPS and allow the drive
+to flush its DRAM cache to the persistent flash storage when a power loss occurs.
+This makes them perform equally well with and without fsync. This feature is called
+"Advanced Power Loss Protection" by Intel; other vendors either call it similarly
+or directly as "Full Capacitor-Based Power Loss Protection".
+
+All software-defined storages that I currently know are slow in terms of latency.
+Notable examples are Ceph and internal SDSes used by cloud providers like Amazon, Google,
+Yandex and so on. They're all slow and can only reach ~0.3ms read and ~0.6ms 4 KB write latency
+with best-in-slot hardware.
+
+And that's in the SSD era when you can buy an SSD that has ~0.04ms latency for 100 $.
+
+I use the following 6 commands with small variations to benchmark any storage:
+
+- Linear write:
+  `fio -ioengine=libaio -direct=1 -invalidate=1 -name=test -bs=4M -iodepth=32 -rw=write -runtime=60 -filename=/dev/sdX`
+- Linear read:
+  `fio -ioengine=libaio -direct=1 -invalidate=1 -name=test -bs=4M -iodepth=32 -rw=read -runtime=60 -filename=/dev/sdX`
+- Random write latency (T1Q1, this hurts storages the most):
+  `fio -ioengine=libaio -direct=1 -invalidate=1 -name=test -bs=4k -iodepth=1 -fsync=1 -rw=randwrite -runtime=60 -filename=/dev/sdX`
+- Random read latency (T1Q1):
+  `fio -ioengine=libaio -direct=1 -invalidate=1 -name=test -bs=4k -iodepth=1 -rw=randread -runtime=60 -filename=/dev/sdX`
+- Parallel write iops (use numjobs if a single CPU core is insufficient to saturate the load):
+  `fio -ioengine=libaio -direct=1 -invalidate=1 -name=test -bs=4k -iodepth=128 [-numjobs=4 -group_reporting] -rw=randwrite -runtime=60 -filename=/dev/sdX`
+- Parallel read iops (use numjobs if a single CPU core is insufficient to saturate the load):
+  `fio -ioengine=libaio -direct=1 -invalidate=1 -name=test -bs=4k -iodepth=128 [-numjobs=4 -group_reporting] -rw=randread -runtime=60 -filename=/dev/sdX`

docs/hugo/content/usage/cli.md

@@ -0,0 +1,183 @@
+---
+title: Vitastor CLI
+weight: 1
+---
+
+vitastor-cli is a command-line tool for administrative tasks like image management.
+
+It supports the following commands:
+
+{{< toc >}}
+
+Global options:
+
+```
+--etcd_address ADDR  Etcd connection address
+--iodepth N          Send N operations in parallel to each OSD when possible (default 32)
+--parallel_osds M    Work with M osds in parallel when possible (default 4)
+--progress 1|0       Report progress (default 1)
+--cas 1|0            Use online CAS writes when possible (default auto)
+--no-color           Disable colored output
+--json               JSON output
+```
+
+## status
+
+`vitastor-cli status`
+
+Show cluster status.
+
+Example output:
+
+```
+  cluster:
+    etcd: 1 / 1 up, 1.8 M database size
+    mon:  1 up, master stump
+    osd:  8 / 12 up
+
+  data:
+    raw:   498.5 G used, 301.2 G / 799.7 G available, 399.8 G down
+    state: 156.6 G clean, 97.6 G misplaced
+    pools: 2 / 3 active
+    pgs:   30 active
+           34 active+has_misplaced
+           32 offline
+
+  io:
+    client:    0 B/s rd, 0 op/s rd, 0 B/s wr, 0 op/s wr
+    rebalance: 989.8 M/s, 7.9 K op/s
+```
+
+## df
+
+`vitastor-cli df`
+
+Show pool space statistics.
+
+Example output:
+
+```
+NAME      SCHEME  PGS  TOTAL    USED    AVAILABLE  USED%   EFFICIENCY
+testpool  2/1     32   100 G    34.2 G  60.7 G     39.23%  100%
+size1     1/1     32   199.9 G  10 G    121.5 G    39.23%  100%
+kaveri    2/1     32   0 B      10 G    0 B        100%    0%
+```
+
+In the example above, "kaveri" pool has "zero" efficiency because all its OSD are down.
+
+## ls
+
+`vitastor-cli ls [-l] [-p POOL] [--sort FIELD] [-r] [-n N] [<glob> ...]`
+
+List images (only matching `<glob>` pattern(s) if passed).
+
+Options:
+
+```
+-p|--pool POOL  Filter images by pool ID or name
+-l|--long       Also report allocated size and I/O statistics
+--del           Also include delete operation statistics
+--sort FIELD    Sort by specified field (name, size, used_size, <read|write|delete>_<iops|bps|lat|queue>)
+-r|--reverse    Sort in descending order
+-n|--count N    Only list first N items
+```
+
+Example output:
+
+```
+NAME                 POOL      SIZE  USED    READ   IOPS  QUEUE  LAT   WRITE  IOPS  QUEUE  LAT   FLAGS  PARENT
+debian9              testpool  20 G  12.3 G  0 B/s  0     0      0 us  0 B/s  0     0      0 us     RO
+pve/vm-100-disk-0    testpool  20 G  0 B     0 B/s  0     0      0 us  0 B/s  0     0      0 us      -  debian9
+pve/base-101-disk-0  testpool  20 G  0 B     0 B/s  0     0      0 us  0 B/s  0     0      0 us     RO  debian9
+pve/vm-102-disk-0    testpool  32 G  36.4 M  0 B/s  0     0      0 us  0 B/s  0     0      0 us      -  pve/base-101-disk-0
+debian9-test         testpool  20 G  36.6 M  0 B/s  0     0      0 us  0 B/s  0     0      0 us      -  debian9
+bench                testpool  10 G  10 G    0 B/s  0     0      0 us  0 B/s  0     0      0 us      -
+bench-kaveri         kaveri    10 G  10 G    0 B/s  0     0      0 us  0 B/s  0     0      0 us      -
+```
+
+## create
+
+`vitastor-cli create -s|--size <size> [-p|--pool <id|name>] [--parent <parent_name>[@<snapshot>]] <name>`
+
+Create an image. You may use K/M/G/T suffixes for `<size>`. If `--parent` is specified,
+a copy-on-write image clone is created. Parent must be a snapshot (readonly image).
+Pool must be specified if there is more than one pool.
+
+```
+vitastor-cli create --snapshot <snapshot> [-p|--pool <id|name>] <image>
+vitastor-cli snap-create [-p|--pool <id|name>] <image>@<snapshot>
+```
+
+Create a snapshot of image `<name>` (either form can be used). May be used live if only a single writer is active.
+
+## modify
+
+`vitastor-cli modify <name> [--rename <new-name>] [--resize <size>] [--readonly | --readwrite] [-f|--force]`
+
+Rename, resize image or change its readonly status. Images with children can't be made read-write.
+If the new size is smaller than the old size, extra data will be purged.
+You should resize file system in the image, if present, before shrinking it.
+
+```
+-f|--force  Proceed with shrinking or setting readwrite flag even if the image has children.
+```
+
+## rm
+
+`vitastor-cli rm <from> [<to>] [--writers-stopped]`
+
+Remove `<from>` or all layers between `<from>` and `<to>` (`<to>` must be a child of `<from>`),
+rebasing all their children accordingly. --writers-stopped allows merging to be a bit
+more effective in case of a single 'slim' read-write child and 'fat' removed parent:
+the child is merged into parent and parent is renamed to child in that case.
+In other cases parent layers are always merged into children.
+
+## flatten
+
+`vitastor-cli flatten <layer>`
+
+Flatten a layer, i.e. merge data and detach it from parents.
+
+## rm-data
+
+`vitastor-cli rm-data --pool <pool> --inode <inode> [--wait-list] [--min-offset <offset>]`
+
+Remove inode data without changing metadata.
+
+```
+--wait-list   Retrieve full objects listings before starting to remove objects.
+              Requires more memory, but allows to show correct removal progress.
+--min-offset  Purge only data starting with specified offset.
+```
+
+## merge-data
+
+`vitastor-cli merge-data <from> <to> [--target <target>]`
+
+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>`.
+
+## alloc-osd
+
+`vitastor-cli alloc-osd`
+
+Allocate a new OSD number and reserve it by creating empty `/osd/stats/<n>` key.
+
+## simple-offsets
+
+`vitastor-cli simple-offsets <device>`
+
+Calculate offsets for simple&stupid (no superblock) OSD deployment.
+
+Options:
+
+```
+--object_size 128k       Set blockstore block size
+--bitmap_granularity 4k  Set bitmap granularity
+--journal_size 16M       Set journal size
+--device_block_size 4k   Set device block size
+--journal_offset 0       Set journal offset
+--device_size 0          Set device size
+--format text            Result format: json, options, env, or text
+```

docs/hugo/content/usage/nbd.md

@@ -0,0 +1,20 @@
+---
+title: NBD
+weight: 6
+---
+
+To create a local block device for a Vitastor image, use NBD. For example:
+
+```
+vitastor-nbd map --etcd_address 10.115.0.10:2379/v3 --image testimg
+```
+
+It will output the device name, like /dev/nbd0 which you can then format and mount as a normal block device.
+
+You can also use `--pool <POOL> --inode <INODE> --size <SIZE>` instead of `--image <IMAGE>` if you want.
+
+To unmap the device run:
+
+```
+vitastor-nbd unmap /dev/nbd0
+```

docs/hugo/content/usage/qemu.md

@@ -0,0 +1,39 @@
+---
+title: QEMU and qemu-img
+weight: 2
+---
+
+You need patched QEMU version to use Vitastor driver.
+
+To start a VM using plain QEMU command-line with Vitastor disk, use the following commands:
+
+Old syntax (-drive):
+
+```
+qemu-system-x86_64 -enable-kvm -m 1024 \
+    -drive 'file=vitastor:etcd_host=192.168.7.2\:2379/v3:image=debian9',
+        format=raw,if=none,id=drive-virtio-disk0,cache=none \
+    -device 'virtio-blk-pci,scsi=off,bus=pci.0,addr=0x5,drive=drive-virtio-disk0,
+        id=virtio-disk0,bootindex=1,write-cache=off' \
+    -vnc 0.0.0.0:0
+```
+
+New syntax (-blockdev):
+
+```
+qemu-system-x86_64 -enable-kvm -m 1024 \
+    -blockdev '{"node-name":"drive-virtio-disk0","driver":"vitastor","image":"debian9",
+        "cache":{"direct":true,"no-flush":false},"auto-read-only":true,"discard":"unmap"}' \
+    -device 'virtio-blk-pci,scsi=off,bus=pci.0,addr=0x5,drive=drive-virtio-disk0,
+        id=virtio-disk0,bootindex=1,write-cache=off' \
+    -vnc 0.0.0.0:0
+```
+
+For qemu-img, you should use `vitastor:etcd_host=<HOST>:image=<IMAGE>` as filename. For example:
+
+```
+qemu-img convert -f qcow2 debian10.qcow2 -p -O raw 'vitastor:etcd_host=192.168.7.2\:2379/v3:image=debian10'
+```
+
+You can also specify `:pool=<POOL>:inode=<INODE>:size=<SIZE>` instead of `:image=<IMAGE>`
+if you don't want to use inode metadata.

docs/hugo/i18n/ru.yaml

@@ -0,0 +1,37 @@
+---
+nav_navigation: Навигация
+nav_tags: Теги
+nav_more: Подробнее
+nav_top: К началу
+
+form_placeholder_search: Поиск
+
+error_page_title: Открыта несуществующая страница
+error_message_title: Потерялись?
+error_message_code: Ошибка 404
+error_message_text: >
+  Похоже, страница, которую вы открыли, не существует. Попробуйте найти
+  нужную информацию с <a class="gdoc-error__link" href="{{ . }}">главной страницы</a>.
+
+button_toggle_dark: Переключить тёмный/светлый/авто режим
+button_nav_open: Показать навигацию
+button_nav_close: Скрыть навигацию
+button_menu_open: Открыть меню
+button_menu_close: Закрыть меню
+button_homepage: На главную
+
+title_anchor_prefix: "Ссылка на:"
+
+posts_read_more: Читать подробнее
+posts_read_time:
+  one: "Одна минута на чтение"
+  other: "{{ . }} минут(ы) на чтение"
+posts_update_prefix: Обновлено
+
+footer_build_with: >
+  Сделано на <a href="https://gohugo.io/" class="gdoc-footer__link">Hugo</a> с
+  <svg class="icon gdoc_heart"><use xlink:href="#gdoc_heart"></use></svg>
+footer_legal_notice: Правовая информация
+footer_privacy_policy: Приватность
+
+language_switch_no_tranlation_prefix: "Страница не переведена:"

docs/hugo/layouts/partials/site-footer.html

@@ -0,0 +1,34 @@
+<footer class="gdoc-footer">
+    <div class="container flex">
+        <div class="flex flex-wrap" style="flex: 1">
+            <span class="gdoc-footer__item gdoc-footer__item--row">
+                &copy; Vitaliy Filippov, 2021+
+            </span>
+        </div>
+        <div class="flex flex-wrap">
+            {{ with .Site.Params.GeekdocLegalNotice }}
+            <span class="gdoc-footer__item gdoc-footer__item--row">
+                <a href="{{ . | relURL }}" class="gdoc-footer__link">{{ i18n "footer_legal_notice" }}</a>
+            </span>
+            {{ end }}
+            {{ with .Site.Params.GeekdocPrivacyPolicy }}
+            <span class="gdoc-footer__item gdoc-footer__item--row">
+                <a href="{{ . | relURL }}" class="gdoc-footer__link">{{ i18n "footer_privacy_policy" }}</a>
+            </span>
+            {{ end }}
+        </div>
+        {{ if (default true .Site.Params.GeekdocBackToTop) }}
+        <div class="flex flex-25 justify-end">
+            <span class="gdoc-footer__item gdoc-footer__item--row" style="margin-right: 50px">
+                {{ i18n "footer_build_with" | safeHTML }}
+            </span>
+            <span class="gdoc-footer__item">
+                <a class="gdoc-footer__link fake-link" href="#" aria-label="{{ i18n "nav_top" }}">
+                    <svg class="icon gdoc_keyboard_arrow_up"><use xlink:href="#gdoc_keyboard_arrow_up"></use></svg>
+                    <span class="hidden-mobile">{{ i18n "nav_top" }}</span>
+                </a>
+            </span>
+        </div>
+        {{ end }}
+    </div>
+</footer>

docs/hugo/static/brand.svg

@@ -0,0 +1,215 @@
+<?xml version="1.0" encoding="UTF-8" standalone="no"?>
+<svg
+   xmlns:osb="http://www.openswatchbook.org/uri/2009/osb"
+   xmlns:dc="http://purl.org/dc/elements/1.1/"
+   xmlns:cc="http://creativecommons.org/ns#"
+   xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
+   xmlns:svg="http://www.w3.org/2000/svg"
+   xmlns="http://www.w3.org/2000/svg"
+   xmlns:xlink="http://www.w3.org/1999/xlink"
+   xmlns:sodipodi="http://sodipodi.sourceforge.net/DTD/sodipodi-0.dtd"
+   xmlns:inkscape="http://www.inkscape.org/namespaces/inkscape"
+   sodipodi:docname="logo_only2.svg"
+   inkscape:version="1.0.2 (e86c870879, 2021-01-15)"
+   id="svg1340"
+   version="1.1"
+   viewBox="0 0 100 86.80192"
+   height="86.801918mm"
+   width="100mm"
+   inkscape:export-filename="/var/home/vitali/SVN/vitastor/presentation/logos/logo_only.png"
+   inkscape:export-xdpi="92.889999"
+   inkscape:export-ydpi="92.889999">
+  <defs
+     id="defs1334">
+    <linearGradient
+       osb:paint="gradient"
+       id="linearGradient866">
+      <stop
+         id="stop862"
+         offset="0"
+         style="stop-color:#c0c0c0;stop-opacity:1" />
+      <stop
+         id="stop864"
+         offset="1"
+         style="stop-color:#000000;stop-opacity:0" />
+    </linearGradient>
+    <linearGradient
+       id="linearGradient846"
+       osb:paint="gradient">
+      <stop
+         style="stop-color:#ffd42a;stop-opacity:1"
+         offset="0"
+         id="stop842" />
+      <stop
+         style="stop-color:#ffa200;stop-opacity:1"
+         offset="1"
+         id="stop844" />
+    </linearGradient>
+    <radialGradient
+       r="50"
+       fy="159.11139"
+       fx="202.36813"
+       cy="159.11139"
+       cx="202.36813"
+       gradientTransform="matrix(1.2462942,-1.2279529,0.77712408,0.78873143,-190.96813,230.1331)"
+       gradientUnits="userSpaceOnUse"
+       id="radialGradient1530"
+       xlink:href="#linearGradient1352"
+       inkscape:collect="always" />
+    <linearGradient
+       inkscape:collect="always"
+       id="linearGradient1352">
+      <stop
+         style="stop-color:#00c9e6;stop-opacity:1"
+         offset="0"
+         id="stop1348" />
+      <stop
+         style="stop-color:#5240d3;stop-opacity:1"
+         offset="1"
+         id="stop1350" />
+    </linearGradient>
+    <linearGradient
+       y2="62.555599"
+       x2="51.484566"
+       y1="62.555599"
+       x1="38.105473"
+       gradientTransform="rotate(-16.930773,271.11609,-412.42594)"
+       gradientUnits="userSpaceOnUse"
+       id="linearGradient1508"
+       xlink:href="#linearGradient1323"
+       inkscape:collect="always" />
+    <linearGradient
+       inkscape:collect="always"
+       id="linearGradient1323">
+      <stop
+         style="stop-color:#000000;stop-opacity:0.47178105"
+         offset="0"
+         id="stop1319" />
+      <stop
+         style="stop-color:#eeaaff;stop-opacity:0;"
+         offset="1"
+         id="stop1321" />
+    </linearGradient>
+    <radialGradient
+       r="21.541935"
+       fy="24.614815"
+       fx="45.312912"
+       cy="24.614815"
+       cx="45.312912"
+       gradientTransform="matrix(1.0933447,0.13113705,-0.12664108,1.0558599,-1.082187,93.974708)"
+       gradientUnits="userSpaceOnUse"
+       id="radialGradient1504"
+       xlink:href="#linearGradient846"
+       inkscape:collect="always" />
+    <filter
+       style="color-interpolation-filters:sRGB"
+       inkscape:label="Drop Shadow"
+       id="filter1497"
+       width="2"
+       height="2"
+       x="-0.5"
+       y="-0.5">
+      <feFlood
+         flood-opacity="0.498039"
+         flood-color="rgb(0,0,0)"
+         result="flood"
+         id="feFlood1487" />
+      <feComposite
+         in="flood"
+         in2="SourceGraphic"
+         operator="in"
+         result="composite1"
+         id="feComposite1489" />
+      <feGaussianBlur
+         in="composite1"
+         stdDeviation="6"
+         result="blur"
+         id="feGaussianBlur1491" />
+      <feOffset
+         dx="0"
+         dy="6"
+         result="offset"
+         id="feOffset1493" />
+      <feComposite
+         in="offset"
+         in2="offset"
+         operator="atop"
+         result="composite2"
+         id="feComposite1495" />
+    </filter>
+    <radialGradient
+       r="21.541935"
+       fy="24.614815"
+       fx="45.312912"
+       cy="24.614815"
+       cx="45.312912"
+       gradientTransform="matrix(1.0933447,0.13113705,-0.12664108,1.0558599,-1.082187,93.974708)"
+       gradientUnits="userSpaceOnUse"
+       id="radialGradient1506"
+       xlink:href="#linearGradient846"
+       inkscape:collect="always" />
+  </defs>
+  <sodipodi:namedview
+     inkscape:window-maximized="1"
+     inkscape:window-y="0"
+     inkscape:window-x="0"
+     inkscape:window-height="992"
+     inkscape:window-width="1920"
+     fit-margin-bottom="0"
+     fit-margin-right="0"
+     fit-margin-left="0"
+     fit-margin-top="-30"
+     showgrid="false"
+     inkscape:document-rotation="0"
+     inkscape:current-layer="layer1"
+     inkscape:document-units="mm"
+     inkscape:cy="47.914558"
+     inkscape:cx="-103.69646"
+     inkscape:zoom="0.7"
+     inkscape:pageshadow="2"
+     inkscape:pageopacity="1"
+     borderopacity="1.0"
+     bordercolor="#666666"
+     pagecolor="#000000"
+     id="base" />
+  <metadata
+     id="metadata1337">
+    <rdf:RDF>
+      <cc:Work
+         rdf:about="">
+        <dc:format>image/svg+xml</dc:format>
+        <dc:type
+           rdf:resource="http://purl.org/dc/dcmitype/StillImage" />
+        <dc:title></dc:title>
+      </cc:Work>
+    </rdf:RDF>
+  </metadata>
+  <g
+     transform="translate(-133.26969,-52.101187)"
+     id="layer1"
+     inkscape:groupmode="layer"
+     inkscape:label="Слой 1">
+    <path
+       style="fill:url(#radialGradient1530);fill-opacity:1;fill-rule:evenodd;stroke:#000000;stroke-width:0;stroke-linecap:butt;stroke-linejoin:miter;stroke-miterlimit:4;stroke-dasharray:none;stroke-opacity:1"
+       d="m 133.26969,59.089473 50,75.000087 50,-75.000087 z"
+       id="path1528"
+       sodipodi:nodetypes="cccc" />
+    <path
+       d="m 194.29572,89.403603 -8.41706,2.562119 -2.50682,7.49308 7.17785,23.579008 9.60097,-14.40173 z"
+       style="fill:url(#linearGradient1508);fill-opacity:1;stroke-width:0;stroke-linecap:butt;stroke-linejoin:miter;stroke-miterlimit:4;stroke-dasharray:none;stroke-opacity:0.501961"
+       id="path1459" />
+    <g
+       transform="translate(135.70225,-49.385894)"
+       id="g1465">
+      <path
+         id="path1461"
+         style="fill:url(#radialGradient1504);fill-opacity:1;fill-rule:evenodd;stroke:#000000;stroke-width:0;stroke-linecap:butt;stroke-linejoin:miter;stroke-miterlimit:4;stroke-dasharray:none;stroke-opacity:1;filter:url(#filter1497)"
+         d="m 28.817436,101.36529 c 3.112699,10.74423 6.225077,21.48892 9.333984,32.23438 2.519532,0 5.039063,0 7.558594,0 -0.985406,8.09729 -2.085815,16.18202 -2.951172,24.29297 -0.06053,0.88723 1.098131,1.61652 1.76,0.9155 1.007514,-1.05482 1.676008,-2.3829 2.528566,-3.56053 7.51538,-11.37722 14.987447,-22.78299 22.482919,-34.17333 -3.239584,0 -6.479167,0 -9.71875,0 2.887267,-6.79562 5.775365,-13.59088 8.662109,-20.38672 -13.284505,0 -26.56901,0 -39.853516,0 0.06576,0.22591 0.131511,0.45182 0.197266,0.67773 z" />
+      <path
+         sodipodi:nodetypes="cccccccc"
+         id="path1463"
+         d="m 30.735882,102.2764 h 35.342242 l -8.662729,20.3854 h 9.173783 l -22.106472,33.62346 3.027029,-24.27377 H 39.34604 Z"
+         style="fill:url(#radialGradient1506);fill-opacity:1;fill-rule:evenodd;stroke:#000000;stroke-width:0;stroke-linecap:butt;stroke-linejoin:miter;stroke-miterlimit:4;stroke-dasharray:none;stroke-opacity:1" />
+    </g>
+  </g>
+</svg>

docs/hugo/static/custom.css

@@ -0,0 +1,138 @@
+/* Global customization */
+
+:root {
+  --code-max-height: 60rem;
+}
+
+/* Light mode theming */
+:root,
+:root[color-mode="light"] {
+  --header-background: #404050;
+  --header-font-color: #ffffff;
+
+  --body-background: #ffffff;
+  --body-font-color: #343a40;
+
+  --button-background: #62cb97;
+  --button-border-color: #4ec58a;
+
+  --link-color: #c54e8a;
+  --link-color-visited: #c54e8a;
+
+  --code-background: #f5f6f8;
+  --code-accent-color: #e3e7eb;
+  --code-accent-color-lite: #eff1f3;
+
+  --accent-color: #e9ecef;
+  --accent-color-lite: #f8f9fa;
+
+  --control-icons: #b2bac1;
+
+  --footer-background: #606070;
+  --footer-font-color: #ffffff;
+  --footer-link-color: #ffcc5c;
+  --footer-link-color-visited: #ffcc5c;
+}
+@media (prefers-color-scheme: light) {
+  :root {
+    --header-background: #404050;
+    --header-font-color: #ffffff;
+
+    --body-background: #ffffff;
+    --body-font-color: #343a40;
+
+    --button-background: #62cb97;
+    --button-border-color: #4ec58a;
+
+    --link-color: #c54e8a;
+    --link-color-visited: #c54e8a;
+
+    --code-background: #f5f6f8;
+    --code-accent-color: #e3e7eb;
+    --code-accent-color-lite: #eff1f3;
+
+    --accent-color: #e9ecef;
+    --accent-color-lite: #f8f9fa;
+
+    --control-icons: #b2bac1;
+
+    --footer-background: #606070;
+    --footer-font-color: #ffffff;
+    --footer-link-color: #ffcc5c;
+    --footer-link-color-visited: #ffcc5c;
+  }
+}
+
+/* Dark mode theming */
+:root[color-mode="dark"] {
+  --header-background: #202830;
+  --header-font-color: #ffffff;
+
+  --body-background: #343a44;
+  --body-font-color: #ced3d8;
+
+  --button-background: #62cb97;
+  --button-border-color: #4ec58a;
+
+  --link-color: #7ac29e;
+  --link-color-visited: #7ac29e;
+
+  --code-background: #2f353a;
+  --code-accent-color: #262b2f;
+  --code-accent-color-lite: #2b3035;
+
+  --accent-color: #2b3035;
+  --accent-color-lite: #2f353a;
+
+  --control-icons: #b2bac1;
+
+  --footer-background: #2f333e;
+  --footer-font-color: #cccccc;
+  --footer-link-color: #7ac29e;
+  --footer-link-color-visited: #7ac29e;
+}
+@media (prefers-color-scheme: dark) {
+  :root {
+    --header-background: #404070;
+    --header-font-color: #ffffff;
+
+    --body-background: #343a40;
+    --body-font-color: #ced3d8;
+
+    --button-background: #62cb97;
+    --button-border-color: #4ec58a;
+
+    --link-color: #7ac29e;
+    --link-color-visited: #7ac29e;
+
+    --code-background: #2f353a;
+    --code-accent-color: #262b2f;
+    --code-accent-color-lite: #2b3035;
+
+    --accent-color: #2b3035;
+    --accent-color-lite: #2f353a;
+
+    --control-icons: #b2bac1;
+
+    --footer-background: #2f333e;
+    --footer-font-color: #cccccc;
+    --footer-link-color: #7ac29e;
+    --footer-link-color-visited: #7ac29e;
+  }
+}
+
+.gdoc-brand__img {
+  width: 48px;
+  height: auto;
+  margin-top: -4px;
+  margin-bottom: -4px;
+}
+
+.gdoc-menu-header > span {
+  display: flex;
+  flex-direction: row-reverse;
+}
+
+span.gdoc-language {
+  margin-right: 20px;
+}

docs/hugo/static/favicon/favicon.svg

@@ -0,0 +1,196 @@
+<?xml version="1.0" encoding="UTF-8" standalone="no"?>
+<svg
+   xmlns:osb="http://www.openswatchbook.org/uri/2009/osb"
+   xmlns:dc="http://purl.org/dc/elements/1.1/"
+   xmlns:cc="http://creativecommons.org/ns#"
+   xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
+   xmlns:svg="http://www.w3.org/2000/svg"
+   xmlns="http://www.w3.org/2000/svg"
+   xmlns:xlink="http://www.w3.org/1999/xlink"
+   xmlns:sodipodi="http://sodipodi.sourceforge.net/DTD/sodipodi-0.dtd"
+   xmlns:inkscape="http://www.inkscape.org/namespaces/inkscape"
+   sodipodi:docname="favicon.svg"
+   inkscape:version="1.0.2 (e86c870879, 2021-01-15)"
+   id="svg1340"
+   version="1.1"
+   viewBox="0 0 100 100"
+   height="100mm"
+   width="100mm"
+   inkscape:export-filename="/var/home/vitali/SVN/vitastor/docs/static/favicon/favicon-64x64.png"
+   inkscape:export-xdpi="16.26"
+   inkscape:export-ydpi="16.26">
+  <defs
+     id="defs1334">
+    <linearGradient
+       osb:paint="gradient"
+       id="linearGradient866">
+      <stop
+         id="stop862"
+         offset="0"
+         style="stop-color:#c0c0c0;stop-opacity:1" />
+      <stop
+         id="stop864"
+         offset="1"
+         style="stop-color:#000000;stop-opacity:0" />
+    </linearGradient>
+    <linearGradient
+       id="linearGradient846"
+       osb:paint="gradient">
+      <stop
+         style="stop-color:#ffd42a;stop-opacity:1"
+         offset="0"
+         id="stop842" />
+      <stop
+         style="stop-color:#ffa200;stop-opacity:1"
+         offset="1"
+         id="stop844" />
+    </linearGradient>
+    <radialGradient
+       r="50"
+       fy="159.11139"
+       fx="202.36813"
+       cy="159.11139"
+       cx="202.36813"
+       gradientTransform="matrix(1.2462942,-1.2279529,0.77712408,0.78873143,-190.96813,230.1331)"
+       gradientUnits="userSpaceOnUse"
+       id="radialGradient1530"
+       xlink:href="#linearGradient1352"
+       inkscape:collect="always" />
+    <linearGradient
+       inkscape:collect="always"
+       id="linearGradient1352">
+      <stop
+         style="stop-color:#00c9e6;stop-opacity:1"
+         offset="0"
+         id="stop1348" />
+      <stop
+         style="stop-color:#5240d3;stop-opacity:1"
+         offset="1"
+         id="stop1350" />
+    </linearGradient>
+    <linearGradient
+       y2="62.555599"
+       x2="51.484566"
+       y1="62.555599"
+       x1="38.105473"
+       gradientTransform="rotate(-16.930773,271.11609,-412.42594)"
+       gradientUnits="userSpaceOnUse"
+       id="linearGradient1508"
+       xlink:href="#linearGradient1323"
+       inkscape:collect="always" />
+    <linearGradient
+       inkscape:collect="always"
+       id="linearGradient1323">
+      <stop
+         style="stop-color:#000000;stop-opacity:0.47178105"
+         offset="0"
+         id="stop1319" />
+      <stop
+         style="stop-color:#eeaaff;stop-opacity:0;"
+         offset="1"
+         id="stop1321" />
+    </linearGradient>
+    <filter
+       style="color-interpolation-filters:sRGB"
+       inkscape:label="Drop Shadow"
+       id="filter1497"
+       width="2"
+       height="2"
+       x="-0.5"
+       y="-0.5">
+      <feFlood
+         flood-opacity="0.498039"
+         flood-color="rgb(0,0,0)"
+         result="flood"
+         id="feFlood1487" />
+      <feComposite
+         in="flood"
+         in2="SourceGraphic"
+         operator="in"
+         result="composite1"
+         id="feComposite1489" />
+      <feGaussianBlur
+         in="composite1"
+         stdDeviation="6"
+         result="blur"
+         id="feGaussianBlur1491" />
+      <feOffset
+         dx="0"
+         dy="6"
+         result="offset"
+         id="feOffset1493" />
+      <feComposite
+         in="offset"
+         in2="offset"
+         operator="atop"
+         result="composite2"
+         id="feComposite1495" />
+    </filter>
+    <radialGradient
+       r="21.541935"
+       fy="24.614815"
+       fx="45.312912"
+       cy="24.614815"
+       cx="45.312912"
+       gradientTransform="matrix(1.6678615,0.20004527,-0.19318681,1.6106796,108.48083,22.966962)"
+       gradientUnits="userSpaceOnUse"
+       id="radialGradient1506"
+       xlink:href="#linearGradient846"
+       inkscape:collect="always" />
+  </defs>
+  <sodipodi:namedview
+     inkscape:window-maximized="1"
+     inkscape:window-y="0"
+     inkscape:window-x="0"
+     inkscape:window-height="992"
+     inkscape:window-width="1920"
+     fit-margin-bottom="0"
+     fit-margin-right="0"
+     fit-margin-left="0"
+     fit-margin-top="0"
+     showgrid="false"
+     inkscape:document-rotation="0"
+     inkscape:current-layer="layer1"
+     inkscape:document-units="mm"
+     inkscape:cy="83.752268"
+     inkscape:cx="-103.69645"
+     inkscape:zoom="0.7"
+     inkscape:pageshadow="2"
+     inkscape:pageopacity="0"
+     borderopacity="1.0"
+     bordercolor="#666666"
+     pagecolor="#000000"
+     id="base" />
+  <metadata
+     id="metadata1337">
+    <rdf:RDF>
+      <cc:Work
+         rdf:about="">
+        <dc:format>image/svg+xml</dc:format>
+        <dc:type
+           rdf:resource="http://purl.org/dc/dcmitype/StillImage" />
+        <dc:title></dc:title>
+      </cc:Work>
+    </rdf:RDF>
+  </metadata>
+  <g
+     transform="translate(-133.26969,-35.630924)"
+     id="layer1"
+     inkscape:groupmode="layer"
+     inkscape:label="Слой 1">
+    <path
+       style="fill:url(#radialGradient1530);fill-opacity:1;fill-rule:evenodd;stroke:#000000;stroke-width:0;stroke-linecap:butt;stroke-linejoin:miter;stroke-miterlimit:4;stroke-dasharray:none;stroke-opacity:1"
+       d="m 133.26969,59.089473 50,75.000087 50,-75.000087 z"
+       id="path1528"
+       sodipodi:nodetypes="cccc" />
+    <path
+       d="m 194.29572,89.403603 -8.41706,2.562119 -2.50682,7.49308 7.17785,23.579008 9.60097,-14.40173 z"
+       style="fill:url(#linearGradient1508);fill-opacity:1;stroke-width:0;stroke-linecap:butt;stroke-linejoin:miter;stroke-miterlimit:4;stroke-dasharray:none;stroke-opacity:0.501961"
+       id="path1459" />
+    <path
+       sodipodi:nodetypes="cccccccc"
+       id="path1463"
+       d="m 157.01826,35.630924 h 53.91343 l -13.21471,31.09726 h 13.99432 l -33.7227,51.291496 4.61762,-37.02885 h -12.45344 z"
+       style="fill:url(#radialGradient1506);fill-opacity:1;fill-rule:evenodd;stroke:#000000;stroke-width:0;stroke-linecap:butt;stroke-linejoin:miter;stroke-miterlimit:4;stroke-dasharray:none;stroke-opacity:1" />
+  </g>
+</svg>

docs/params/common.yml

@@ -0,0 +1,35 @@
+- name: config_path
+  type: string
+  default: "/etc/vitastor/vitastor.conf"
+  info: |
+    Path to the JSON configuration file. Configuration file is optional,
+    a non-existing configuration file does not prevent Vitastor from
+    running if required parameters are specified.
+  info_ru: |
+    Путь к файлу конфигурации в формате JSON. Файл конфигурации необязателен,
+    без него Vitastor тоже будет работать, если переданы необходимые параметры.
+- name: etcd_address
+  type: string or array of strings
+  type_ru: строка или массив строк
+  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.
+  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"]`.
+- name: etcd_prefix
+  type: string
+  default: "/vitastor"
+  info: |
+    Prefix for all keys in etcd used by Vitastor. You can change prefix and, for
+    example, use a single etcd cluster for multiple Vitastor clusters.
+  info_ru: |
+    Префикс для ключей etcd, которые использует Vitastor. Вы можете задать другой
+    префикс, например, чтобы запустить несколько кластеров Vitastor с одним
+    кластером etcd.
+- name: log_level
+  type: int
+  default: 0
+  info: Log level. Raise if you want more verbose output.
+  info_ru: Уровень логгирования. Повысьте, если хотите более подробный вывод.

docs/params/head/common.en.md

@@ -0,0 +1,6 @@
+---
+title: Common Parameters
+weight: 1
+---
+
+These are the most common parameters which apply to all components of Vitastor.

docs/params/head/common.ru.md

@@ -0,0 +1,6 @@
+---
+title: Общие параметры
+weight: 1
+---
+
+Это наиболее общие параметры, используемые всеми компонентами Vitastor.

docs/params/head/layout-cluster.en.md

@@ -0,0 +1,7 @@
+---
+title: Cluster-Wide Disk Layout Parameters
+weight: 2
+---
+
+These parameters apply to clients and OSDs, are fixed at the moment of OSD drive
+initialization and can't be changed after it without losing data.

docs/params/head/layout-cluster.ru.md

@@ -0,0 +1,7 @@
+---
+title: Дисковые параметры уровня кластера
+weight: 2
+---
+
+Данные параметры используются клиентами и OSD, задаются в момент инициализации
+диска OSD и не могут быть изменены после этого без потери данных.

docs/params/head/layout-osd.en.md

@@ -0,0 +1,7 @@
+---
+title: OSD Disk Layout Parameters
+weight: 3
+---
+
+These parameters apply to OSDs, are fixed at the moment of OSD drive
+initialization and can't be changed after it without losing data.

docs/params/head/layout-osd.ru.md

@@ -0,0 +1,8 @@
+---
+title: Дисковые параметры OSD
+weight: 3
+---
+
+Данные параметры используются только OSD и, также как и общекластерные
+дисковые параметры, задаются в момент инициализации дисков OSD и не могут быть
+изменены после этого без потери данных.

docs/params/head/monitor.en.md

@@ -0,0 +1,6 @@
+---
+title: Monitor Parameters
+weight: 6
+---
+
+These parameters only apply to Monitors.

docs/params/head/monitor.ru.md

@@ -0,0 +1,6 @@
+---
+title: Параметры мониторов
+weight: 6
+---
+
+Данные параметры используются только мониторами Vitastor.

docs/params/head/network.en.md

@@ -0,0 +1,7 @@
+---
+title: Network Protocol Parameters
+weight: 4
+---
+
+These parameters apply to clients and OSDs and affect network connection logic
+between clients, OSDs and etcd.

docs/params/head/network.ru.md

@@ -0,0 +1,7 @@
+---
+title: Параметры сетевого протокола
+weight: 4
+---
+
+Данные параметры используются клиентами и OSD и влияют на логику сетевого
+взаимодействия между клиентами, OSD, а также etcd.

docs/params/head/osd.en.md

@@ -0,0 +1,7 @@
+---
+title: Runtime OSD Parameters
+weight: 5
+---
+
+These parameters only apply to OSDs, are not fixed at the moment of OSD drive
+initialization and can be changed with an OSD restart.

docs/params/head/osd.ru.md

@@ -0,0 +1,8 @@
+---
+title: Изменяемые параметры OSD
+weight: 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.

docs/params/layout-cluster.yml

@@ -0,0 +1,200 @@
+- name: block_size
+  type: int
+  default: 131072
+  info: |
+    Size of objects (data blocks) into which all physical and virtual drives are
+    subdivided in Vitastor. One of current main settings in Vitastor, affects
+    memory usage, write amplification and I/O load distribution effectiveness.
+
+    Recommended default block size is 128 KB for SSD and 4 MB for HDD. In fact,
+    it's possible to use 4 MB for SSD too - it will lower memory usage, but
+    may increase average WA and reduce linear performance.
+
+    OSDs with different block sizes (for example, SSD and SSD+HDD OSDs) can
+    currently coexist in one etcd instance only within separate Vitastor
+    clusters with different etcd_prefix'es.
+
+    Also block size can't be changed after OSD initialization without losing
+    data.
+
+    You must always specify block_size in etcd in /vitastor/config/global if
+    you change it so all clients can know about it.
+
+    OSD memory usage is roughly (SIZE / BLOCK * 68 bytes) which is roughly
+    544 MB per 1 TB of used disk space with the default 128 KB block size.
+  info_ru: |
+    Размер объектов (блоков данных), на которые делятся физические и виртуальные
+    диски в Vitastor. Одна из ключевых на данный момент настроек, влияет на
+    потребление памяти, объём избыточной записи (write amplification) и
+    эффективность распределения нагрузки по OSD.
+
+    Рекомендуемые по умолчанию размеры блока - 128 килобайт для SSD и 4
+    мегабайта для HDD. В принципе, для SSD можно тоже использовать 4 мегабайта,
+    это понизит использование памяти, но ухудшит распределение нагрузки и в
+    среднем увеличит WA.
+
+    OSD с разными размерами блока (например, SSD и SSD+HDD OSD) на данный
+    момент могут сосуществовать в рамках одного etcd только в виде двух независимых
+    кластеров Vitastor с разными etcd_prefix.
+
+    Также размер блока нельзя менять после инициализации OSD без потери данных.
+
+    Если вы меняете размер блока, обязательно прописывайте его в etcd в
+    /vitastor/config/global, дабы все клиенты его знали.
+
+    Потребление памяти OSD составляет примерно (РАЗМЕР / БЛОК * 68 байт),
+    т.е. примерно 544 МБ памяти на 1 ТБ занятого места на диске при
+    стандартном 128 КБ блоке.
+- name: bitmap_granularity
+  type: int
+  default: 4096
+  info: |
+    Required virtual disk write alignment ("sector size"). Must be a multiple
+    of disk_alignment. It's called bitmap granularity because Vitastor tracks
+    an allocation bitmap for each object containing 2 bits per each
+    (bitmap_granularity) bytes.
+
+    This parameter can't be changed after OSD initialization without losing
+    data. Also it's fixed for the whole Vitastor cluster i.e. two different
+    values can't be used in a single Vitastor cluster.
+
+    Clients MUST be aware of this parameter value, so put it into etcd key
+    /vitastor/config/global if you change it for any reason.
+  info_ru: |
+    Требуемое выравнивание записи на виртуальные диски (размер их "сектора").
+    Должен быть кратен disk_alignment. Называется гранулярностью битовой карты
+    потому, что Vitastor хранит битовую карту для каждого объекта, содержащую
+    по 2 бита на каждые (bitmap_granularity) байт.
+
+    Данный параметр нельзя менять после инициализации OSD без потери данных.
+    Также он фиксирован для всего кластера Vitastor, т.е. разные значения
+    не могут сосуществовать в одном кластере.
+
+    Клиенты ДОЛЖНЫ знать правильное значение этого параметра, так что если вы
+    его меняете, обязательно прописывайте изменённое значение в etcd в ключ
+    /vitastor/config/global.
+- name: immediate_commit
+  type: string
+  default: false
+  info: |
+    Another parameter which is really important for performance.
+
+    Desktop SSDs are very fast (100000+ iops) for simple random writes
+    without cache flush. However, they are really slow (only around 1000 iops)
+    if you try to fsync() each write, that is, when you want to guarantee that
+    each change gets immediately persisted to the physical media.
+
+    Server-grade SSDs with "Advanced/Enhanced Power Loss Protection" or with
+    "Supercapacitor-based Power Loss Protection", on the other hand, are equally
+    fast with and without fsync because their cache is protected from sudden
+    power loss by a built-in supercapacitor-based "UPS".
+
+    Some software-defined storage systems always fsync each write and thus are
+    really slow when used with desktop SSDs. Vitastor, however, can also
+    efficiently utilize desktop SSDs by postponing fsync until the client calls
+    it explicitly.
+
+    This is what this parameter regulates. When it's set to "all" the whole
+    Vitastor cluster commits each change to disks immediately and clients just
+    ignore fsyncs because they know for sure that they're unneeded. This reduces
+    the amount of network roundtrips performed by clients and improves
+    performance. So it's always better to use server grade SSDs with
+    supercapacitors even with Vitastor, especially given that they cost only
+    a bit more than desktop models.
+
+    There is also a common SATA SSD (and HDD too!) firmware bug (or feature)
+    that makes server SSDs which have supercapacitors slow with fsync. To check
+    if your SSDs are affected, compare benchmark results from `fio -name=test
+    -ioengine=libaio -direct=1 -bs=4k -rw=randwrite -iodepth=1` with and without
+    `-fsync=1`. Results should be the same. If fsync=1 result is worse you can
+    try to work around this bug by "disabling" drive write-back cache by running
+    `hdparm -W 0 /dev/sdXX` or `echo write through > /sys/block/sdXX/device/scsi_disk/*/cache_type`
+    (IMPORTANT: don't mistake it with `/sys/block/sdXX/queue/write_cache` - it's
+    unsafe to change by hand). The same may apply to newer HDDs with internal
+    SSD cache or "media-cache" - for example, a lot of Seagate EXOS drives have
+    it (they have internal SSD cache even though it's not stated in datasheets).
+
+    This parameter must be set both in etcd in /vitastor/config/global and in
+    OSD command line or configuration. Setting it to "all" or "small" requires
+    enabling disable_journal_fsync and disable_meta_fsync, setting it to "all"
+    also requires enabling disable_data_fsync.
+
+    TLDR: For optimal performance, set immediate_commit to "all" if you only use
+    SSDs with supercapacitor-based power loss protection (nonvolatile
+    write-through cache) for both data and journals in the whole Vitastor
+    cluster. Set it to "small" if you only use such SSDs for journals. Leave
+    empty if your drives have write-back cache.
+  info_ru: |
+    Ещё один важный для производительности параметр.
+
+    Модели SSD для настольных компьютеров очень быстрые (100000+ операций в
+    секунду) при простой случайной записи без сбросов кэша. Однако они очень
+    медленные (всего порядка 1000 iops), если вы пытаетесь сбрасывать кэш после
+    каждой записи, то есть, если вы пытаетесь гарантировать, что каждое
+    изменение физически записывается в энергонезависимую память.
+
+    С другой стороны, серверные SSD с конденсаторами - функцией, называемой
+    "Advanced/Enhanced Power Loss Protection" или просто "Supercapacitor-based
+    Power Loss Protection" - одинаково быстрые и со сбросом кэша, и без
+    него, потому что их кэш защищён от потери питания встроенным "источником
+    бесперебойного питания" на основе суперконденсаторов и на самом деле они
+    его никогда не сбрасывают.
+
+    Некоторые программные СХД всегда сбрасывают кэши дисков при каждой записи
+    и поэтому работают очень медленно с настольными SSD. Vitastor, однако, может
+    откладывать fsync до явного его вызова со стороны клиента и таким образом
+    эффективно утилизировать настольные SSD.
+
+    Данный параметр влияет как раз на это. Когда он установлен в значение "all",
+    весь кластер Vitastor мгновенно фиксирует каждое изменение на физические
+    носители и клиенты могут просто игнорировать запросы fsync, т.к. они точно
+    знают, что fsync-и не нужны. Это уменьшает число необходимых обращений к OSD
+    по сети и улучшает производительность. Поэтому даже с Vitastor лучше всегда
+    использовать только серверные модели SSD с суперконденсаторами, особенно
+    учитывая то, что стоят они ненамного дороже настольных.
+
+    Также в прошивках SATA SSD (и даже HDD!) очень часто встречается либо баг,
+    либо просто особенность логики, из-за которой серверные SSD, имеющие
+    конденсаторы и защиту от потери питания, всё равно медленно работают с
+    fsync. Чтобы понять, подвержены ли этой проблеме ваши SSD, сравните
+    результаты тестов `fio -name=test -ioengine=libaio -direct=1 -bs=4k
+    -rw=randwrite -iodepth=1` без и с опцией `-fsync=1`. Результаты должны
+    быть одинаковые. Если результат с `fsync=1` хуже, вы можете попробовать
+    обойти проблему, "отключив" кэш записи диска командой `hdparm -W 0 /dev/sdXX`
+    либо `echo write through > /sys/block/sdXX/device/scsi_disk/*/cache_type`
+    (ВАЖНО: не перепутайте с `/sys/block/sdXX/queue/write_cache` - этот параметр
+    менять руками небезопасно). Такая же проблема может встречаться и в новых
+    HDD-дисках с внутренним SSD или "медиа" кэшем - например, она встречается во
+    многих дисках Seagate EXOS (у них есть внутренний SSD-кэш, хотя это и не
+    указано в спецификациях).
+
+    Данный параметр нужно указывать и в etcd в /vitastor/config/global, и в
+    командной строке или конфигурации OSD. Значения "all" и "small" требуют
+    включения disable_journal_fsync и disable_meta_fsync, значение "all" также
+    требует включения disable_data_fsync.
+
+    Итого, вкратце: для оптимальной производительности установите
+    immediate_commit в значение "all", если вы используете в кластере только SSD
+    с суперконденсаторами и для данных, и для журналов. Если вы используете
+    такие SSD для всех журналов, но не для данных - можете установить параметр
+    в "small". Если и какие-то из дисков журналов имеют волатильный кэш записи -
+    оставьте параметр пустым.
+- name: client_dirty_limit
+  type: int
+  default: 33554432
+  info: |
+    Without immediate_commit=all this parameter sets the limit of "dirty"
+    (not committed by fsync) data allowed by the client before forcing an
+    additional fsync and committing the data. Also note that the client always
+    holds a copy of uncommitted data in memory so this setting also affects
+    RAM usage of clients.
+
+    This parameter doesn't affect OSDs themselves.
+  info_ru: |
+    При работе без immediate_commit=all - это лимит объёма "грязных" (не
+    зафиксированных fsync-ом) данных, при достижении которого клиент будет
+    принудительно вызывать fsync и фиксировать данные. Также стоит иметь в виду,
+    что в этом случае до момента fsync клиент хранит копию незафиксированных
+    данных в памяти, то есть, настройка влияет на потребление памяти клиентами.
+
+    Параметр не влияет на сами OSD.

docs/params/layout-osd.yml

@@ -0,0 +1,205 @@
+- name: data_device
+  type: string
+  info: |
+    Path to the block device to use for data. It's highly recommendded to use
+    stable paths for all device names: `/dev/disk/by-partuuid/xxx...` instead
+    of just `/dev/sda` or `/dev/nvme0n1` to not mess up after server restart.
+    Files can also be used instead of block devices, but this is implemented
+    only for testing purposes and not for production.
+  info_ru: |
+    Путь к диску (блочному устройству) для хранения данных. Крайне рекомендуется
+    использовать стабильные пути: `/dev/disk/by-partuuid/xxx...` вместо простых
+    `/dev/sda` или `/dev/nvme0n1`, чтобы пути не могли спутаться после
+    перезагрузки сервера. Также вместо блочных устройств можно указывать файлы,
+    но это реализовано только для тестирования, а не для боевой среды.
+- name: meta_device
+  type: string
+  info: |
+    Path to the block device to use for the metadata. Metadata must be on a fast
+    SSD or performance will suffer. If this option is skipped, `data_device` is
+    used for the metadata.
+  info_ru: |
+    Путь к диску метаданных. Метаданные должны располагаться на быстром
+    SSD-диске, иначе производительность пострадает. Если эта опция не указана,
+    для метаданных используется `data_device`.
+- name: journal_device
+  type: string
+  info: |
+    Path to the block device to use for the journal. Journal must be on a fast
+    SSD or performance will suffer. If this option is skipped, `meta_device` is
+    used for the journal, and if it's also empty, journal is put on
+    `data_device`. It's almost always fine to put metadata and journal on the
+    same device, in this case you only need to set `meta_device`.
+  info_ru: |
+    Путь к диску журнала. Журнал должен располагаться на быстром SSD-диске,
+    иначе производительность пострадает. Если эта опция не указана,
+    для журнала используется `meta_device`, если же пуста и она, журнал
+    располагается на `data_device`. Нормально располагать журнал и метаданные
+    на одном устройстве, в этом случае достаточно указать только `meta_device`.
+- name: journal_offset
+  type: int
+  default: 0
+  info: Offset on the device in bytes where the journal is stored.
+  info_ru: Смещение на устройстве в байтах, по которому располагается журнал.
+- name: journal_size
+  type: int
+  info: |
+    Journal size in bytes. Doesn't have to be large, 16-32 MB is usually fine.
+    By default, the whole journal device will be used for the journal. You must
+    set it to some value manually (or use make-osd.sh) if you colocate the
+    journal with data or metadata.
+  info_ru: |
+    Размер журнала в байтах. Большим быть не обязан, 16-32 МБ обычно достаточно.
+    По умолчанию для журнала используется всё устройство журнала. Если же вы
+    размещаете журнал на устройстве данных или метаданных, то вы должны
+    установить эту опцию в какое-то значение сами (или использовать скрипт
+    make-osd.sh).
+- name: meta_offset
+  type: int
+  default: 0
+  info: |
+    Offset on the device in bytes where the metadata area is stored.
+    Again, set it to something if you colocate metadata with journal or data.
+  info_ru: |
+    Смещение на устройстве в байтах, по которому располагаются метаданные.
+    Эту опцию нужно задать, если метаданные у вас хранятся на том же
+    устройстве, что данные или журнал.
+- name: data_offset
+  type: int
+  default: 0
+  info: |
+    Offset on the device in bytes where the data area is stored.
+    Again, set it to something if you colocate data with journal or metadata.
+  info_ru: |
+    Смещение на устройстве в байтах, по которому располагаются данные.
+    Эту опцию нужно задать, если данные у вас хранятся на том же
+    устройстве, что метаданные или журнал.
+- name: data_size
+  type: int
+  info: |
+    Data area size in bytes. By default, the whole data device up to the end
+    will be used for the data area, but you can restrict it if you want to use
+    a smaller part. Note that there is no option to set metadata area size -
+    it's derived from the data area size.
+  info_ru: |
+    Размер области данных в байтах. По умолчанию под данные будет использована
+    вся доступная область устройства данных до конца устройства, но вы можете
+    использовать эту опцию, чтобы ограничить её меньшим размером. Заметьте, что
+    опции размера области метаданных нет - она вычисляется из размера области
+    данных автоматически.
+- name: meta_block_size
+  type: int
+  default: 4096
+  info: |
+    Physical block size of the metadata device. 4096 for most current
+    HDDs and SSDs.
+  info_ru: |
+    Размер физического блока устройства метаданных. 4096 для большинства
+    современных SSD и HDD.
+- name: journal_block_size
+  type: int
+  default: 4096
+  info: |
+    Physical block size of the journal device. Must be a multiple of
+    `disk_alignment`. 4096 for most current HDDs and SSDs.
+  info_ru: |
+    Размер физического блока устройства журнала. Должен быть кратен
+    `disk_alignment`. 4096 для большинства современных SSD и HDD.
+- name: disable_data_fsync
+  type: bool
+  default: false
+  info: |
+    Do not issue fsyncs to the data device, i.e. do not flush its cache.
+    Safe ONLY if your data device has write-through cache. If you disable
+    the cache yourself using `hdparm` or `scsi_disk/cache_type` then make sure
+    that the cache disable command is run every time before starting Vitastor
+    OSD, for example, in the systemd unit. See also `immediate_commit` option
+    for the instructions to disable cache and how to benefit from it.
+  info_ru: |
+    Не отправлять fsync-и устройству данных, т.е. не сбрасывать его кэш.
+    Безопасно, ТОЛЬКО если ваше устройство данных имеет кэш со сквозной
+    записью (write-through). Если вы отключаете кэш через `hdparm` или
+    `scsi_disk/cache_type`, то удостоверьтесь, что команда отключения кэша
+    выполняется перед каждым запуском Vitastor OSD, например, в systemd unit-е.
+    Смотрите также опцию `immediate_commit` для инструкций по отключению кэша
+    и о том, как из этого извлечь выгоду.
+- name: disable_meta_fsync
+  type: bool
+  default: false
+  info: |
+    Same as disable_data_fsync, but for the metadata device. If the metadata
+    device is not set or if the data device is used for the metadata the option
+    is ignored and disable_data_fsync value is used instead of it.
+  info_ru: |
+    То же, что disable_data_fsync, но для устройства метаданных. Если устройство
+    метаданных не задано или если оно равно устройству данных, значение опции
+    игнорируется и вместо него используется значение опции disable_data_fsync.
+- name: disable_journal_fsync
+  type: bool
+  default: false
+  info: |
+    Same as disable_data_fsync, but for the journal device. If the journal
+    device is not set or if the metadata device is used for the journal the
+    option is ignored and disable_meta_fsync value is used instead of it. If
+    the same device is used for data, metadata and journal the option is also
+    ignored and disable_data_fsync value is used instead of it.
+  info_ru: |
+    То же, что disable_data_fsync, но для устройства журнала. Если устройство
+    журнала не задано или если оно равно устройству метаданных, значение опции
+    игнорируется и вместо него используется значение опции disable_meta_fsync.
+    Если одно и то же устройство используется и под данные, и под журнал, и под
+    метаданные - значение опции также игнорируется и вместо него используется
+    значение опции disable_data_fsync.
+- name: disable_device_lock
+  type: bool
+  default: false
+  info: |
+    Do not lock data, metadata and journal block devices exclusively with
+    flock(). Though it's not recommended, but you can use it you want to run
+    multiple OSD with a single device and different offsets, without using
+    partitions.
+  info_ru: |
+    Не блокировать устройства данных, метаданных и журнала от открытия их
+    другими OSD с помощью flock(). Так делать не рекомендуется, но теоретически
+    вы можете это использовать, чтобы запускать несколько OSD на одном
+    устройстве с разными смещениями и без использования разделов.
+- name: disk_alignment
+  type: int
+  default: 4096
+  info: |
+    Required physical disk write alignment. Most current SSD and HDD drives
+    use 4 KB physical sectors even if they report 512 byte logical sector
+    size, so 4 KB is a good default setting.
+
+    Note, however, that physical sector size also affects WA, because with block
+    devices it's impossible to write anything smaller than a block. So, when
+    Vitastor has to write a single metadata entry that's only about 32 bytes in
+    size, it actually has to write the whole 4 KB sector.
+
+    Because of this it can actually be beneficial to use SSDs which work well
+    with 512 byte sectors and use 512 byte disk_alignment, journal_block_size
+    and meta_block_size. But the only SSD that may fit into this category is
+    Intel Optane (probably, not tested yet).
+
+    Clients don't need to be aware of disk_alignment, so it's not required to
+    put a modified value into etcd key /vitastor/config/global.
+  info_ru: |
+    Требуемое выравнивание записи на физические диски. Почти все современные
+    SSD и HDD диски используют 4 КБ физические секторы, даже если показывают
+    логический размер сектора 512 байт, поэтому 4 КБ - хорошее значение по
+    умолчанию.
+
+    Однако стоит понимать, что физический размер сектора тоже влияет на
+    избыточную запись (WA), потому что ничего меньше блока (сектора) на блочное
+    устройство записать невозможно. Таким образом, когда Vitastor-у нужно
+    записать на диск всего лишь одну 32-байтную запись метаданных, фактически
+    приходится перезаписывать 4 КБ сектор целиком.
+
+    Поэтому, на самом деле, может быть выгодно найти SSD, хорошо работающие с
+    меньшими, 512-байтными, блоками и использовать 512-байтные disk_alignment,
+    journal_block_size и meta_block_size. Однако единственные SSD, которые
+    теоретически могут попасть в эту категорию - это Intel Optane (но и это
+    пока не проверялось автором).
+
+    Клиентам не обязательно знать про disk_alignment, так что помещать значение
+    этого параметра в etcd в /vitastor/config/global не нужно.

docs/params/monitor.yml

@@ -0,0 +1,65 @@
+- name: etcd_mon_ttl
+  type: sec
+  min: 10
+  default: 30
+  info: Monitor etcd lease refresh interval in seconds
+  info_ru: Интервал обновления etcd резервации (lease) монитором
+- name: etcd_mon_timeout
+  type: ms
+  default: 1000
+  info: etcd request timeout used by monitor
+  info_ru: Таймаут выполнения запросов к etcd от монитора
+- name: etcd_mon_retries
+  type: int
+  default: 5
+  info: Maximum number of attempts for one monitor etcd request
+  info_ru: Максимальное число попыток выполнения запросов к etcd монитором
+- name: mon_change_timeout
+  type: ms
+  min: 100
+  default: 1000
+  info: Optimistic retry interval for monitor etcd modification requests
+  info_ru: Время повтора при коллизиях при запросах модификации в etcd, производимых монитором
+- name: mon_stats_timeout
+  type: ms
+  min: 100
+  default: 1000
+  info: |
+    Interval for monitor to wait before updating aggregated statistics in
+    etcd after receiving OSD statistics updates
+  info_ru: |
+    Интервал, который монитор ожидает при изменении статистики по отдельным
+    OSD перед обновлением агрегированной статистики в etcd
+- name: osd_out_time
+  type: sec
+  default: 600
+  info: |
+    Time after which a failed OSD is removed from the data distribution.
+    I.e. time which the monitor waits before attempting to restore data
+    redundancy using other OSDs.
+  info_ru: |
+    Время, через которое отключенный OSD исключается из распределения данных.
+    То есть, время, которое монитор ожидает перед попыткой переместить данные
+    на другие OSD и таким образом восстановить избыточность хранения.
+- name: placement_levels
+  type: json
+  default: '`{"host":100,"osd":101}`'
+  info: |
+    Levels for the placement tree. You can define arbitrary tree levels by
+    defining them in this parameter. The configuration parameter value should
+    contain a JSON object with level names as keys and integer priorities as
+    values.  Smaller priority means higher level in tree. For example,
+    "datacenter" should have smaller priority than "osd". "host" and "osd"
+    levels are always predefined and can't be removed. If one of them is not
+    present in the configuration, then it is defined with the default priority
+    (100 for "host", 101 for "osd").
+  info_ru: |
+    Определения уровней для дерева размещения OSD. Вы можете определять
+    произвольные уровни, помещая их в данный параметр конфигурации. Значение
+    параметра должно содержать JSON-объект, ключи которого будут являться
+    названиями уровней, а значения - целочисленными приоритетами. Меньшие
+    приоритеты соответствуют верхним уровням дерева. Например, уровень
+    "датацентр" должен иметь меньший приоритет, чем "OSD". Уровни с названиями
+    "host" и "osd" являются предопределёнными и не могут быть удалены. Если
+    один из них отсутствует в конфигурации, он доопределяется с приоритетом по
+    умолчанию (100 для уровня "host", 101 для "osd").

docs/params/network.yml

@@ -0,0 +1,225 @@
+- name: tcp_header_buffer_size
+  type: int
+  default: 65536
+  info: |
+    Size of the buffer used to read data using an additional copy. Vitastor
+    packet headers are 128 bytes, payload is always at least 4 KB, so it is
+    usually beneficial to try to read multiple packets at once even though
+    it requires to copy the data an additional time. The rest of each packet
+    is received without an additional copy. You can try to play with this
+    parameter and see how it affects random iops and linear bandwidth if you
+    want.
+  info_ru: |
+    Размер буфера для чтения данных с дополнительным копированием. Пакеты
+    Vitastor содержат 128-байтные заголовки, за которыми следуют данные размером
+    от 4 КБ и для мелких операций ввода-вывода обычно выгодно за 1 вызов читать
+    сразу несколько пакетов, даже не смотря на то, что это требует лишний раз
+    скопировать данные. Часть каждого пакета за пределами значения данного
+    параметра читается без дополнительного копирования. Вы можете попробовать
+    поменять этот параметр и посмотреть, как он влияет на производительность
+    случайного и линейного доступа.
+- name: use_sync_send_recv
+  type: bool
+  default: false
+  info: |
+    If true, synchronous send/recv syscalls are used instead of io_uring for
+    socket communication. Useless for OSDs because they require io_uring anyway,
+    but may be required for clients with old kernel versions.
+  info_ru: |
+    Если установлено в истину, то вместо io_uring для передачи данных по сети
+    будут использоваться обычные синхронные системные вызовы send/recv. Для OSD
+    это бессмысленно, так как OSD в любом случае нуждается в io_uring, но, в
+    принципе, это может применяться для клиентов со старыми версиями ядра.
+- name: use_rdma
+  type: bool
+  default: true
+  info: |
+    Try to use RDMA for communication if it's available. Disable if you don't
+    want Vitastor to use RDMA. RDMA increases the performance, but TCP-only
+    clients can still talk to an RDMA-enabled cluster, so you don't need to
+    make sure that all clients support RDMA when enabling it.
+  info_ru: |
+    Пытаться использовать RDMA для связи при наличии доступных устройств.
+    Отключите, если вы не хотите, чтобы Vitastor использовал RDMA.
+    RDMA улучшает производительность, но 
+    Клиенты и клиентов and TCP-only clients in the cluster at the
+    same time - TCP-only clients are still able to use an RDMA-enabled cluster.
+- name: rdma_device
+  type: string
+  info: |
+    RDMA device name to use for Vitastor OSD communications (for example,
+    "rocep5s0f0"). Please note that Vitastor RDMA requires Implicit On-Demand
+    Paging (Implicit ODP) and Scatter/Gather (SG) support from the RDMA device
+    to work. For example, Mellanox ConnectX-3 and older adapters don't have
+    Implicit ODP, so they're unsupported by Vitastor. Run `ibv_devinfo -v` as
+    root to list available RDMA devices and their features.
+  info_ru: |
+    Название RDMA-устройства для связи с Vitastor OSD (например, "rocep5s0f0").
+    Имейте в виду, что поддержка RDMA в Vitastor требует функций устройства
+    Implicit On-Demand Paging (Implicit ODP) и Scatter/Gather (SG). Например,
+    адаптеры Mellanox ConnectX-3 и более старые не поддерживают Implicit ODP и
+    потому не поддерживаются в Vitastor. Запустите `ibv_devinfo -v` от имени
+    суперпользователя, чтобы посмотреть список доступных RDMA-устройств, их
+    параметры и возможности.
+- name: rdma_port_num
+  type: int
+  default: 1
+  info: |
+    RDMA device port number to use. Only for devices that have more than 1 port.
+    See `phys_port_cnt` in `ibv_devinfo -v` output to determine how many ports
+    your device has.
+  info_ru: |
+    Номер порта RDMA-устройства, который следует использовать. Имеет смысл
+    только для устройств, у которых более 1 порта. Чтобы узнать, сколько портов
+    у вашего адаптера, посмотрите `phys_port_cnt` в выводе команды
+    `ibv_devinfo -v`.
+- name: rdma_gid_index
+  type: int
+  default: 0
+  info: |
+    Global address identifier index of the RDMA device to use. Different GID
+    indexes may correspond to different protocols like RoCEv1, RoCEv2 and iWARP.
+    Search for "GID" in `ibv_devinfo -v` output to determine which GID index
+    you need.
+
+    **IMPORTANT:** If you want to use RoCEv2 (as recommended) then the correct
+    rdma_gid_index is usually 1 (IPv6) or 3 (IPv4).
+  info_ru: |
+    Номер глобального идентификатора адреса RDMA-устройства, который следует
+    использовать. Разным gid_index могут соответствовать разные протоколы связи:
+    RoCEv1, RoCEv2, iWARP. Чтобы понять, какой нужен вам - смотрите строчки со
+    словом "GID" в выводе команды `ibv_devinfo -v`.
+
+    **ВАЖНО:** Если вы хотите использовать RoCEv2 (как мы и рекомендуем), то
+    правильный rdma_gid_index, как правило, 1 (IPv6) или 3 (IPv4).
+- name: rdma_mtu
+  type: int
+  default: 4096
+  info: |
+    RDMA Path MTU to use. Must be 1024, 2048 or 4096. There is usually no
+    sense to change it from the default 4096.
+  info_ru: |
+    Максимальная единица передачи (Path MTU) для RDMA. Должно быть равно 1024,
+    2048 или 4096. Обычно нет смысла менять значение по умолчанию, равное 4096.
+- name: rdma_max_sge
+  type: int
+  default: 128
+  info: |
+    Maximum number of scatter/gather entries to use for RDMA. OSDs negotiate
+    the actual value when establishing connection anyway, so it's usually not
+    required to change this parameter.
+  info_ru: |
+    Максимальное число записей разделения/сборки (scatter/gather) для RDMA.
+    OSD в любом случае согласовывают реальное значение при установке соединения,
+    так что менять этот параметр обычно не нужно.
+- name: rdma_max_msg
+  type: int
+  default: 1048576
+  info: Maximum size of a single RDMA send or receive operation in bytes.
+  info_ru: Максимальный размер одной RDMA-операции отправки или приёма.
+- name: rdma_max_recv
+  type: int
+  default: 8
+  info: |
+    Maximum number of parallel RDMA receive operations. Note that this number
+    of receive buffers `rdma_max_msg` in size are allocated for each client,
+    so this setting actually affects memory usage. This is because RDMA receive
+    operations are (sadly) still not zero-copy in Vitastor. It may be fixed in
+    later versions.
+  info_ru: |
+    Максимальное число параллельных RDMA-операций получения данных. Следует
+    иметь в виду, что данное число буферов размером `rdma_max_msg` выделяется
+    для каждого подключённого клиентского соединения, так что данная настройка
+    влияет на потребление памяти. Это так потому, что RDMA-приём данных в
+    Vitastor, увы, всё равно не является zero-copy, т.е. всё равно 1 раз
+    копирует данные в памяти. Данная особенность, возможно, будет исправлена в
+    более новых версиях Vitastor.
+- name: peer_connect_interval
+  type: sec
+  min: 1
+  default: 5
+  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
+  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
+  info: |
+    OSD connection inactivity time after which clients and other OSDs send
+    keepalive requests to check state of the connection.
+  info_ru: |
+    Время неактивности соединения с OSD, после которого клиенты или другие OSD
+    посылают запрос проверки состояния соединения.
+- name: osd_ping_timeout
+  type: sec
+  min: 1
+  default: 5
+  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
+    is scheduled.
+  info_ru: |
+    Максимальное время ожидания ответа на запрос проверки состояния соединения.
+    Если OSD не отвечает за это время, соединение отключается и производится
+    повторная попытка соединения.
+- name: up_wait_retry_interval
+  type: ms
+  min: 50
+  default: 500
+  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
+    the time for the clients to wait before re-attempting such I/O requests.
+  info_ru: |
+    Когда OSD получают от клиентов запросы ввода-вывода, относящиеся к не
+    поднятым на данный момент на них PG, либо к PG в процессе синхронизации,
+    они отвечают клиентам специальным кодом ошибки, означающим, что клиент
+    должен некоторое время подождать перед повторением запроса. Именно это время
+    ожидания задаёт данный параметр.
+- name: max_etcd_attempts
+  type: int
+  default: 5
+  info: |
+    Maximum number of attempts for etcd requests which can't be retried
+    indefinitely.
+  info_ru: |
+    Максимальное число попыток выполнения запросов к etcd для тех запросов,
+    которые нельзя повторять бесконечно.
+- name: etcd_quick_timeout
+  type: ms
+  default: 1000
+  info: |
+    Timeout for etcd requests which should complete quickly, like lease refresh.
+  info_ru: |
+    Максимальное время выполнения запросов к etcd, которые должны завершаться
+    быстро, таких, как обновление резервации (lease).
+- name: etcd_slow_timeout
+  type: ms
+  default: 5000
+  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
+    гарантировать быстрое выполнение.
+- name: etcd_keepalive_timeout
+  type: sec
+  default: max(30, etcd_report_interval*2)
+  info: |
+    Timeout for etcd connection HTTP Keep-Alive. Should be higher than
+    etcd_report_interval to guarantee that keepalive actually works.
+  info_ru: |
+    Таймаут для HTTP Keep-Alive в соединениях к etcd. Должен быть больше, чем
+    etcd_report_interval, чтобы keepalive гарантированно работал.
+- name: etcd_ws_keepalive_timeout
+  type: sec
+  default: 30
+  info: |
+    etcd websocket ping interval required to keep the connection alive and
+    detect disconnections quickly.
+  info_ru: |
+    Интервал проверки живости вебсокет-подключений к etcd.

docs/params/osd.yml

@@ -0,0 +1,345 @@
+- name: etcd_report_interval
+  type: sec
+  default: 5
+  info: |
+    Interval at which OSDs report their state to etcd. Affects OSD lease time
+    and thus the failover speed. Lease time is equal to this parameter value
+    plus max_etcd_attempts * etcd_quick_timeout because it should be guaranteed
+    that every OSD always refreshes its lease in time.
+  info_ru: |
+    Интервал, с которым OSD обновляет своё состояние в etcd. Значение параметра
+    влияет на время резервации (lease) OSD и поэтому на скорость переключения
+    при падении OSD. Время lease равняется значению этого параметра плюс
+    max_etcd_attempts * etcd_quick_timeout.
+- name: run_primary
+  type: bool
+  default: true
+  info: |
+    Start primary OSD logic on this OSD. As of now, can be turned off only for
+    debugging purposes. It's possible to implement additional feature for the
+    monitor which may allow to separate primary and secondary OSDs, but it's
+    unclear why anyone could need it, so it's not implemented.
+  info_ru: |
+    Запускать логику первичного OSD на данном OSD. На данный момент отключать
+    эту опцию может иметь смысл только в целях отладки. В теории, можно
+    реализовать дополнительный режим для монитора, который позволит отделять
+    первичные OSD от вторичных, но пока не понятно, зачем это может кому-то
+    понадобиться, поэтому это не реализовано.
+- name: osd_network
+  type: string or array of strings
+  type_ru: строка или массив строк
+  info: |
+    Network mask of the network (IPv4 or IPv6) to use for OSDs. Note that
+    although it's possible to specify multiple networks here, this does not
+    mean that OSDs will create multiple listening sockets - they'll only
+    pick the first matching address of an UP + RUNNING interface. Separate
+    networks for cluster and client connections are also not implemented, but
+    they are mostly useless anyway, so it's not a big deal.
+  info_ru: |
+    Маска подсети (IPv4 или IPv6) для использования для соединений с OSD.
+    Имейте в виду, что хотя сейчас и можно передать в этот параметр несколько
+    подсетей, это не означает, что OSD будут создавать несколько слушающих
+    сокетов - они лишь будут выбирать адрес первого поднятого (состояние UP +
+    RUNNING), подходящий под заданную маску. Также не реализовано разделение
+    кластерной и публичной сетей OSD. Правда, от него обычно всё равно довольно
+    мало толку, так что особенной проблемы в этом нет.
+- name: bind_address
+  type: string
+  default: "0.0.0.0"
+  info: |
+    Instead of the network mask, you can also set OSD listen address explicitly
+    using this parameter. May be useful if you want to start OSDs on interfaces
+    that are not UP + RUNNING.
+  info_ru: |
+    Этим параметром можно явным образом задать адрес, на котором будет ожидать
+    соединений OSD (вместо использования маски подсети). Может быть полезно,
+    например, чтобы запускать OSD на неподнятых интерфейсах (не UP + RUNNING).
+- name: bind_port
+  type: int
+  info: |
+    By default, OSDs pick random ports to use for incoming connections
+    automatically. With this option you can set a specific port for a specific
+    OSD by hand.
+  info_ru: |
+    По умолчанию OSD сами выбирают случайные порты для входящих подключений.
+    С помощью данной опции вы можете задать порт для отдельного OSD вручную.
+- name: autosync_interval
+  type: sec
+  default: 5
+  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
+    them OSDs quickly fill their journals, become unable to clear them and
+    stall. Also this option limits the amount of recent uncommitted changes
+    which OSDs may lose in case of a power outage in case when clients don't
+    issue fsyncs at all.
+  info_ru: |
+    Временной интервал отправки автоматических fsync-ов (операций очистки кэша)
+    каждым OSD для случая, когда режим immediate_commit отключён. fsync-и нужны
+    OSD, чтобы успевать очищать журнал - без них OSD быстро заполняют журналы и
+    перестают обрабатывать операции записи. Также эта опция ограничивает объём
+    недавних незафиксированных изменений, которые OSD могут терять при
+    отключении питания, если клиенты вообще не отправляют fsync.
+- name: autosync_writes
+  type: int
+  default: 128
+  info: |
+    Same as autosync_interval, but sets the maximum number of uncommitted write
+    operations before issuing an fsync operation internally.
+  info_ru: |
+    Аналогично autosync_interval, но задаёт не временной интервал, а
+    максимальное количество незафиксированных операций записи перед
+    принудительной отправкой fsync-а.
+- name: recovery_queue_depth
+  type: int
+  default: 4
+  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
+    and rebalancing, but it's planned to implement more.
+  info_ru: |
+    Максимальное число операций восстановления на одном первичном OSD в любой
+    момент времени. На данный момент единственный параметр, который можно менять
+    для ускорения или замедления восстановления и перебалансировки данных, но
+    в планах реализация других параметров.
+- name: recovery_sync_batch
+  type: int
+  default: 16
+  info: Maximum number of recovery operations before issuing an additional fsync.
+  info_ru: Максимальное число операций восстановления перед дополнительным fsync.
+- name: readonly
+  type: bool
+  default: false
+  info: |
+    Read-only mode. If this is enabled, an OSD will never issue any writes to
+    the underlying device. This may be useful for recovery purposes.
+  info_ru: |
+    Режим "только чтение". Если включить этот режим, OSD не будет писать ничего
+    на диск. Может быть полезно в целях восстановления.
+- name: no_recovery
+  type: bool
+  default: false
+  info: |
+    Disable automatic background recovery of objects. Note that it doesn't
+    affect implicit recovery of objects happening during writes - a write is
+    always made to a full set of at least pg_minsize OSDs.
+  info_ru: |
+    Отключить автоматическое фоновое восстановление объектов. Обратите внимание,
+    что эта опция не отключает восстановление объектов, происходящее при
+    записи - запись всегда производится в полный набор из как минимум pg_minsize
+    OSD.
+- name: no_rebalance
+  type: bool
+  default: false
+  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.
+  info_ru: |
+    Отключить фоновое перемещение объектов между разными OSD. Отключение
+    означает, что PG, находящиеся в состоянии `has_misplaced`, будут оставлены
+    в нём на неопределённый срок.
+- name: print_stats_interval
+  type: sec
+  default: 3
+  info: |
+    Time interval at which OSDs print simple human-readable operation
+    statistics on stdout.
+  info_ru: |
+    Временной интервал, с которым OSD печатают простую человекочитаемую
+    статистику выполнения операций в стандартный вывод.
+- name: slow_log_interval
+  type: sec
+  default: 10
+  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
+    "slow".
+  info_ru: |
+    Временной интервал, с которым OSD выводят в стандартный вывод список
+    медленных или зависших операций, если таковые имеются. Также время, при
+    превышении которого операция считается "медленной".
+- name: max_write_iodepth
+  type: int
+  default: 128
+  info: |
+    Parallel client write operation limit per one OSD. Operations that exceed
+    this limit are pushed to a temporary queue instead of being executed
+    immediately.
+  info_ru: |
+    Максимальное число одновременных клиентских операций записи на один OSD.
+    Операции, превышающие этот лимит, не исполняются сразу, а сохраняются во
+    временной очереди.
+- name: min_flusher_count
+  type: int
+  default: 1
+  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.
+    Minimum number is set by this parameter.
+  info_ru: |
+    Flusher - это микро-поток (корутина), которая копирует данные из журнала в
+    основную область устройства данных. Их число настраивается динамически между
+    минимальным и максимальным значением. Этот параметр задаёт минимальное число.
+- name: max_flusher_count
+  type: int
+  default: 256
+  info: |
+    Maximum number of journal flushers (see above min_flusher_count).
+  info_ru: |
+    Максимальное число микро-потоков очистки журнала (см. выше min_flusher_count).
+- name: inmemory_metadata
+  type: bool
+  default: true
+  info: |
+    This parameter makes Vitastor always keep metadata area of the block device
+    in memory. It's required for good performance because it allows to avoid
+    additional read-modify-write cycles during metadata modifications. Metadata
+    area size is currently roughly 224 MB per 1 TB of data. You can turn it off
+    to reduce memory usage by this value, but it will hurt performance. This
+    restriction is likely to be removed in the future along with the upgrade
+    of the metadata storage scheme.
+  info_ru: |
+    Данный параметр заставляет Vitastor всегда держать область метаданных диска
+    в памяти. Это нужно, чтобы избегать дополнительных операций чтения с диска
+    при записи. Размер области метаданных на данный момент составляет примерно
+    224 МБ на 1 ТБ данных. При включении потребление памяти снизится примерно
+    на эту величину, но при этом также снизится и производительность. В будущем,
+    после обновления схемы хранения метаданных, это ограничение, скорее всего,
+    будет ликвидировано.
+- name: inmemory_journal
+  type: bool
+  default: true
+  info: |
+    This parameter make Vitastor always keep journal area of the block
+    device in memory. Turning it off will, again, reduce memory usage, but
+    hurt performance because flusher coroutines will have to read data from
+    the disk back before copying it into the main area. The memory usage benefit
+    is typically very small because it's sufficient to have 16-32 MB journal
+    for SSD OSDs. However, in theory it's possible that you'll want to turn it
+    off for hybrid (HDD+SSD) OSDs with large journals on quick devices.
+  info_ru: |
+    Данный параметр заставляет Vitastor всегда держать в памяти журналы OSD.
+    Отключение параметра, опять же, снижает потребление памяти, но ухудшает
+    производительность, так как для копирования данных из журнала в основную
+    область устройства OSD будут вынуждены читать их обратно с диска. Выигрыш
+    по памяти при этом обычно крайне низкий, так как для SSD OSD обычно
+    достаточно 16- или 32-мегабайтного журнала. Однако в теории отключение
+    параметра может оказаться полезным для гибридных OSD (HDD+SSD) с большими
+    журналами, расположенными на быстром по сравнению с HDD устройстве.
+- name: journal_sector_buffer_count
+  type: int
+  default: 32
+  info: |
+    Maximum number of buffers that can be used for writing journal metadata
+    blocks. The only situation when you should increase it to a larger value
+    is when you enable journal_no_same_sector_overwrites. In this case set
+    it to, for example, 1024.
+  info_ru: |
+    Максимальное число буферов, разрешённых для использования под записываемые
+    в журнал блоки метаданных. Единственная ситуация, в которой этот параметр
+    нужно менять - это если вы включаете journal_no_same_sector_overwrites. В
+    этом случае установите данный параметр, например, в 1024.
+- name: journal_no_same_sector_overwrites
+  type: bool
+  default: false
+  info: |
+    Enable this option for SSDs like Intel D3-S4510 and D3-S4610 which REALLY
+    don't like when a program overwrites the same sector multiple times in a
+    row and slow down significantly (from 25000+ iops to ~3000 iops). When
+    this option is set, Vitastor will always move to the next sector of the
+    journal after writing it instead of possibly overwriting it the second time.
+
+    Most (99%) other SSDs don't need this option.
+  info_ru: |
+    Включайте данную опцию для SSD вроде Intel D3-S4510 и D3-S4610, которые
+    ОЧЕНЬ не любят, когда ПО перезаписывает один и тот же сектор несколько раз
+    подряд. Такие SSD при многократной перезаписи одного и того же сектора
+    сильно замедляются - условно, с 25000 и более iops до 3000 iops. Когда
+    данная опция установлена, Vitastor всегда переходит к следующему сектору
+    журнала после записи вместо потенциально повторной перезаписи того же
+    самого сектора.
+
+    Почти все другие SSD (99% моделей) не требуют данной опции.
+- name: throttle_small_writes
+  type: bool
+  default: false
+  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
+    small writes complete very quickly because they're first written to the
+    journal device, but moving them to the main device is slow. So if an OSD
+    allows clients to issue a lot of small writes it will perform very good
+    for several seconds and then the journal will fill up and the performance
+    will drop to almost zero. Throttling is meant to prevent this problem by
+    artifically slowing quick writes down based on the amount of free space in
+    the journal. When throttling is used, the performance of small writes will
+    decrease smoothly instead of abrupt drop at the moment when the journal
+    fills up.
+  info_ru: |
+    Разрешить мягкое ограничение скорости журналируемой записи. Полезно для
+    гибридных OSD с быстрыми устройствами метаданных и медленными устройствами
+    данных. Идея заключается в том, что мелкие записи в этой ситуации могут
+    завершаться очень быстро, так как они изначально записываются на быстрое
+    журнальное устройство (SSD). Но перемещать их потом на основное медленное
+    устройство долго. Поэтому если OSD быстро примет от клиентов очень много
+    мелких операций записи, он быстро заполнит свой журнал, после чего
+    производительность записи резко упадёт практически до нуля. Ограничение
+    скорости записи призвано решить эту проблему с помощью искусственного
+    замедления операций записи на основании объёма свободного места в журнале.
+    Когда эта опция включена, производительность мелких операций записи будет
+    снижаться плавно, а не резко в момент окончательного заполнения журнала.
+- name: throttle_target_iops
+  type: int
+  default: 100
+  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
+    (HDDs).
+  info_ru: |
+    Расчётное максимальное число ограничиваемых операций в секунду при условии
+    отсутствия свободного места в журнале. Устанавливайте приблизительно равным
+    максимальной производительности случайной записи ваших устройств данных
+    (HDD) в операциях в секунду.
+- name: throttle_target_mbs
+  type: int
+  default: 100
+  info: |
+    Target maximum bandwidth in MB/s of throttled operations per second under
+    the condition of full journal. Set it to approximate linear write
+    performance of your data devices (HDDs).
+  info_ru: |
+    Расчётный максимальный размер в МБ/с ограничиваемых операций в секунду при
+    условии отсутствия свободного места в журнале. Устанавливайте приблизительно
+    равным максимальной производительности линейной записи ваших устройств
+    данных (HDD).
+- name: throttle_target_parallelism
+  type: int
+  default: 1
+  info: |
+    Target maximum parallelism of throttled operations under the condition of
+    full journal. Set it to approximate internal parallelism of your data
+    devices (1 for HDDs, 4-8 for SSDs).
+  info_ru: |
+    Расчётный максимальный параллелизм ограничиваемых операций в секунду при
+    условии отсутствия свободного места в журнале. Устанавливайте приблизительно
+    равным внутреннему параллелизму ваших устройств данных (1 для HDD, 4-8
+    для SSD).
+- name: throttle_threshold_us
+  type: us
+  default: 50
+  info: |
+    Minimal computed delay to be applied to throttled operations. Usually
+    doesn't need to be changed.
+  info_ru: |
+    Минимальная применимая к ограничиваемым операциям задержка. Обычно не
+    требует изменений.
+- name: osd_memlock
+  type: bool
+  default: false
+  info: >
+    Lock all OSD memory to prevent it from being unloaded into swap with
+    mlockall(). Requires sufficient ulimit -l (max locked memory).
+  info_ru: >
+    Блокировать всю память OSD с помощью mlockall, чтобы запретить её выгрузку
+    в пространство подкачки. Требует достаточного значения ulimit -l (лимита
+    заблокированной памяти).

mon/lp-optimizer.js

@@ -50,7 +50,7 @@ async function lp_solve(text)
     return { score, vars };
 }
 
-async function optimize_initial({ osd_tree, pg_count, pg_size = 3, pg_minsize = 2, max_combinations = 10000, parity_space = 1, round_robin = 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)
     {
@@ -92,7 +92,7 @@ async function optimize_initial({ osd_tree, pg_count, pg_size = 3, pg_minsize =
         console.log(lp);
         throw new Error('Problem is infeasible or unbounded - is it a bug?');
     }
-    const int_pgs = make_int_pgs(lp_result.vars, pg_count, round_robin);
+    const int_pgs = make_int_pgs(lp_result.vars, pg_count, ordered);
     const eff = pg_list_space_efficiency(int_pgs, all_weights, pg_minsize, parity_space);
     const res = {
         score: lp_result.score,
@@ -140,20 +140,20 @@ function make_int_pgs(weights, pg_count, round_robin)
     return int_pgs;
 }
 
-function calc_intersect_weights(pg_size, pg_count, prev_weights, all_pgs)
+function calc_intersect_weights(old_pg_size, pg_size, pg_count, prev_weights, all_pgs, ordered)
 {
     const move_weights = {};
-    if ((1 << pg_size) < pg_count)
+    if ((1 << old_pg_size) < pg_count)
     {
         const intersect = {};
         for (const pg_name in prev_weights)
         {
             const pg = pg_name.substr(3).split(/_/);
-            for (let omit = 1; omit < (1 << pg_size); omit++)
+            for (let omit = 1; omit < (1 << old_pg_size); omit++)
             {
                 let pg_omit = [ ...pg ];
-                let intersect_count = pg_size;
+                let intersect_count = old_pg_size;
-                for (let i = 0; i < pg_size; i++)
+                for (let i = 0; i < old_pg_size; i++)
                 {
                     if (omit & (1 << i))
                     {
@@ -161,6 +161,8 @@ function calc_intersect_weights(pg_size, pg_count, prev_weights, all_pgs)
                         intersect_count--;
                     }
                 }
+                if (!ordered)
+                    pg_omit = pg_omit.filter(n => n).sort();
                 pg_omit = pg_omit.join(':');
                 intersect[pg_omit] = Math.max(intersect[pg_omit] || 0, intersect_count);
             }
@@ -174,10 +176,10 @@ function calc_intersect_weights(pg_size, pg_count, prev_weights, all_pgs)
                 for (let i = 0; i < pg_size; i++)
                 {
                     if (omit & (1 << i))
-                    {
                         pg_omit[i] = '';
-                    }
                 }
+                if (!ordered)
+                    pg_omit = pg_omit.filter(n => n).sort();
                 pg_omit = pg_omit.join(':');
                 max_int = Math.max(max_int, intersect[pg_omit] || 0);
             }
@@ -186,15 +188,18 @@ function calc_intersect_weights(pg_size, pg_count, prev_weights, all_pgs)
     }
     else
     {
-        const prev_pg_hashed = Object.keys(prev_weights).map(pg_name => pg_name.substr(3).split(/_/).reduce((a, c) => { a[c] = 1; return a; }, {}));
+        const prev_pg_hashed = Object.keys(prev_weights).map(pg_name => pg_name
+            .substr(3).split(/_/).reduce((a, c, i) => { a[c] = i+1; return a; }, {}));
         for (const pg of all_pgs)
         {
             if (!prev_weights['pg_'+pg.join('_')])
             {
                 let max_int = 0;
-                for (const prev_hash in prev_pg_hashed)
+                for (const prev_hash of prev_pg_hashed)
                 {
-                    const intersect_count = pg.reduce((a, osd) => a + (prev_hash[osd] ? 1 : 0), 0);
+                    const intersect_count = ordered
+                        ? pg.reduce((a, osd, i) => a + (prev_hash[osd] == 1+i ? 1 : 0), 0)
+                        : pg.reduce((a, osd, i) => a + (prev_hash[osd] ? 1 : 0), 0);
                     if (max_int < intersect_count)
                     {
                         max_int = intersect_count;
@@ -243,7 +248,7 @@ function add_valid_previous(osd_tree, prev_weights, all_pgs)
 }
 
 // Try to minimize data movement
-async function optimize_change({ prev_pgs: prev_int_pgs, osd_tree, pg_size = 3, pg_minsize = 2, max_combinations = 10000, parity_space = 1 })
+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)
     {
@@ -266,9 +271,13 @@ async function optimize_change({ prev_pgs: prev_int_pgs, osd_tree, pg_size = 3,
             prev_pg_per_osd[osd].push([ pg_name, (i >= pg_minsize ? parity_space : 1) ]);
         }
     }
+    const old_pg_size = prev_int_pgs[0].length;
     // Get all combinations
     let all_pgs = random_combinations(osd_tree, pg_size, max_combinations, parity_space > 1);
-    add_valid_previous(osd_tree, prev_weights, all_pgs);
+    if (old_pg_size == pg_size)
+    {
+        add_valid_previous(osd_tree, prev_weights, all_pgs);
+    }
     all_pgs = Object.values(all_pgs);
     const pg_per_osd = {};
     for (const pg of all_pgs)
@@ -282,7 +291,7 @@ async function optimize_change({ prev_pgs: prev_int_pgs, osd_tree, pg_size = 3,
         }
     }
     // Penalize PGs based on their similarity to old PGs
-    const move_weights = calc_intersect_weights(pg_size, pg_count, prev_weights, all_pgs);
+    const move_weights = calc_intersect_weights(old_pg_size, pg_size, pg_count, prev_weights, all_pgs, ordered);
     // Calculate total weight - old PG weights
     const all_pg_names = all_pgs.map(pg => 'pg_'+pg.join('_'));
     const all_pgs_hash = all_pg_names.reduce((a, c) => { a[c] = true; return a; }, {});
@@ -373,11 +382,35 @@ async function optimize_change({ prev_pgs: prev_int_pgs, osd_tree, pg_size = 3,
         {
             differs++;
         }
-        for (let j = 0; j < pg_size; j++)
+    }
+    if (ordered)
+    {
+        for (let i = 0; i < pg_count; i++)
         {
-            if (new_pgs[i][j] != prev_int_pgs[i][j])
+            for (let j = 0; j < pg_size; j++)
             {
-                osd_differs++;
+                if (new_pgs[i][j] != prev_int_pgs[i][j])
+                {
+                    osd_differs++;
+                }
+            }
+        }
+    }
+    else
+    {
+        for (let i = 0; i < pg_count; i++)
+        {
+            const old_map = prev_int_pgs[i].reduce((a, c) => { a[c] = (a[c]|0) + 1; return a; }, {});
+            for (let j = 0; j < pg_size; j++)
+            {
+                if ((0|old_map[new_pgs[i][j]]) > 0)
+                {
+                    old_map[new_pgs[i][j]]--;
+                }
+                else
+                {
+                    osd_differs++;
+                }
             }
         }
     }

mon/make-osd-hybrid.js

@@ -0,0 +1,414 @@
+#!/usr/bin/nodejs
+// systemd unit generator for hybrid (HDD+SSD) vitastor OSDs
+// Copyright (c) Vitaliy Filippov, 2019+
+// License: VNPL-1.1
+
+// USAGE: nodejs make-osd-hybrid.js [--disable_ssd_cache 0] [--disable_hdd_cache 0] /dev/sda /dev/sdb /dev/sdc /dev/sdd ...
+// I.e. - just pass all HDDs and SSDs mixed, the script will decide where
+// to put journals on its own
+
+const fs = require('fs');
+const fsp = fs.promises;
+const child_process = require('child_process');
+
+const options = {
+    debug: 1,
+    journal_size: 1024*1024*1024,
+    min_meta_size: 1024*1024*1024,
+    object_size: 1024*1024,
+    bitmap_granularity: 4096,
+    device_block_size: 4096,
+    disable_ssd_cache: 1,
+    disable_hdd_cache: 1,
+};
+
+run().catch(console.fatal);
+
+async function run()
+{
+    const device_list = parse_options();
+    await system_or_die("mkdir -p /var/log/vitastor; chown vitastor /var/log/vitastor");
+    // Collect devices
+    const all_devices = await collect_devices(device_list);
+    const ssds = all_devices.filter(d => d.ssd);
+    const hdds = all_devices.filter(d => !d.ssd);
+    // Collect existing OSD units
+    const osd_units = await collect_osd_units();
+    // Count assigned HDD journals and unallocated space for each SSD
+    await check_journal_count(ssds, osd_units);
+    // Create new OSDs
+    await create_new_hybrid_osds(hdds, ssds, osd_units);
+    process.exit(0);
+}
+
+function parse_options()
+{
+    const devices = [];
+    const opt = {};
+    for (let i = 2; i < process.argv.length; i++)
+    {
+        const arg = process.argv[i];
+        if (arg == '--help' || arg == '-h')
+        {
+            opt.help = true;
+            break;
+        }
+        else if (arg.substr(0, 2) == '--')
+            opt[arg.substr(2)] = process.argv[++i];
+        else
+            devices.push(arg);
+    }
+    if (opt.help || !devices.length)
+    {
+        console.log(
+            'Prepare hybrid (HDD+SSD) Vitastor OSDs\n'+
+            '(c) Vitaliy Filippov, 2019+, license: VNPL-1.1\n\n'+
+            'USAGE: nodejs make-osd-hybrid.js [OPTIONS] /dev/sda /dev/sdb /dev/sdc ...\n'+
+            'Just pass all your SSDs and HDDs in any order, the script will distribute OSDs for you.\n\n'+
+            'OPTIONS (with defaults):\n'+
+            Object.keys(options).map(k => `  --${k} ${options[k]}`).join('\n')
+        );
+        process.exit(0);
+    }
+    for (const k in opt)
+        options[k] = opt[k];
+    return devices;
+}
+
+// Collect devices
+async function collect_devices(devices_to_check)
+{
+    const devices = [];
+    for (const dev of devices_to_check)
+    {
+        if (dev.substr(0, 5) != '/dev/')
+        {
+            console.log(`${dev} does not start with /dev/, skipping`);
+            continue;
+        }
+        if (!await file_exists('/sys/block/'+dev.substr(5)))
+        {
+            console.log(`${dev} is a partition, skipping`);
+            continue;
+        }
+        // Check if the device is an SSD
+        const rot = '/sys/block/'+dev.substr(5)+'/queue/rotational';
+        if (!await file_exists(rot))
+        {
+            console.log(`${dev} does not have ${rot} to check whether it's an SSD, skipping`);
+            continue;
+        }
+        const ssd = !parseInt(await fsp.readFile(rot, { encoding: 'utf-8' }));
+        // Check if the device has partition table
+        let [ has_partition_table, parts ] = await system(`sfdisk --dump ${dev} --json`);
+        if (has_partition_table != 0)
+        {
+            // Check if the device has any data
+            const [ has_data, out ] = await system(`blkid ${dev}`);
+            if (has_data == 0)
+            {
+                console.log(`${dev} contains data, skipping:\n  ${out.trim().replace(/\n/g, '\n  ')}`);
+                continue;
+            }
+        }
+        parts = parts ? JSON.parse(parts).partitiontable : null;
+        if (parts && parts.label != 'gpt')
+        {
+            console.log(`${dev} contains "${parts.label}" partition table, only GPT is supported, skipping`);
+            continue;
+        }
+        devices.push({
+            path: dev,
+            ssd,
+            parts,
+        });
+    }
+    return devices;
+}
+
+// Collect existing OSD units
+async function collect_osd_units()
+{
+    const units = [];
+    for (const unit of (await system("ls /etc/systemd/system/vitastor-osd*.service"))[1].trim().split('\n'))
+    {
+        if (!unit)
+        {
+            continue;
+        }
+        let cmd = /^ExecStart\s*=\s*(([^\n]*\\\n)*[^\n]*)/.exec(await fsp.readFile(unit, { encoding: 'utf-8' }));
+        if (!cmd)
+        {
+            console.log('ExecStart= not found in '+unit+', skipping')
+            continue;
+        }
+        let kv = {}, key;
+        cmd = cmd[1].replace(/^bash\s+-c\s+'/, '')
+            .replace(/>>\s*\S+2>\s*&1\s*'$/, '')
+            .replace(/\s*\\\n\s*/g, ' ')
+            .replace(/([^\s']+)|'([^']+)'/g, (m, m1, m2) =>
+            {
+                m1 = m1||m2;
+                if (key == null)
+                {
+                    if (m1.substr(0, 2) != '--')
+                    {
+                        console.log('Strange command line in '+unit+', stopping');
+                        process.exit(1);
+                    }
+                    key = m1.substr(2);
+                }
+                else
+                {
+                    kv[key] = m1;
+                    key = null;
+                }
+            });
+        units.push(kv);
+    }
+    return units;
+}
+
+// Count assigned HDD journals and unallocated space for each SSD
+async function check_journal_count(ssds, osd_units)
+{
+    const units_by_journal = osd_units.reduce((a, c) =>
+    {
+        if (c.journal_device)
+            a[c.journal_device] = c;
+        return a;
+    }, {});
+    for (const dev of ssds)
+    {
+        dev.journals = 0;
+        if (dev.parts)
+        {
+            for (const part of dev.parts.partitions)
+            {
+                if (part.uuid && units_by_journal['/dev/disk/by-partuuid/'+part.uuid.toLowerCase()])
+                {
+                    dev.journals++;
+                }
+            }
+            dev.free = free_from_parttable(dev.parts);
+        }
+        else
+        {
+            dev.free = parseInt(await system_or_die("blockdev --getsize64 "+dev.path));
+        }
+    }
+}
+
+async function create_new_hybrid_osds(hdds, ssds, osd_units)
+{
+    const units_by_disk = osd_units.reduce((a, c) => { a[c.data_device] = c; return a; }, {});
+    for (const dev of hdds)
+    {
+        if (!dev.parts)
+        {
+            // HDD is not partitioned yet, create a single partition
+            // + is the "default value" for sfdisk
+            await system_or_die('sfdisk '+dev.path, 'label: gpt\n\n+ +\n');
+            dev.parts = JSON.parse(await system_or_die('sfdisk --dump '+dev.path+' --json')).partitiontable;
+        }
+        if (dev.parts.partitions.length != 1)
+        {
+            console.log(dev.path+' has more than 1 partition, skipping');
+        }
+        else if ((dev.parts.partitions[0].start + dev.parts.partitions[0].size) != (1 + dev.parts.lastlba))
+        {
+            console.log(dev.path+'1 is not a whole-disk partition, skipping');
+        }
+        else if (!dev.parts.partitions[0].uuid)
+        {
+            console.log(dev.parts.partitions[0].node+' does not have UUID. Please repartition '+dev.path+' with GPT');
+        }
+        else if (!units_by_disk['/dev/disk/by-partuuid/'+dev.parts.partitions[0].uuid.toLowerCase()])
+        {
+            await create_hybrid_osd(dev, ssds);
+        }
+    }
+}
+
+async function create_hybrid_osd(dev, ssds)
+{
+    // Create a new OSD
+    // Calculate metadata size
+    const data_device = '/dev/disk/by-partuuid/'+dev.parts.partitions[0].uuid.toLowerCase();
+    const data_size = dev.parts.partitions[0].size * dev.parts.sectorsize;
+    const meta_entry_size = 24 + 2*options.object_size/options.bitmap_granularity/8;
+    const entries_per_block = Math.floor(options.device_block_size / meta_entry_size);
+    const object_count = Math.floor(data_size / options.object_size);
+    let meta_size = Math.ceil(1 + object_count / entries_per_block) * options.device_block_size;
+    // Leave some extra space for future metadata formats and round metadata area size to multiples of 1 MB
+    meta_size = 2*meta_size;
+    meta_size = Math.ceil(meta_size/1024/1024) * 1024*1024;
+    if (meta_size < options.min_meta_size)
+        meta_size = options.min_meta_size;
+    let journal_size = Math.ceil(options.journal_size/1024/1024) * 1024*1024;
+    // Pick an SSD for journal, balancing the number of journals across SSDs
+    let selected_ssd;
+    for (const ssd of ssds)
+        if (ssd.free >= (meta_size+journal_size) && (!selected_ssd || selected_ssd.journals > ssd.journals))
+            selected_ssd = ssd;
+    if (!selected_ssd)
+    {
+        console.error('Could not find free space for SSD journal and metadata for '+dev.path);
+        process.exit(1);
+    }
+    // Allocate an OSD number
+    const osd_num = (await system_or_die("vitastor-cli alloc-osd")).trim();
+    if (!osd_num)
+    {
+        console.error('Failed to run vitastor-cli alloc-osd');
+        process.exit(1);
+    }
+    console.log('Creating OSD '+osd_num+' on '+dev.path+' (HDD) with journal and metadata on '+selected_ssd.path+' (SSD)');
+    // Add two partitions: journal and metadata
+    const new_parts = await add_partitions(selected_ssd, [ journal_size, meta_size ]);
+    selected_ssd.journals++;
+    const journal_device = '/dev/disk/by-partuuid/'+new_parts[0].uuid.toLowerCase();
+    const meta_device = '/dev/disk/by-partuuid/'+new_parts[1].uuid.toLowerCase();
+    // Wait until the device symlinks appear
+    while (!await file_exists(journal_device))
+    {
+        await new Promise(ok => setTimeout(ok, 100));
+    }
+    while (!await file_exists(meta_device))
+    {
+        await new Promise(ok => setTimeout(ok, 100));
+    }
+    // Zero out metadata and journal
+    await system_or_die("dd if=/dev/zero of="+journal_device+" bs=1M count="+(journal_size/1024/1024)+" oflag=direct");
+    await system_or_die("dd if=/dev/zero of="+meta_device+" bs=1M count="+(meta_size/1024/1024)+" oflag=direct");
+    // Create unit file for the OSD
+    const has_scsi_cache_type = options.disable_ssd_cache &&
+        (await system("ls /sys/block/"+selected_ssd.path.substr(5)+"/device/scsi_disk/*/cache_type"))[0] == 0;
+    const write_through = options.disable_ssd_cache && (
+        has_scsi_cache_type || selected_ssd.path.substr(5, 4) == 'nvme'
+        && (await system_or_die("/sys/block/"+selected_ssd.path.substr(5)+"/queue/write_cache")).trim() == "write through");
+    await fsp.writeFile('/etc/systemd/system/vitastor-osd'+osd_num+'.service',
+`[Unit]
+Description=Vitastor object storage daemon osd.${osd_num}
+After=network-online.target local-fs.target time-sync.target
+Wants=network-online.target local-fs.target time-sync.target
+PartOf=vitastor.target
+
+[Service]
+LimitNOFILE=1048576
+LimitNPROC=1048576
+LimitMEMLOCK=infinity
+ExecStart=bash -c '/usr/bin/vitastor-osd \\
+    --osd_num ${osd_num} ${write_through
+        ? "--disable_meta_fsync 1 --disable_journal_fsync 1 --immediate_commit "+(options.disable_hdd_cache ? "all" : "small")
+        : ""} \\
+    --throttle_small_writes 1 \\
+    --disk_alignment ${options.device_block_size} \\
+    --journal_block_size ${options.device_block_size} \\
+    --meta_block_size ${options.device_block_size} \\
+    --journal_no_same_sector_overwrites true \\
+    --journal_sector_buffer_count 1024 \\
+    --block_size ${options.object_size} \\
+    --data_device ${data_device} \\
+    --journal_device ${journal_device} \\
+    --meta_device ${meta_device} >>/var/log/vitastor/osd${osd_num}.log 2>&1'
+WorkingDirectory=/
+ExecStartPre=+chown vitastor:vitastor ${data_device}
+ExecStartPre=+chown vitastor:vitastor ${journal_device}
+ExecStartPre=+chown vitastor:vitastor ${meta_device}${
+    has_scsi_cache_type
+    ? "\nExecStartPre=+bash -c 'D=$$$(readlink "+journal_device+"); echo write through > $$$(dirname /sys/block/*/$$\${D##*/})/device/scsi_disk/*/cache_type'"
+    : ""}${
+    options.disable_hdd_cache
+    ? "\nExecStartPre=+bash -c 'D=$$$(readlink "+data_device+"); echo write through > $$$(dirname /sys/block/*/$$\${D##*/})/device/scsi_disk/*/cache_type'"
+    : ""}
+User=vitastor
+PrivateTmp=false
+TasksMax=infinity
+Restart=always
+StartLimitInterval=0
+RestartSec=10
+
+[Install]
+WantedBy=vitastor.target
+`);
+    await system_or_die("systemctl enable vitastor-osd"+osd_num);
+}
+
+async function add_partitions(dev, sizes)
+{
+    let script = 'label: gpt\n\n';
+    if (dev.parts)
+    {
+        // Old partitions
+        for (const part of dev.parts.partitions)
+        {
+            script += part.node+': '+Object.keys(part).map(k => k == 'node' ? '' : k+'='+part[k]).filter(k => k).join(', ')+'\n';
+        }
+    }
+    // New partitions
+    for (const size of sizes)
+    {
+        script += '+ '+Math.ceil(size/1024)+'KiB\n';
+    }
+    await system_or_die('sfdisk '+dev.path, script);
+    // Get new partition table and find the new partition
+    const newpt = JSON.parse(await system_or_die('sfdisk --dump '+dev.path+' --json')).partitiontable;
+    const old_nodes = dev.parts ? dev.parts.partitions.reduce((a, c) => { a[c.uuid] = true; return a; }, {}) : {};
+    const new_nodes = newpt.partitions.filter(part => !old_nodes[part.uuid]);
+    if (new_nodes.length != sizes.length)
+    {
+        console.error('Failed to partition '+dev.path+': new partitions not found in table');
+        process.exit(1);
+    }
+    dev.parts = newpt;
+    dev.free = free_from_parttable(newpt);
+    return new_nodes;
+}
+
+function free_from_parttable(pt)
+{
+    let free = pt.lastlba + 1 - pt.firstlba;
+    for (const part of pt.partitions)
+    {
+        free -= part.size;
+    }
+    free *= pt.sectorsize;
+    return free;
+}
+
+async function system_or_die(cmd, input = '')
+{
+    let [ exitcode, stdout, stderr ] = await system(cmd, input);
+    if (exitcode != 0)
+    {
+        console.error(cmd+' failed: '+stderr);
+        process.exit(1);
+    }
+    return stdout;
+}
+
+async function system(cmd, input = '')
+{
+    if (options.debug)
+    {
+        process.stderr.write('+ '+cmd+(input ? " <<EOF\n"+input.replace(/\s*$/, '\n')+"EOF" : '')+'\n');
+    }
+    const cp = child_process.spawn(cmd, { shell: true });
+    let stdout = '', stderr = '', finish_cb;
+    cp.stdout.on('data', buf => stdout += buf.toString());
+    cp.stderr.on('data', buf => stderr += buf.toString());
+    cp.on('exit', () => finish_cb && finish_cb());
+    cp.stdin.write(input);
+    cp.stdin.end();
+    if (cp.exitCode == null)
+    {
+        await new Promise(ok => finish_cb = ok);
+    }
+    return [ cp.exitCode, stdout, stderr ];
+}
+
+async function file_exists(filename)
+{
+    return new Promise((ok, no) => fs.access(filename, fs.constants.R_OK, err => ok(!err)));
+}

mon/make-osd.sh

@@ -25,6 +25,10 @@ OPT=$(vitastor-cli simple-offsets --format options $DEV | tr '\n' ' ')
 META=$(vitastor-cli simple-offsets --format json $DEV | jq .data_offset)
 dd if=/dev/zero of=$DEV bs=1048576 count=$(((META+1048575)/1048576)) oflag=direct
 
+mkdir -p /var/log/vitastor
+id vitastor &>/dev/null || useradd vitastor
+chown vitastor /var/log/vitastor
+
 cat >/etc/systemd/system/vitastor-osd$OSD_NUM.service <<EOF
 [Unit]
 Description=Vitastor object storage daemon osd.$OSD_NUM
@@ -36,14 +40,14 @@ PartOf=vitastor.target
 LimitNOFILE=1048576
 LimitNPROC=1048576
 LimitMEMLOCK=infinity
-ExecStart=/usr/bin/vitastor-osd \\
+ExecStart=bash -c '/usr/bin/vitastor-osd \\
     --osd_num $OSD_NUM \\
     --disable_data_fsync 1 \\
     --immediate_commit all \\
     --disk_alignment 4096 --journal_block_size 4096 --meta_block_size 4096 \\
     --journal_no_same_sector_overwrites true \\
     --journal_sector_buffer_count 1024 \\
-    $OPT
+    $OPT >>/var/log/vitastor/osd$OSD_NUM.log 2>&1'
 WorkingDirectory=/
 ExecStartPre=+chown vitastor:vitastor $DEV
 User=vitastor

mon/mon.js

@@ -31,6 +31,7 @@ const etcd_allow = new RegExp('^'+[
     'osd/inodestats/[1-9]\\d*',
     'osd/space/[1-9]\\d*',
     'mon/master',
+    'mon/member/[a-f0-9]+',
     'pg/state/[1-9]\\d*/[1-9]\\d*',
     'pg/stats/[1-9]\\d*/[1-9]\\d*',
     'pg/history/[1-9]\\d*/[1-9]\\d*',
@@ -83,8 +84,13 @@ const etcd_tree = {
             osd_idle_timeout: 5, // seconds. min: 1
             osd_ping_timeout: 5, // seconds. min: 1
             up_wait_retry_interval: 500, // ms. min: 50
+            max_etcd_attempts: 5,
+            etcd_quick_timeout: 1000, // ms
+            etcd_slow_timeout: 5000, // ms
+            etcd_keepalive_timeout: 30, // seconds, default is max(30, etcd_report_interval*2)
+            etcd_ws_keepalive_interval: 30, // seconds
             // osd
-            etcd_report_interval: 5,
+            etcd_report_interval: 5, // seconds
             run_primary: true,
             osd_network: null, // "192.168.7.0/24" or an array of masks
             bind_address: "0.0.0.0",
@@ -99,6 +105,7 @@ const etcd_tree = {
             no_rebalance: false,
             print_stats_interval: 3,
             slow_log_interval: 10,
+            osd_memlock: false,
             // blockstore - fixed in superblock
             block_size,
             disk_alignment,
@@ -125,6 +132,11 @@ const etcd_tree = {
             inmemory_journal,
             journal_sector_buffer_count,
             journal_no_same_sector_overwrites,
+            throttle_small_writes: false,
+            throttle_target_iops: 100,
+            throttle_target_mbs: 100,
+            throttle_target_parallelism: 1,
+            throttle_threshold_us: 50,
         }, */
         global: {},
         /* node_placement: {
@@ -148,6 +160,8 @@ const etcd_tree = {
                 root_node?: 'rack1',
                 // restrict pool to OSDs having all of these tags
                 osd_tags?: 'nvme' | [ 'nvme', ... ],
+                // prefer to put primary on OSD with these tags
+                primary_affinity_tags?: 'nvme' | [ 'nvme', ... ],
             },
             ...
         }, */
@@ -212,21 +226,28 @@ const etcd_tree = {
             }, */
         },
         inodestats: {
-            /* <inode_t>: {
+            /* <pool_id>: {
-                read: { count: uint64_t, usec: uint64_t, bytes: uint64_t },
+                <inode_t>: {
-                write: { count: uint64_t, usec: uint64_t, bytes: uint64_t },
+                    read: { count: uint64_t, usec: uint64_t, bytes: uint64_t },
-                delete: { count: uint64_t, usec: uint64_t, bytes: uint64_t },
+                    write: { count: uint64_t, usec: uint64_t, bytes: uint64_t },
+                    delete: { count: uint64_t, usec: uint64_t, bytes: uint64_t },
+                },
             }, */
         },
         space: {
             /* <osd_num_t>: {
-                <inode_t>: uint64_t, // bytes
+                <pool_id>: {
+                    <inode_t>: uint64_t, // bytes
+                },
             }, */
         },
     },
     mon: {
         master: {
-            /* ip: [ string ], */
+            /* ip: [ string ], id: uint64_t */
+        },
+        standby: {
+            /* <uint64_t>: { ip: [ string ] }, */
         },
     },
     pg: {
@@ -257,7 +278,7 @@ const etcd_tree = {
                 <pg_id>: {
                     osd_sets: osd_num_t[][],
                     all_peers: osd_num_t[],
-                    epoch: uint32_t,
+                    epoch: uint64_t,
                 },
             }, */
         },
@@ -341,6 +362,9 @@ 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.ws = null;
+        this.ws_alive = false;
+        this.ws_keepalive_timer = null;
         this.on_stop_cb = () => this.on_stop(0).catch(console.error);
     }
 
@@ -383,7 +407,7 @@ class Mon
         for (const pool_id in this.state.config.pools)
         {
             if (!this.state.pool.stats[pool_id] ||
-                !this.state.pool.stats[pool_id].pg_real_size)
+                !Number(this.state.pool.stats[pool_id].pg_real_size))
             {
                 // Generate missing data in etcd
                 this.state.config.pgs.hash = null;
@@ -461,8 +485,20 @@ class Mon
 
     restart_watcher(cur_addr)
     {
+        if (this.ws)
+        {
+            this.ws.close();
+            this.ws = null;
+        }
+        if (this.ws_keepalive_timer)
+        {
+            clearInterval(this.ws_keepalive_timer);
+            this.ws_keepalive_timer = null;
+        }
         if (this.selected_etcd_url == cur_addr)
+        {
             this.selected_etcd_url = null;
+        }
         this.start_watcher(this.config.etcd_mon_retries).catch(this.die);
     }
 
@@ -482,6 +518,7 @@ class Mon
                 const timer_id = setTimeout(() =>
                 {
                     this.ws.close();
+                    this.ws = null;
                     ok(false);
                 }, this.config.etcd_mon_timeout);
                 this.ws = new WebSocket(base+'/watch');
@@ -510,6 +547,20 @@ class Mon
             this.die('Failed to open etcd watch websocket');
         }
         const cur_addr = this.selected_etcd_url;
+        this.ws_alive = true;
+        this.ws_keepalive_timer = setInterval(() =>
+        {
+            if (this.ws_alive)
+            {
+                this.ws_alive = false;
+                this.ws.send(JSON.stringify({ progress_request: {} }));
+            }
+            else
+            {
+                console.log('etcd websocket timed out, restarting it');
+                this.restart_watcher(cur_addr);
+            }
+        }, (Number(this.config.etcd_keepalive_interval) || 30)*1000);
         this.ws.on('error', () => this.restart_watcher(cur_addr));
         this.ws.send(JSON.stringify({
             create_request: {
@@ -522,6 +573,7 @@ class Mon
         }));
         this.ws.on('message', (msg) =>
         {
+            this.ws_alive = true;
             let data;
             try
             {
@@ -558,7 +610,7 @@ class Mon
                     console.log('Revision '+data.result.header.revision+' events: ');
                 }
                 this.etcd_watch_revision = BigInt(data.result.header.revision)+BigInt(1);
-                for (const e of data.result.events)
+                for (const e of data.result.events||[])
                 {
                     this.parse_kv(e.kv);
                     const key = e.kv.key.substr(this.etcd_prefix.length);
@@ -631,11 +683,25 @@ class Mon
         }, this.etcd_start_timeout, 0);
     }
 
+    get_mon_state()
+    {
+        return { ip: this.local_ips(), hostname: os.hostname() };
+    }
+
     async get_lease()
     {
         const max_ttl = this.config.etcd_mon_ttl + this.config.etcd_mon_timeout/1000*this.config.etcd_mon_retries;
-        const res = await this.etcd_call('/lease/grant', { TTL: max_ttl }, this.config.etcd_mon_timeout, -1);
+        // Get lease
+        let res = await this.etcd_call('/lease/grant', { TTL: max_ttl }, this.config.etcd_mon_timeout, -1);
         this.etcd_lease_id = res.ID;
+        // Register in /mon/member, just for the information
+        const state = this.get_mon_state();
+        res = await this.etcd_call('/kv/put', {
+            key: b64(this.etcd_prefix+'/mon/member/'+this.etcd_lease_id),
+            value: b64(JSON.stringify(state)),
+            lease: ''+this.etcd_lease_id
+        }, this.etcd_start_timeout, 0);
+        // Set refresh timer
         this.lease_timer = setInterval(async () =>
         {
             const res = await this.etcd_call('/lease/keepalive', { ID: this.etcd_lease_id }, this.config.etcd_mon_timeout, this.config.etcd_mon_retries);
@@ -661,7 +727,7 @@ class Mon
 
     async become_master()
     {
-        const state = { ip: this.local_ips() };
+        const state = { ...this.get_mon_state(), id: ''+this.etcd_lease_id };
         while (1)
         {
             const res = await this.etcd_call('/kv/txn', {
@@ -839,27 +905,39 @@ class Mon
         return this.seed + 2147483648;
     }
 
-    pick_primary(pool_id, osd_set, up_osds)
+    pick_primary(pool_id, osd_set, up_osds, aff_osds)
     {
         let alive_set;
         if (this.state.config.pools[pool_id].scheme === 'replicated')
-            alive_set = osd_set.filter(osd_num => osd_num && up_osds[osd_num]);
+        {
+            // Prefer "affinity" OSDs
+            alive_set = osd_set.filter(osd_num => osd_num && aff_osds[osd_num]);
+            if (!alive_set.length)
+                alive_set = osd_set.filter(osd_num => osd_num && up_osds[osd_num]);
+        }
         else
         {
             // Prefer data OSDs for EC because they can actually read something without an additional network hop
             const pg_data_size = (this.state.config.pools[pool_id].pg_size||0) -
                 (this.state.config.pools[pool_id].parity_chunks||0);
-            alive_set = osd_set.slice(0, pg_data_size).filter(osd_num => osd_num && up_osds[osd_num]);
+            alive_set = osd_set.slice(0, pg_data_size).filter(osd_num => osd_num && aff_osds[osd_num]);
             if (!alive_set.length)
-                alive_set = osd_set.filter(osd_num => osd_num && up_osds[osd_num]);
+                alive_set = osd_set.filter(osd_num => osd_num && aff_osds[osd_num]);
+            if (!alive_set.length)
+            {
+                alive_set = osd_set.slice(0, pg_data_size).filter(osd_num => osd_num && up_osds[osd_num]);
+                if (!alive_set.length)
+                    alive_set = osd_set.filter(osd_num => osd_num && up_osds[osd_num]);
+            }
         }
         if (!alive_set.length)
             return 0;
         return alive_set[this.rng() % alive_set.length];
     }
 
-    save_new_pgs_txn(request, pool_id, up_osds, prev_pgs, new_pgs, pg_history)
+    save_new_pgs_txn(request, pool_id, up_osds, osd_tree, prev_pgs, new_pgs, pg_history)
     {
+        const aff_osds = this.get_affinity_osds(this.state.config.pools[pool_id], up_osds, osd_tree);
         const pg_items = {};
         this.reset_rng();
         new_pgs.map((osd_set, i) =>
@@ -867,7 +945,7 @@ class Mon
             osd_set = osd_set.map(osd_num => osd_num === LPOptimizer.NO_OSD ? 0 : osd_num);
             pg_items[i+1] = {
                 osd_set,
-                primary: this.pick_primary(pool_id, osd_set, up_osds),
+                primary: this.pick_primary(pool_id, osd_set, up_osds, aff_osds),
             };
             if (prev_pgs[i] && prev_pgs[i].join(' ') != osd_set.join(' ') &&
                 prev_pgs[i].filter(osd_num => osd_num).length > 0)
@@ -998,6 +1076,13 @@ class Mon
                 console.log('Pool '+pool_id+' has invalid osd_tags (must be a string or array of strings)');
             return false;
         }
+        if (pool_cfg.primary_affinity_tags && typeof(pool_cfg.primary_affinity_tags) != 'string' &&
+            (!(pool_cfg.primary_affinity_tags instanceof Array) || pool_cfg.primary_affinity_tags.filter(t => typeof t != 'string').length > 0))
+        {
+            if (warn)
+                console.log('Pool '+pool_id+' has invalid primary_affinity_tags (must be a string or array of strings)');
+            return false;
+        }
         return true;
     }
 
@@ -1027,6 +1112,17 @@ class Mon
         }
     }
 
+    get_affinity_osds(pool_cfg, up_osds, osd_tree)
+    {
+        let aff_osds = up_osds;
+        if (pool_cfg.primary_affinity_tags)
+        {
+            aff_osds = { ...up_osds };
+            this.filter_osds_by_tags(osd_tree, { x: aff_osds }, pool_cfg.primary_affinity_tags);
+        }
+        return aff_osds;
+    }
+
     async recheck_pgs()
     {
         // Take configuration and state, check it against the stored configuration hash
@@ -1057,7 +1153,7 @@ class Mon
                     {
                         prev_pgs[pg-1] = this.state.config.pgs.items[pool_id][pg].osd_set;
                     }
-                    this.save_new_pgs_txn(etcd_request, pool_id, up_osds, prev_pgs, [], []);
+                    this.save_new_pgs_txn(etcd_request, pool_id, up_osds, osd_tree, prev_pgs, [], []);
                 }
             }
             for (const pool_id in this.state.config.pools)
@@ -1097,7 +1193,7 @@ class Mon
                     pg_size: pool_cfg.pg_size,
                     pg_minsize: pool_cfg.pg_minsize,
                     max_combinations: pool_cfg.max_osd_combinations,
-                    round_robin: pool_cfg.scheme != 'replicated',
+                    ordered: pool_cfg.scheme != 'replicated',
                 };
                 let optimize_result;
                 if (old_pg_count > 0)
@@ -1120,10 +1216,6 @@ class Mon
                         {
                             pg.push(0);
                         }
-                        while (pg.length > pool_cfg.pg_size)
-                        {
-                            pg.pop();
-                        }
                     }
                     if (!this.state.config.pgs.hash)
                     {
@@ -1159,8 +1251,8 @@ class Mon
                 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,
-                    pg_real_size: pg_effsize,
+                    pg_real_size: pg_effsize || pool_cfg.pg_size,
-                    raw_to_usable: pg_effsize / (pool_cfg.scheme === 'replicated'
+                    raw_to_usable: (pg_effsize || pool_cfg.pg_size) / (pool_cfg.scheme === 'replicated'
                         ? 1 : (pool_cfg.pg_size - (pool_cfg.parity_chunks||0))),
                     space_efficiency: optimize_result.space/(optimize_result.total_space||1),
                 };
@@ -1168,7 +1260,7 @@ class Mon
                     key: b64(this.etcd_prefix+'/pool/stats/'+pool_id),
                     value: b64(JSON.stringify(this.state.pool.stats[pool_id])),
                 } });
-                this.save_new_pgs_txn(etcd_request, pool_id, up_osds, real_prev_pgs, optimize_result.int_pgs, pg_history);
+                this.save_new_pgs_txn(etcd_request, pool_id, up_osds, osd_tree, real_prev_pgs, optimize_result.int_pgs, pg_history);
             }
             this.state.config.pgs.hash = tree_hash;
             await this.save_pg_config(etcd_request);
@@ -1185,13 +1277,14 @@ class Mon
                     continue;
                 }
                 const replicated = pool_cfg.scheme === 'replicated';
+                const aff_osds = this.get_affinity_osds(pool_cfg, up_osds, osd_tree);
                 this.reset_rng();
                 for (let pg_num = 1; pg_num <= pool_cfg.pg_count; pg_num++)
                 {
                     const pg_cfg = this.state.config.pgs.items[pool_id][pg_num];
                     if (pg_cfg)
                     {
-                        const new_primary = this.pick_primary(pool_id, pg_cfg.osd_set, up_osds);
+                        const new_primary = this.pick_primary(pool_id, pg_cfg.osd_set, up_osds, aff_osds);
                         if (pg_cfg.primary != new_primary)
                         {
                             console.log(
@@ -1307,21 +1400,30 @@ class Mon
         const tm = prev_stats ? BigInt(timestamp - prev_stats.timestamp) : 0;
         for (const op in op_stats)
         {
-            op_stats[op].bps = prev_stats ? (op_stats[op].bytes - prev_stats.op_stats[op].bytes) * 1000n / tm : 0;
+            if (prev_stats && prev_stats.op_stats && prev_stats.op_stats[op])
-            op_stats[op].iops = prev_stats ? (op_stats[op].count - prev_stats.op_stats[op].count) * 1000n / tm : 0;
+            {
-            op_stats[op].lat = prev_stats ? (op_stats[op].usec - prev_stats.op_stats[op].usec)
+                op_stats[op].bps = (op_stats[op].bytes - prev_stats.op_stats[op].bytes) * 1000n / tm;
-                / ((op_stats[op].count - prev_stats.op_stats[op].count) || 1n) : 0;
+                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)
         {
-            subop_stats[op].iops = prev_stats ? (subop_stats[op].count - prev_stats.subop_stats[op].count) * 1000n / tm : 0;
+            if (prev_stats && prev_stats.subop_stats && prev_stats.subop_stats[op])
-            subop_stats[op].lat = prev_stats ? (subop_stats[op].usec - prev_stats.subop_stats[op].usec)
+            {
-                / ((subop_stats[op].count - prev_stats.subop_stats[op].count) || 1n) : 0;
+                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)
         {
-            recovery_stats[op].bps = prev_stats ? (recovery_stats[op].bytes - prev_stats.recovery_stats[op].bytes) * 1000n / tm : 0;
+            if (prev_stats && prev_stats.recovery_stats && prev_stats.recovery_stats[op])
-            recovery_stats[op].iops = prev_stats ? (recovery_stats[op].count - prev_stats.recovery_stats[op].count) * 1000n / tm : 0;
+            {
+                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 };
     }

mon/simple-offsets.js

@@ -49,7 +49,8 @@ async function run()
     }
     options.journal_offset = Math.ceil(options.journal_offset/options.device_block_size)*options.device_block_size;
     const meta_offset = options.journal_offset + Math.ceil(options.journal_size/options.device_block_size)*options.device_block_size;
-    const entries_per_block = Math.floor(options.device_block_size / (24 + 2*options.object_size/options.bitmap_granularity/8));
+    const meta_entry_size = 24 + 2*options.object_size/options.bitmap_granularity/8;
+    const entries_per_block = Math.floor(options.device_block_size / meta_entry_size);
     const object_count = Math.floor((device_size-meta_offset)/options.object_size);
     const meta_size = Math.ceil(1 + object_count / entries_per_block) * options.device_block_size;
     const data_offset = meta_offset + meta_size;

mon/test-optimize-simple.js

@@ -5,21 +5,45 @@ const LPOptimizer = require('./lp-optimizer.js');
 
 async function run()
 {
-    const osd_tree = { a: { 1: 1 }, b: { 2: 1 }, c: { 3: 1 } };
+    const osd_tree = {
+        100: { 1: 1 },
+        200: { 2: 1 },
+        300: { 3: 1 },
+    };
+
     let res;
 
     console.log('16 PGs, size=3');
-    res = await LPOptimizer.optimize_initial({ osd_tree, pg_size: 3, pg_count: 16 });
+    res = await LPOptimizer.optimize_initial({ osd_tree, pg_size: 3, pg_count: 16, ordered: false });
     LPOptimizer.print_change_stats(res, false);
-
+    assert(res.space == 3, 'Initial distribution');
-    console.log('\nReduce PG size to 2');
+    console.log('\nChange size to 2');
-    res = await LPOptimizer.optimize_change({ prev_pgs: res.int_pgs.map(pg => pg.slice(0, 2)), osd_tree, pg_size: 2 });
+    res = await LPOptimizer.optimize_change({ prev_pgs: res.int_pgs, osd_tree, pg_size: 2, ordered: false });
     LPOptimizer.print_change_stats(res, false);
-
+    assert(res.space >= 3*14/16 && res.osd_differs == 0, 'Redistribution');
     console.log('\nRemove OSD 3');
-    delete osd_tree['c'];
+    const no3_tree = { ...osd_tree };
-    res = await LPOptimizer.optimize_change({ prev_pgs: res.int_pgs, osd_tree, pg_size: 2 });
+    delete no3_tree['300'];
+    res = await LPOptimizer.optimize_change({ prev_pgs: res.int_pgs, osd_tree: no3_tree, pg_size: 2, ordered: false });
     LPOptimizer.print_change_stats(res, false);
+    assert(res.space == 2, 'Redistribution after OSD removal');
+
+    console.log('\n16 PGs, size=3, ordered');
+    res = await LPOptimizer.optimize_initial({ osd_tree, pg_size: 3, pg_count: 16, ordered: true });
+    LPOptimizer.print_change_stats(res, false);
+    assert(res.space == 3, 'Initial distribution');
+    console.log('\nChange size to 2, ordered');
+    res = await LPOptimizer.optimize_change({ prev_pgs: res.int_pgs, osd_tree, pg_size: 2, ordered: true });
+    LPOptimizer.print_change_stats(res, false);
+    assert(res.space >= 3*14/16 && res.osd_differs < 8, 'Redistribution');
+}
+
+function assert(cond, txt)
+{
+    if (!cond)
+    {
+        throw new Error((txt||'test')+' failed');
+    }
 }
 
 run().catch(console.error);

mon/test-optimize-undersized.js

@@ -45,30 +45,45 @@ async function run()
     console.log('Empty tree:');
     let res = await LPOptimizer.optimize_initial({ osd_tree: cur_tree, pg_size: 3, pg_count: 256 });
     LPOptimizer.print_change_stats(res, false);
+    assert(res.space == 0);
     console.log('\nAdding 1st failure domain:');
     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['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['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['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['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['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);
+}
+
+function assert(cond, txt)
+{
+    if (!cond)
+    {
+        throw new Error((txt||'test')+' failed');
+    }
 }
 
 run().catch(console.error);

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 = '0.6.10'
+VERSION = '0.6.17'
 
 LOG = logging.getLogger(__name__)
 
@@ -355,7 +355,25 @@ class VitastorDriver(driver.CloneableImageVD,
     def revert_to_snapshot(self, context, volume, snapshot):
         """Revert a volume to a given snapshot."""
 
-        # FIXME Delete the image, then recreate it from the snapshot
+        vol_name = utils.convert_str(snapshot.volume_name)
+        snap_name = utils.convert_str(snapshot.name)
+
+        # Delete the image and recreate it from the snapshot
+        args = [ 'vitastor-cli', 'rm', vol_name, *(self._vitastor_args()) ]
+        try:
+            self._execute(*args)
+        except processutils.ProcessExecutionError as exc:
+            LOG.error("Failed to delete image "+vol_name+": "+exc)
+            raise exception.VolumeBackendAPIException(data = exc.stderr)
+        args = [
+            'vitastor-cli', 'create', '--parent', vol_name+'@'+snap_name,
+            vol_name, *(self._vitastor_args())
+        ]
+        try:
+            self._execute(*args)
+        except processutils.ProcessExecutionError as exc:
+            LOG.error("Failed to recreate image "+vol_name+" from "+vol_name+"@"+snap_name+": "+exc)
+            raise exception.VolumeBackendAPIException(data = exc.stderr)
 
     def delete_snapshot(self, snapshot):
         """Deletes a snapshot."""
@@ -363,24 +381,15 @@ class VitastorDriver(driver.CloneableImageVD,
         vol_name = utils.convert_str(snapshot.volume_name)
         snap_name = utils.convert_str(snapshot.name)
 
-        # Find the snapshot
+        args = [
-        resp = self._etcd_txn({ 'success': [
+            'vitastor-cli', 'rm', vol_name+'@'+snap_name,
-            { 'request_range': { 'key': 'index/image/'+vol_name+'@'+snap_name } },
+            *(self._vitastor_args())
-        ] })
+        ]
-        if len(resp['responses'][0]['kvs']) == 0:
+        try:
-            raise exception.SnapshotNotFound(snapshot_id = snap_name)
+            self._execute(*args)
-        inode_id = int(resp['responses'][0]['kvs'][0]['value']['id'])
+        except processutils.ProcessExecutionError as exc:
-        pool_id = int(resp['responses'][0]['kvs'][0]['value']['pool_id'])
+            LOG.error("Failed to remove snapshot "+vol_name+'@'+snap_name+": "+exc)
-        parents = {}
+            raise exception.VolumeBackendAPIException(data = exc.stderr)
-        parents[(pool_id << 48) | (inode_id & 0xffffffffffff)] = True
-
-        # Check if there are child volumes
-        children = self._child_count(parents)
-        if children > 0:
-            raise exception.SnapshotIsBusy(snapshot_name = snap_name)
-
-        # FIXME: We can't delete snapshots because we can't merge layers yet
-        raise exception.VolumeBackendAPIException(data = 'Snapshot delete (layer merge) is not implemented yet')
 
     def _child_count(self, parents):
         children = 0

rpm/build-tarball.sh

@@ -25,4 +25,4 @@ 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-0.6.10/#' --exclude 'rpm/*.rpm' -czf $VITASTOR/../vitastor-0.6.10$(rpm --eval '%dist').tar.gz *
+tar --transform 's#^#vitastor-0.6.17/#' --exclude 'rpm/*.rpm' -czf $VITASTOR/../vitastor-0.6.17$(rpm --eval '%dist').tar.gz *

rpm/vitastor-el7.Dockerfile

@@ -34,7 +34,7 @@ ADD . /root/vitastor
 RUN set -e; \
     cd /root/vitastor/rpm; \
     sh build-tarball.sh; \
-    cp /root/vitastor-0.6.10.el7.tar.gz ~/rpmbuild/SOURCES; \
+    cp /root/vitastor-0.6.17.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:        0.6.10
+Version:        0.6.17
 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-0.6.10.el7.tar.gz
+Source0:        vitastor-0.6.17.el7.tar.gz
 
 BuildRequires:  liburing-devel >= 0.6
 BuildRequires:  gperftools-devel
@@ -119,6 +119,7 @@ cp -r mon %buildroot/usr/lib/vitastor
 
 %files -n vitastor-client
 %_bindir/vitastor-nbd
+%_bindir/vitastor-nfs
 %_bindir/vitastor-cli
 %_bindir/vitastor-rm
 %_bindir/vita

rpm/vitastor-el8.Dockerfile

@@ -33,7 +33,7 @@ ADD . /root/vitastor
 RUN set -e; \
     cd /root/vitastor/rpm; \
     sh build-tarball.sh; \
-    cp /root/vitastor-0.6.10.el8.tar.gz ~/rpmbuild/SOURCES; \
+    cp /root/vitastor-0.6.17.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:        0.6.10
+Version:        0.6.17
 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-0.6.10.el8.tar.gz
+Source0:        vitastor-0.6.17.el8.tar.gz
 
 BuildRequires:  liburing-devel >= 0.6
 BuildRequires:  gperftools-devel
@@ -116,6 +116,7 @@ cp -r mon %buildroot/usr/lib/vitastor
 
 %files -n vitastor-client
 %_bindir/vitastor-nbd
+%_bindir/vitastor-nfs
 %_bindir/vitastor-cli
 %_bindir/vitastor-rm
 %_bindir/vita

src/CMakeLists.txt

@@ -15,7 +15,7 @@ if("${CMAKE_INSTALL_PREFIX}" MATCHES "^/usr/local/?$")
 	set(CMAKE_INSTALL_RPATH "${CMAKE_INSTALL_PREFIX}/${CMAKE_INSTALL_LIBDIR}")
 endif()
 
-add_definitions(-DVERSION="0.6.10")
+add_definitions(-DVERSION="0.6.17")
 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)
@@ -111,12 +111,11 @@ if (${WITH_FIO})
 	# libfio_vitastor_sec.so
 	add_library(fio_vitastor_sec SHARED
 		fio_sec_osd.cpp
+		rw_blocking.cpp
+		addr_util.cpp
 	)
 	target_link_libraries(fio_vitastor_sec
-		vitastor_common
 		tcmalloc_minimal
-		${LIBURING_LIBRARIES}
-		${IBVERBS_LIBRARIES}
 	)
 endif (${WITH_FIO})
 
@@ -125,6 +124,18 @@ add_library(vitastor_client SHARED
 	cluster_client.cpp
 	cluster_client_list.cpp
 	vitastor_c.cpp
+	cli_common.cpp
+	cli_alloc_osd.cpp
+	cli_simple_offsets.cpp
+	cli_status.cpp
+	cli_df.cpp
+	cli_ls.cpp
+	cli_create.cpp
+	cli_modify.cpp
+	cli_flatten.cpp
+	cli_merge.cpp
+	cli_rm_data.cpp
+	cli_rm.cpp
 )
 set_target_properties(vitastor_client PROPERTIES PUBLIC_HEADER "vitastor_c.h")
 target_link_libraries(vitastor_client
@@ -153,10 +164,24 @@ target_link_libraries(vitastor-nbd
 	vitastor_client
 )
 
+# vitastor-nfs
+add_executable(vitastor-nfs
+	nfs_proxy.cpp
+	nfs_conn.cpp
+	nfs_portmap.cpp
+	sha256.c
+	nfs/xdr_impl.cpp
+	nfs/rpc_xdr.cpp
+	nfs/portmap_xdr.cpp
+	nfs/nfs_xdr.cpp
+)
+target_link_libraries(vitastor-nfs
+	vitastor_client
+)
+
 # vitastor-cli
 add_executable(vitastor-cli
-	cli.cpp cli_alloc_osd.cpp cli_simple_offsets.cpp cli_df.cpp
+	cli.cpp
-	cli_ls.cpp cli_create.cpp cli_modify.cpp cli_flatten.cpp cli_merge.cpp cli_rm.cpp cli_snap_rm.cpp
 )
 target_link_libraries(vitastor-cli
 	vitastor_client
@@ -245,7 +270,7 @@ target_include_directories(test_cluster_client PUBLIC ${CMAKE_SOURCE_DIR}/src/mo
 
 ### Install
 
-install(TARGETS vitastor-osd vitastor-dump-journal vitastor-nbd vitastor-cli RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
+install(TARGETS vitastor-osd vitastor-dump-journal vitastor-nbd vitastor-nfs vitastor-cli RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
 install_symlink(vitastor-cli ${CMAKE_INSTALL_PREFIX}/${CMAKE_INSTALL_BINDIR}/vitastor-rm)
 install_symlink(vitastor-cli ${CMAKE_INSTALL_PREFIX}/${CMAKE_INSTALL_BINDIR}/vita)
 install(

src/addr_util.cpp

@@ -1,4 +1,9 @@
+#include <sys/socket.h>
+#include <unistd.h>
 #include <arpa/inet.h>
+#include <net/if.h>
+#include <sys/types.h>
+#include <ifaddrs.h>
 #include <string.h>
 #include <stdio.h>
 
@@ -6,7 +11,7 @@
 
 #include "addr_util.h"
 
-bool string_to_addr(std::string str, bool parse_port, int default_port, struct sockaddr *addr)
+bool string_to_addr(std::string str, bool parse_port, int default_port, struct sockaddr_storage *addr)
 {
     if (parse_port)
     {
@@ -22,7 +27,7 @@ bool string_to_addr(std::string str, bool parse_port, int default_port, struct s
     }
     if (inet_pton(AF_INET, str.c_str(), &((struct sockaddr_in*)addr)->sin_addr) == 1)
     {
-        addr->sa_family = AF_INET;
+        addr->ss_family = AF_INET;
         ((struct sockaddr_in*)addr)->sin_port = htons(default_port);
         return true;
     }
@@ -30,31 +35,204 @@ bool string_to_addr(std::string str, bool parse_port, int default_port, struct s
         str = str.substr(1, str.length()-2);
     if (inet_pton(AF_INET6, str.c_str(), &((struct sockaddr_in6*)addr)->sin6_addr) == 1)
     {
-        addr->sa_family = AF_INET6;
+        addr->ss_family = AF_INET6;
         ((struct sockaddr_in6*)addr)->sin6_port = htons(default_port);
         return true;
     }
     return false;
 }
 
-std::string addr_to_string(const sockaddr &addr)
+std::string addr_to_string(const sockaddr_storage &addr)
 {
     char peer_str[256];
     bool ok = false;
     int port;
-    if (addr.sa_family == AF_INET)
+    if (addr.ss_family == AF_INET)
     {
         ok = !!inet_ntop(AF_INET, &((sockaddr_in*)&addr)->sin_addr, peer_str, 256);
         port = ntohs(((sockaddr_in*)&addr)->sin_port);
     }
-    else if (addr.sa_family == AF_INET6)
+    else if (addr.ss_family == AF_INET6)
     {
         ok = !!inet_ntop(AF_INET6, &((sockaddr_in6*)&addr)->sin6_addr, peer_str, 256);
         port = ntohs(((sockaddr_in6*)&addr)->sin6_port);
     }
     else
-        throw std::runtime_error("Unknown address family "+std::to_string(addr.sa_family));
+        throw std::runtime_error("Unknown address family "+std::to_string(addr.ss_family));
     if (!ok)
         throw std::runtime_error(std::string("inet_ntop: ") + strerror(errno));
     return std::string(peer_str)+":"+std::to_string(port);
 }
+
+static bool cidr_match(const in_addr &addr, const in_addr &net, uint8_t bits)
+{
+    if (bits == 0)
+    {
+        // C99 6.5.7 (3): u32 << 32 is undefined behaviour
+        return true;
+    }
+    return !((addr.s_addr ^ net.s_addr) & htonl(0xFFFFFFFFu << (32 - bits)));
+}
+
+static bool cidr6_match(const in6_addr &address, const in6_addr &network, uint8_t bits)
+{
+    const uint32_t *a = address.s6_addr32;
+    const uint32_t *n = network.s6_addr32;
+    int bits_whole, bits_incomplete;
+    bits_whole = bits >> 5;         // number of whole u32
+    bits_incomplete = bits & 0x1F;  // number of bits in incomplete u32
+    if (bits_whole && memcmp(a, n, bits_whole << 2))
+        return false;
+    if (bits_incomplete)
+    {
+        uint32_t mask = htonl((0xFFFFFFFFu) << (32 - bits_incomplete));
+        if ((a[bits_whole] ^ n[bits_whole]) & mask)
+            return false;
+    }
+    return true;
+}
+
+struct addr_mask_t
+{
+    sa_family_t family;
+    in_addr ipv4;
+    in6_addr ipv6;
+    uint8_t bits;
+};
+
+std::vector<std::string> getifaddr_list(std::vector<std::string> mask_cfg, bool include_v6)
+{
+    std::vector<addr_mask_t> masks;
+    for (auto mask: mask_cfg)
+    {
+        unsigned bits = 0;
+        int p = mask.find('/');
+        if (p != std::string::npos)
+        {
+            char null_byte = 0;
+            if (sscanf(mask.c_str()+p+1, "%u%c", &bits, &null_byte) != 1 || bits > 128)
+            {
+                throw std::runtime_error((include_v6 ? "Invalid IPv4 address mask: " : "Invalid IP address mask: ") + mask);
+            }
+            mask = mask.substr(0, p);
+        }
+        in_addr ipv4;
+        in6_addr ipv6;
+        if (inet_pton(AF_INET, mask.c_str(), &ipv4) == 1)
+        {
+            if (bits > 32)
+            {
+                throw std::runtime_error((include_v6 ? "Invalid IPv4 address mask: " : "Invalid IP address mask: ") + mask);
+            }
+            masks.push_back((addr_mask_t){ .family = AF_INET, .ipv4 = ipv4, .bits = (uint8_t)bits });
+        }
+        else if (include_v6 && inet_pton(AF_INET6, mask.c_str(), &ipv6) == 1)
+        {
+            masks.push_back((addr_mask_t){ .family = AF_INET6, .ipv6 = ipv6, .bits = (uint8_t)bits });
+        }
+        else
+        {
+            throw std::runtime_error((include_v6 ? "Invalid IPv4 address mask: " : "Invalid IP address mask: ") + mask);
+        }
+    }
+    std::vector<std::string> addresses;
+    ifaddrs *list, *ifa;
+    if (getifaddrs(&list) == -1)
+    {
+        throw std::runtime_error(std::string("getifaddrs: ") + strerror(errno));
+    }
+    for (ifa = list; ifa != NULL; ifa = ifa->ifa_next)
+    {
+        if (!ifa->ifa_addr)
+        {
+            continue;
+        }
+        int family = ifa->ifa_addr->sa_family;
+        if ((family == AF_INET || family == AF_INET6 && include_v6) &&
+            (ifa->ifa_flags & (IFF_UP | IFF_RUNNING | IFF_LOOPBACK)) == (IFF_UP | IFF_RUNNING))
+        {
+            void *addr_ptr;
+            if (family == AF_INET)
+            {
+                addr_ptr = &((sockaddr_in *)ifa->ifa_addr)->sin_addr;
+            }
+            else
+            {
+                addr_ptr = &((sockaddr_in6 *)ifa->ifa_addr)->sin6_addr;
+            }
+            if (masks.size() > 0)
+            {
+                int i;
+                for (i = 0; i < masks.size(); i++)
+                {
+                    if (masks[i].family == family && (family == AF_INET
+                        ? cidr_match(*(in_addr*)addr_ptr, masks[i].ipv4, masks[i].bits)
+                        : cidr6_match(*(in6_addr*)addr_ptr, masks[i].ipv6, masks[i].bits)))
+                    {
+                        break;
+                    }
+                }
+                if (i >= masks.size())
+                {
+                    continue;
+                }
+            }
+            char addr[INET6_ADDRSTRLEN];
+            if (!inet_ntop(family, addr_ptr, addr, INET6_ADDRSTRLEN))
+            {
+                throw std::runtime_error(std::string("inet_ntop: ") + strerror(errno));
+            }
+            addresses.push_back(std::string(addr));
+        }
+    }
+    freeifaddrs(list);
+    return addresses;
+}
+
+int create_and_bind_socket(std::string bind_address, int bind_port, int listen_backlog, int *listening_port)
+{
+    sockaddr_storage addr;
+    if (!string_to_addr(bind_address, 0, bind_port, &addr))
+    {
+        throw std::runtime_error("bind address "+bind_address+" is not valid");
+    }
+
+    int listen_fd = socket(addr.ss_family, SOCK_STREAM, 0);
+    if (listen_fd < 0)
+    {
+        throw std::runtime_error(std::string("socket: ") + strerror(errno));
+    }
+    int enable = 1;
+    setsockopt(listen_fd, SOL_SOCKET, SO_REUSEADDR, &enable, sizeof(enable));
+
+    if (bind(listen_fd, (sockaddr*)&addr, sizeof(addr)) < 0)
+    {
+        close(listen_fd);
+        throw std::runtime_error(std::string("bind: ") + strerror(errno));
+    }
+    if (listening_port)
+    {
+        if (bind_port == 0)
+        {
+            socklen_t len = sizeof(addr);
+            if (getsockname(listen_fd, (sockaddr *)&addr, &len) == -1)
+            {
+                close(listen_fd);
+                throw std::runtime_error(std::string("getsockname: ") + strerror(errno));
+            }
+            *listening_port = ntohs(((sockaddr_in*)&addr)->sin_port);
+        }
+        else
+        {
+            *listening_port = bind_port;
+        }
+    }
+
+    if (listen(listen_fd, listen_backlog ? listen_backlog : 128) < 0)
+    {
+        close(listen_fd);
+        throw std::runtime_error(std::string("listen: ") + strerror(errno));
+    }
+
+    return listen_fd;
+}

src/addr_util.h

@@ -2,6 +2,9 @@
 
 #include <sys/socket.h>
 #include <string>
+#include <vector>
 
-bool string_to_addr(std::string str, bool parse_port, int default_port, struct sockaddr *addr);
+bool string_to_addr(std::string str, bool parse_port, int default_port, struct sockaddr_storage *addr);
-std::string addr_to_string(const sockaddr &addr);
+std::string addr_to_string(const sockaddr_storage &addr);
+std::vector<std::string> getifaddr_list(std::vector<std::string> mask_cfg = std::vector<std::string>(), bool include_v6 = false);
+int create_and_bind_socket(std::string bind_address, int bind_port, int listen_backlog, int *listening_port);

src/allocator.cpp

@@ -25,7 +25,7 @@ allocator::allocator(uint64_t blocks)
     size = free = blocks;
     last_one_mask = (blocks % 64) == 0
         ? UINT64_MAX
-        : ((1l << (blocks % 64)) - 1);
+        : (((uint64_t)1 << (blocks % 64)) - 1);
     for (uint64_t i = 0; i < total; i++)
     {
         mask[i] = 0;
@@ -79,7 +79,7 @@ void allocator::set(uint64_t addr, bool value)
             }
             if (value)
             {
-                mask[last] = mask[last] | (1l << bit);
+                mask[last] = mask[last] | ((uint64_t)1 << bit);
                 if (mask[last] != (!is_last || cur_addr/64 < size/64
                     ? UINT64_MAX : last_one_mask))
                 {
@@ -88,7 +88,7 @@ void allocator::set(uint64_t addr, bool value)
             }
             else
             {
-                mask[last] = mask[last] & ~(1l << bit);
+                mask[last] = mask[last] & ~((uint64_t)1 << bit);
             }
             is_last = false;
             if (p2 > 1)

src/blockstore.h

@@ -21,7 +21,7 @@
 // Memory alignment for direct I/O (usually 512 bytes)
 // All other alignments must be a multiple of this one
 #ifndef MEM_ALIGNMENT
-#define MEM_ALIGNMENT 512
+#define MEM_ALIGNMENT 4096
 #endif
 
 // Default block size is 128 KB, current allowed range is 4K - 128M

src/blockstore_flush.cpp

@@ -185,7 +185,7 @@ void journal_flusher_t::release_trim()
 void journal_flusher_t::dump_diagnostics()
 {
     const char *unflushable_type = "";
-    obj_ver_id unflushable = { 0 };
+    obj_ver_id unflushable = {};
     // Try to find out if there is a flushable object for information
     for (object_id cur_oid: flush_queue)
     {
@@ -415,8 +415,11 @@ stop_flusher:
         flusher->active_flushers++;
 resume_1:
         // Find it in clean_db
-        clean_it = bs->clean_db.find(cur.oid);
+        {
-        old_clean_loc = (clean_it != bs->clean_db.end() ? clean_it->second.location : UINT64_MAX);
+            auto & clean_db = bs->clean_db_shard(cur.oid);
+            auto clean_it = clean_db.find(cur.oid);
+            old_clean_loc = (clean_it != clean_db.end() ? clean_it->second.location : UINT64_MAX);
+        }
         // Scan dirty versions of the object
         if (!scan_dirty(1))
         {
@@ -486,8 +489,8 @@ resume_1:
         if (bs->clean_entry_bitmap_size)
         {
             new_clean_bitmap = (bs->inmemory_meta
-                ? meta_new.buf + meta_new.pos*bs->clean_entry_size + sizeof(clean_disk_entry)
+                ? (uint8_t*)meta_new.buf + meta_new.pos*bs->clean_entry_size + sizeof(clean_disk_entry)
-                : bs->clean_bitmap + (clean_loc >> bs->block_order)*(2*bs->clean_entry_bitmap_size));
+                : (uint8_t*)bs->clean_bitmap + (clean_loc >> bs->block_order)*(2*bs->clean_entry_bitmap_size));
             if (clean_init_bitmap)
             {
                 memset(new_clean_bitmap, 0, bs->clean_entry_bitmap_size);
@@ -533,7 +536,7 @@ resume_1:
                 return false;
             }
             // zero out old metadata entry
-            memset(meta_old.buf + meta_old.pos*bs->clean_entry_size, 0, bs->clean_entry_size);
+            memset((uint8_t*)meta_old.buf + meta_old.pos*bs->clean_entry_size, 0, bs->clean_entry_size);
             await_sqe(15);
             data->iov = (struct iovec){ meta_old.buf, bs->meta_block_size };
             data->callback = simple_callback_w;
@@ -544,7 +547,7 @@ resume_1:
         }
         if (has_delete)
         {
-            clean_disk_entry *new_entry = (clean_disk_entry*)(meta_new.buf + meta_new.pos*bs->clean_entry_size);
+            clean_disk_entry *new_entry = (clean_disk_entry*)((uint8_t*)meta_new.buf + meta_new.pos*bs->clean_entry_size);
             if (new_entry->oid.inode != 0 && new_entry->oid != cur.oid)
             {
                 printf("Fatal error (metadata corruption or bug): tried to delete metadata entry %lu (%lx:%lx v%lu) while deleting %lx:%lx\n",
@@ -553,11 +556,11 @@ resume_1:
                 exit(1);
             }
             // zero out new metadata entry
-            memset(meta_new.buf + meta_new.pos*bs->clean_entry_size, 0, bs->clean_entry_size);
+            memset((uint8_t*)meta_new.buf + meta_new.pos*bs->clean_entry_size, 0, bs->clean_entry_size);
         }
         else
         {
-            clean_disk_entry *new_entry = (clean_disk_entry*)(meta_new.buf + meta_new.pos*bs->clean_entry_size);
+            clean_disk_entry *new_entry = (clean_disk_entry*)((uint8_t*)meta_new.buf + meta_new.pos*bs->clean_entry_size);
             if (new_entry->oid.inode != 0 && new_entry->oid != cur.oid)
             {
                 printf("Fatal error (metadata corruption or bug): tried to overwrite non-zero metadata entry %lu (%lx:%lx v%lu) with %lx:%lx v%lu\n",
@@ -575,7 +578,7 @@ resume_1:
             if (bs->clean_entry_bitmap_size)
             {
                 void *bmp_ptr = bs->clean_entry_bitmap_size > sizeof(void*) ? dirty_end->second.bitmap : &dirty_end->second.bitmap;
-                memcpy((void*)(new_entry+1) + bs->clean_entry_bitmap_size, bmp_ptr, bs->clean_entry_bitmap_size);
+                memcpy((uint8_t*)(new_entry+1) + bs->clean_entry_bitmap_size, bmp_ptr, bs->clean_entry_bitmap_size);
             }
         }
         await_sqe(6);
@@ -762,7 +765,7 @@ bool journal_flusher_co::scan_dirty(int wait_base)
                         if (bs->journal.inmemory)
                         {
                             // Take it from memory
-                            memcpy(it->buf, bs->journal.buffer + submit_offset, submit_len);
+                            memcpy(it->buf, (uint8_t*)bs->journal.buffer + submit_offset, submit_len);
                         }
                         else
                         {
@@ -826,7 +829,7 @@ bool journal_flusher_co::modify_meta_read(uint64_t meta_loc, flusher_meta_write_
     wr.pos = ((meta_loc >> bs->block_order) % (bs->meta_block_size / bs->clean_entry_size));
     if (bs->inmemory_meta)
     {
-        wr.buf = bs->metadata_buffer + wr.sector;
+        wr.buf = (uint8_t*)bs->metadata_buffer + wr.sector;
         return true;
     }
     wr.it = flusher->meta_sectors.find(wr.sector);
@@ -870,10 +873,11 @@ void journal_flusher_co::update_clean_db()
 #endif
         bs->data_alloc->set(old_clean_loc >> bs->block_order, false);
     }
+    auto & clean_db = bs->clean_db_shard(cur.oid);
     if (has_delete)
     {
-        auto clean_it = bs->clean_db.find(cur.oid);
+        auto clean_it = clean_db.find(cur.oid);
-        bs->clean_db.erase(clean_it);
+        clean_db.erase(clean_it);
 #ifdef BLOCKSTORE_DEBUG
         printf("Free block %lu from %lx:%lx v%lu (delete)\n",
             clean_loc >> bs->block_order,
@@ -884,7 +888,7 @@ void journal_flusher_co::update_clean_db()
     }
     else
     {
-        bs->clean_db[cur.oid] = {
+        clean_db[cur.oid] = {
             .version = cur.version,
             .location = clean_loc,
         };

src/blockstore_flush.h

@@ -49,7 +49,6 @@ class journal_flusher_co
     std::function<void(ring_data_t*)> simple_callback_r, simple_callback_w;
 
     bool skip_copy, has_delete, has_writes;
-    blockstore_clean_db_t::iterator clean_it;
     std::vector<copy_buffer_t> v;
     std::vector<copy_buffer_t>::iterator it;
     int copy_count;

src/blockstore_impl.cpp

@@ -118,7 +118,7 @@ void blockstore_impl_t::loop()
         // has_writes == 0 - no writes before the current queue item
         // has_writes == 1 - some writes in progress
         // has_writes == 2 - tried to submit some writes, but failed
-        int has_writes = 0, op_idx = 0, new_idx = 0;
+        int has_writes = 0, op_idx = 0, new_idx = 0, done_lists = 0;
         for (; op_idx < submit_queue.size(); op_idx++, new_idx++)
         {
             auto op = submit_queue[op_idx];
@@ -142,7 +142,6 @@ void blockstore_impl_t::loop()
                     continue;
                 }
             }
-            unsigned ring_space = ringloop->space_left();
             unsigned prev_sqe_pos = ringloop->save();
             // 0 = can't submit
             // 1 = in progress
@@ -199,9 +198,14 @@ void blockstore_impl_t::loop()
             }
             else if (op->opcode == BS_OP_LIST)
             {
-                // LIST doesn't need to be blocked by previous modifications
+                // LIST doesn't have to be blocked by previous modifications
-                process_list(op);
+                // But don't do a lot of LISTs at once, because they're blocking and potentially slow
-                wr_st = 2;
+                if (single_tick_list_limit <= 0 || done_lists < single_tick_list_limit)
+                {
+                    process_list(op);
+                    done_lists++;
+                    wr_st = 2;
+                }
             }
             if (wr_st == 2)
             {
@@ -212,7 +216,6 @@ void blockstore_impl_t::loop()
                 ringloop->restore(prev_sqe_pos);
                 if (PRIV(op)->wait_for == WAIT_SQE)
                 {
-                    PRIV(op)->wait_detail = 1 + ring_space;
                     // ring is full, stop submission
                     break;
                 }
@@ -282,7 +285,7 @@ void blockstore_impl_t::check_wait(blockstore_op_t *op)
 {
     if (PRIV(op)->wait_for == WAIT_SQE)
     {
-        if (ringloop->space_left() < PRIV(op)->wait_detail)
+        if (ringloop->sqes_left() < PRIV(op)->wait_detail)
         {
             // stop submission if there's still no free space
 #ifdef BLOCKSTORE_DEBUG
@@ -372,7 +375,7 @@ void blockstore_impl_t::enqueue_op(blockstore_op_t *op)
                     };
                 }
                 unstable_writes.clear();
-                op->callback = [this, old_callback](blockstore_op_t *op)
+                op->callback = [old_callback](blockstore_op_t *op)
                 {
                     obj_ver_id *vers = (obj_ver_id*)op->buf;
                     delete[] vers;
@@ -425,22 +428,104 @@ static bool replace_stable(object_id oid, uint64_t version, int search_start, in
     return false;
 }
 
+blockstore_clean_db_t& blockstore_impl_t::clean_db_shard(object_id oid)
+{
+    uint64_t pg_num = 0;
+    uint64_t pool_id = (oid.inode >> (64-POOL_ID_BITS));
+    auto sh_it = clean_db_settings.find(pool_id);
+    if (sh_it != clean_db_settings.end())
+    {
+        // like map_to_pg()
+        pg_num = (oid.stripe / sh_it->second.pg_stripe_size) % sh_it->second.pg_count + 1;
+    }
+    return clean_db_shards[(pool_id << (64-POOL_ID_BITS)) | pg_num];
+}
+
+void blockstore_impl_t::reshard_clean_db(pool_id_t pool, uint32_t pg_count, uint32_t pg_stripe_size)
+{
+    uint64_t pool_id = (uint64_t)pool;
+    std::map<pool_pg_id_t, blockstore_clean_db_t> new_shards;
+    auto sh_it = clean_db_shards.lower_bound((pool_id << (64-POOL_ID_BITS)));
+    while (sh_it != clean_db_shards.end() &&
+        (sh_it->first >> (64-POOL_ID_BITS)) == pool_id)
+    {
+        for (auto & pair: sh_it->second)
+        {
+            // like map_to_pg()
+            uint64_t pg_num = (pair.first.stripe / pg_stripe_size) % pg_count + 1;
+            uint64_t shard_id = (pool_id << (64-POOL_ID_BITS)) | pg_num;
+            new_shards[shard_id][pair.first] = pair.second;
+        }
+        clean_db_shards.erase(sh_it++);
+    }
+    for (sh_it = new_shards.begin(); sh_it != new_shards.end(); sh_it++)
+    {
+        auto & to = clean_db_shards[sh_it->first];
+        to.swap(sh_it->second);
+    }
+    clean_db_settings[pool_id] = (pool_shard_settings_t){
+        .pg_count = pg_count,
+        .pg_stripe_size = pg_stripe_size,
+    };
+}
+
 void blockstore_impl_t::process_list(blockstore_op_t *op)
 {
-    uint32_t list_pg = op->offset;
+    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;
     // Check PG
-    if (pg_count != 0 && (pg_stripe_size < MIN_BLOCK_SIZE || list_pg >= pg_count))
+    if (pg_count != 0 && (pg_stripe_size < MIN_BLOCK_SIZE || list_pg > pg_count))
     {
         op->retval = -EINVAL;
         FINISH_OP(op);
         return;
     }
-    // Copy clean_db entries (sorted)
+    // Check if the DB needs resharding
-    int stable_count = 0, stable_alloc = clean_db.size() / (pg_count ? pg_count : 1);
+    // (we don't know about PGs from the beginning, we only create "shards" here)
+    uint64_t first_shard = 0, last_shard = UINT64_MAX;
+    if (min_inode != 0 &&
+        // Check if min_inode == max_inode == pool_id<<N, i.e. this is a pool listing
+        (min_inode >> (64-POOL_ID_BITS)) == (max_inode >> (64-POOL_ID_BITS)))
+    {
+        pool_id_t pool_id = (min_inode >> (64-POOL_ID_BITS));
+        if (pg_count > 1)
+        {
+            // Per-pg listing
+            auto sh_it = clean_db_settings.find(pool_id);
+            if (sh_it == clean_db_settings.end() ||
+                sh_it->second.pg_count != pg_count ||
+                sh_it->second.pg_stripe_size != pg_stripe_size)
+            {
+                reshard_clean_db(pool_id, pg_count, pg_stripe_size);
+            }
+            first_shard = last_shard = ((uint64_t)pool_id << (64-POOL_ID_BITS)) | list_pg;
+        }
+        else
+        {
+            // Per-pool listing
+            first_shard = ((uint64_t)pool_id << (64-POOL_ID_BITS));
+            last_shard = ((uint64_t)(pool_id+1) << (64-POOL_ID_BITS)) - 1;
+        }
+    }
+    // Copy clean_db entries
+    int stable_count = 0, stable_alloc = 0;
+    if (min_inode != max_inode)
+    {
+        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;
+            stable_alloc += clean_db.size();
+        }
+    }
+    else
+    {
+        stable_alloc = 32768;
+    }
     obj_ver_id *stable = (obj_ver_id*)malloc(sizeof(obj_ver_id) * stable_alloc);
     if (!stable)
     {
@@ -448,7 +533,11 @@ void blockstore_impl_t::process_list(blockstore_op_t *op)
         FINISH_OP(op);
         return;
     }
+    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)
         {
@@ -463,26 +552,28 @@ void blockstore_impl_t::process_list(blockstore_op_t *op)
         }
         for (; clean_it != clean_end; clean_it++)
         {
-            if (!pg_count || ((clean_it->first.stripe / pg_stripe_size) % pg_count) == list_pg) // like map_to_pg()
+            if (stable_count >= stable_alloc)
             {
-                if (stable_count >= stable_alloc)
+                stable_alloc *= 2;
+                stable = (obj_ver_id*)realloc(stable, sizeof(obj_ver_id) * stable_alloc);
+                if (!stable)
                 {
-                    stable_alloc += 32768;
+                    op->retval = -ENOMEM;
-                    stable = (obj_ver_id*)realloc(stable, sizeof(obj_ver_id) * stable_alloc);
+                    FINISH_OP(op);
-                    if (!stable)
+                    return;
-                    {
-                        op->retval = -ENOMEM;
-                        FINISH_OP(op);
-                        return;
-                    }
                 }
-                stable[stable_count++] = {
-                    .oid = clean_it->first,
-                    .version = clean_it->second.version,
-                };
             }
+            stable[stable_count++] = {
+                .oid = clean_it->first,
+                .version = clean_it->second.version,
+            };
         }
     }
+    if (first_shard != last_shard)
+    {
+        // If that's not a per-PG listing, sort clean entries
+        std::sort(stable, stable+stable_count);
+    }
     int clean_stable_count = stable_count;
     // Copy dirty_db entries (sorted, too)
     int unstable_count = 0, unstable_alloc = 0;
@@ -508,7 +599,7 @@ void blockstore_impl_t::process_list(blockstore_op_t *op)
         }
         for (; dirty_it != dirty_end; dirty_it++)
         {
-            if (!pg_count || ((dirty_it->first.oid.stripe / pg_stripe_size) % pg_count) == list_pg) // like map_to_pg()
+            if (!pg_count || ((dirty_it->first.oid.stripe / pg_stripe_size) % pg_count + 1) == list_pg) // like map_to_pg()
             {
                 if (IS_DELETE(dirty_it->second.state))
                 {

src/blockstore_impl.h

@@ -55,9 +55,10 @@
 #define IS_DELETE(st) (((st) & 0x0F) == BS_ST_DELETE)
 
 #define BS_SUBMIT_CHECK_SQES(n) \
-    if (ringloop->space_left() < (n))\
+    if (ringloop->sqes_left() < (n))\
     {\
         /* Pause until there are more requests available */\
+        PRIV(op)->wait_detail = (n);\
         PRIV(op)->wait_for = WAIT_SQE;\
         return 0;\
     }
@@ -71,6 +72,7 @@
     if (!sqe)\
     {\
         /* Pause until there are more requests available */\
+        PRIV(op)->wait_detail = 1;\
         PRIV(op)->wait_for = WAIT_SQE;\
         return 0;\
     }
@@ -80,6 +82,7 @@
     if (!sqe)\
     {\
         /* Pause until there are more requests available */\
+        PRIV(op)->wait_detail = 1;\
         PRIV(op)->wait_for = WAIT_SQE;\
         return 0;\
     }
@@ -201,6 +204,17 @@ typedef std::map<obj_ver_id, dirty_entry> blockstore_dirty_db_t;
 
 #include "blockstore_flush.h"
 
+typedef uint32_t pool_id_t;
+typedef uint64_t pool_pg_id_t;
+
+#define POOL_ID_BITS 16
+
+struct pool_shard_settings_t
+{
+    uint32_t pg_count;
+    uint32_t pg_stripe_size;
+};
+
 class blockstore_impl_t
 {
     /******* OPTIONS *******/
@@ -238,11 +252,14 @@ class blockstore_impl_t
     int throttle_target_parallelism = 1;
     // Minimum difference in microseconds between target and real execution times to throttle the response
     int throttle_threshold_us = 50;
+    // Maximum number of LIST operations to be processed between
+    int single_tick_list_limit = 1;
     /******* END OF OPTIONS *******/
 
     struct ring_consumer_t ring_consumer;
 
-    blockstore_clean_db_t clean_db;
+    std::map<pool_id_t, pool_shard_settings_t> clean_db_settings;
+    std::map<pool_pg_id_t, blockstore_clean_db_t> clean_db_shards;
     uint8_t *clean_bitmap = NULL;
     blockstore_dirty_db_t dirty_db;
     std::vector<blockstore_op_t*> submit_queue;
@@ -280,7 +297,7 @@ class blockstore_impl_t
 
     friend class blockstore_init_meta;
     friend class blockstore_init_journal;
-    friend class blockstore_journal_check_t;
+    friend struct blockstore_journal_check_t;
     friend class journal_flusher_t;
     friend class journal_flusher_co;
 
@@ -291,6 +308,9 @@ class blockstore_impl_t
     void open_journal();
     uint8_t* get_clean_entry_bitmap(uint64_t block_loc, int offset);
 
+    blockstore_clean_db_t& clean_db_shard(object_id oid);
+    void reshard_clean_db(pool_id_t pool_id, uint32_t pg_count, uint32_t pg_stripe_size);
+
     // Journaling
     void prepare_journal_sector_write(int sector, blockstore_op_t *op);
     void handle_journal_write(ring_data_t *data, uint64_t flush_id);

src/blockstore_init.cpp

@@ -131,6 +131,7 @@ resume_1:
     }
     // Skip superblock
     bs->meta_offset += bs->meta_block_size;
+    bs->meta_len -= bs->meta_block_size;
     prev_done = 0;
     done_len = 0;
     done_pos = 0;
@@ -148,7 +149,7 @@ resume_1:
         {
             GET_SQE();
             data->iov = {
-                metadata_buffer + (bs->inmemory_meta
+                (uint8_t*)metadata_buffer + (bs->inmemory_meta
                     ? metadata_read
                     : (prev == 1 ? bs->metadata_buf_size : 0)),
                 bs->meta_len - metadata_read > bs->metadata_buf_size ? bs->metadata_buf_size : bs->meta_len - metadata_read,
@@ -169,13 +170,13 @@ resume_1:
         if (prev_done)
         {
             void *done_buf = bs->inmemory_meta
-                ? (metadata_buffer + done_pos)
+                ? ((uint8_t*)metadata_buffer + done_pos)
-                : (metadata_buffer + (prev_done == 2 ? bs->metadata_buf_size : 0));
+                : ((uint8_t*)metadata_buffer + (prev_done == 2 ? bs->metadata_buf_size : 0));
             unsigned count = bs->meta_block_size / bs->clean_entry_size;
             for (int sector = 0; sector < done_len; sector += bs->meta_block_size)
             {
                 // handle <count> entries
-                handle_entries(done_buf + sector, count, bs->block_order);
+                handle_entries((uint8_t*)done_buf + sector, count, bs->block_order);
                 done_cnt += count;
             }
             prev_done = 0;
@@ -215,17 +216,18 @@ void blockstore_init_meta::handle_entries(void* entries, unsigned count, int blo
 {
     for (unsigned i = 0; i < count; i++)
     {
-        clean_disk_entry *entry = (clean_disk_entry*)(entries + i*bs->clean_entry_size);
+        clean_disk_entry *entry = (clean_disk_entry*)((uint8_t*)entries + i*bs->clean_entry_size);
         if (!bs->inmemory_meta && bs->clean_entry_bitmap_size)
         {
             memcpy(bs->clean_bitmap + (done_cnt+i)*2*bs->clean_entry_bitmap_size, &entry->bitmap, 2*bs->clean_entry_bitmap_size);
         }
         if (entry->oid.inode > 0)
         {
-            auto clean_it = bs->clean_db.find(entry->oid);
+            auto & clean_db = bs->clean_db_shard(entry->oid);
-            if (clean_it == bs->clean_db.end() || clean_it->second.version < entry->version)
+            auto clean_it = clean_db.find(entry->oid);
+            if (clean_it == clean_db.end() || clean_it->second.version < entry->version)
             {
-                if (clean_it != bs->clean_db.end())
+                if (clean_it != clean_db.end())
                 {
                     // free the previous block
 #ifdef BLOCKSTORE_DEBUG
@@ -245,7 +247,7 @@ void blockstore_init_meta::handle_entries(void* entries, unsigned count, int blo
                 printf("Allocate block (clean entry) %lu: %lx:%lx v%lu\n", done_cnt+i, entry->oid.inode, entry->oid.stripe, entry->version);
 #endif
                 bs->data_alloc->set(done_cnt+i, true);
-                bs->clean_db[entry->oid] = (struct clean_entry){
+                clean_db[entry->oid] = (struct clean_entry){
                     .version = entry->version,
                     .location = (done_cnt+i) << block_order,
                 };
@@ -440,7 +442,7 @@ resume_1:
                 if (!bs->journal.inmemory)
                     submitted_buf = memalign_or_die(MEM_ALIGNMENT, JOURNAL_BUFFER_SIZE);
                 else
-                    submitted_buf = bs->journal.buffer + journal_pos;
+                    submitted_buf = (uint8_t*)bs->journal.buffer + journal_pos;
                 data->iov = {
                     submitted_buf,
                     end - journal_pos < JOURNAL_BUFFER_SIZE ? end - journal_pos : JOURNAL_BUFFER_SIZE,
@@ -570,7 +572,7 @@ int blockstore_init_journal::handle_journal_part(void *buf, uint64_t done_pos, u
     resume:
         while (pos < bs->journal.block_size)
         {
-            journal_entry *je = (journal_entry*)(buf + proc_pos - done_pos + pos);
+            journal_entry *je = (journal_entry*)((uint8_t*)buf + proc_pos - done_pos + pos);
             if (je->magic != JOURNAL_MAGIC || je_crc32(je) != je->crc32 ||
                 je->type < JE_MIN || je->type > JE_MAX || started && je->crc32_prev != crc32_last)
             {
@@ -619,7 +621,7 @@ int blockstore_init_journal::handle_journal_part(void *buf, uint64_t done_pos, u
                 if (location >= done_pos && location+je->small_write.len <= done_pos+len)
                 {
                     // data is within this buffer
-                    data_crc32 = crc32c(0, buf + location - done_pos, je->small_write.len);
+                    data_crc32 = crc32c(0, (uint8_t*)buf + location - done_pos, je->small_write.len);
                 }
                 else
                 {
@@ -634,7 +636,7 @@ int blockstore_init_journal::handle_journal_part(void *buf, uint64_t done_pos, u
                                 ? location+je->small_write.len : done[i].pos+done[i].len);
                             uint64_t part_begin = (location < done[i].pos ? done[i].pos : location);
                             covered += part_end - part_begin;
-                            data_crc32 = crc32c(data_crc32, done[i].buf + part_begin - done[i].pos, part_end - part_begin);
+                            data_crc32 = crc32c(data_crc32, (uint8_t*)done[i].buf + part_begin - done[i].pos, part_end - part_begin);
                         }
                     }
                     if (covered < je->small_write.len)
@@ -650,14 +652,15 @@ int blockstore_init_journal::handle_journal_part(void *buf, uint64_t done_pos, u
                     // interesting thing is that we must clear the corrupt entry if we're not readonly,
                     // because we don't write next entries in the same journal block
                     printf("Journal entry data is corrupt (data crc32 %x != %x)\n", data_crc32, je->small_write.crc32_data);
-                    memset(buf + proc_pos - done_pos + pos, 0, bs->journal.block_size - pos);
+                    memset((uint8_t*)buf + proc_pos - done_pos + pos, 0, bs->journal.block_size - pos);
                     bs->journal.next_free = prev_free;
-                    init_write_buf = buf + proc_pos - done_pos;
+                    init_write_buf = (uint8_t*)buf + proc_pos - done_pos;
                     init_write_sector = proc_pos;
                     return 0;
                 }
-                auto clean_it = bs->clean_db.find(je->small_write.oid);
+                auto & clean_db = bs->clean_db_shard(je->small_write.oid);
-                if (clean_it == bs->clean_db.end() ||
+                auto clean_it = clean_db.find(je->small_write.oid);
+                if (clean_it == clean_db.end() ||
                     clean_it->second.version < je->small_write.version)
                 {
                     obj_ver_id ov = {
@@ -665,7 +668,7 @@ int blockstore_init_journal::handle_journal_part(void *buf, uint64_t done_pos, u
                         .version = je->small_write.version,
                     };
                     void *bmp = NULL;
-                    void *bmp_from = (void*)je + sizeof(journal_entry_small_write);
+                    void *bmp_from = (uint8_t*)je + sizeof(journal_entry_small_write);
                     if (bs->clean_entry_bitmap_size <= sizeof(void*))
                     {
                         memcpy(&bmp, bmp_from, bs->clean_entry_bitmap_size);
@@ -735,8 +738,9 @@ int blockstore_init_journal::handle_journal_part(void *buf, uint64_t done_pos, u
                         erase_dirty_object(dirty_it);
                     }
                 }
-                auto clean_it = bs->clean_db.find(je->big_write.oid);
+                auto & clean_db = bs->clean_db_shard(je->big_write.oid);
-                if (clean_it == bs->clean_db.end() ||
+                auto clean_it = clean_db.find(je->big_write.oid);
+                if (clean_it == clean_db.end() ||
                     clean_it->second.version < je->big_write.version)
                 {
                     // oid, version, block
@@ -745,7 +749,7 @@ int blockstore_init_journal::handle_journal_part(void *buf, uint64_t done_pos, u
                         .version = je->big_write.version,
                     };
                     void *bmp = NULL;
-                    void *bmp_from = (void*)je + sizeof(journal_entry_big_write);
+                    void *bmp_from = (uint8_t*)je + sizeof(journal_entry_big_write);
                     if (bs->clean_entry_bitmap_size <= sizeof(void*))
                     {
                         memcpy(&bmp, bmp_from, bs->clean_entry_bitmap_size);
@@ -841,8 +845,9 @@ int blockstore_init_journal::handle_journal_part(void *buf, uint64_t done_pos, u
                     dirty_it--;
                     dirty_exists = dirty_it->first.oid == je->del.oid;
                 }
-                auto clean_it = bs->clean_db.find(je->del.oid);
+                auto & clean_db = bs->clean_db_shard(je->del.oid);
-                bool clean_exists = (clean_it != bs->clean_db.end() &&
+                auto clean_it = clean_db.find(je->del.oid);
+                bool clean_exists = (clean_it != clean_db.end() &&
                     clean_it->second.version < je->del.version);
                 if (!clean_exists && dirty_exists)
                 {
@@ -901,8 +906,9 @@ void blockstore_init_journal::erase_dirty_object(blockstore_dirty_db_t::iterator
             break;
         }
     }
-    auto clean_it = bs->clean_db.find(oid);
+    auto & clean_db = bs->clean_db_shard(oid);
-    uint64_t clean_loc = clean_it != bs->clean_db.end()
+    auto clean_it = clean_db.find(oid);
+    uint64_t clean_loc = clean_it != clean_db.end()
         ? clean_it->second.location : UINT64_MAX;
     if (exists && clean_loc == UINT64_MAX)
     {

src/blockstore_init.h

@@ -6,7 +6,7 @@
 class blockstore_init_meta
 {
     blockstore_impl_t *bs;
-    int wait_state = 0, wait_count = 0;
+    int wait_state = 0;
     bool zero_on_init = false;
     void *metadata_buffer = NULL;
     uint64_t metadata_read = 0;

src/blockstore_journal.cpp

@@ -96,7 +96,8 @@ int blockstore_journal_check_t::check_available(blockstore_op_t *op, int entries
         next_pos = next_pos + data_after;
         if (next_pos > bs->journal.len)
         {
-            next_pos = bs->journal.block_size + data_after;
+            if (right_dir)
+                next_pos = bs->journal.block_size + data_after;
             right_dir = false;
         }
     }
@@ -136,13 +137,13 @@ journal_entry* prefill_single_journal_entry(journal_t & journal, uint16_t type,
         journal.in_sector_pos = 0;
         journal.next_free = (journal.next_free+journal.block_size) < journal.len ? journal.next_free + journal.block_size : journal.block_size;
         memset(journal.inmemory
-            ? journal.buffer + journal.sector_info[journal.cur_sector].offset
+            ? (uint8_t*)journal.buffer + journal.sector_info[journal.cur_sector].offset
-            : journal.sector_buf + journal.block_size*journal.cur_sector, 0, journal.block_size);
+            : (uint8_t*)journal.sector_buf + journal.block_size*journal.cur_sector, 0, journal.block_size);
     }
     journal_entry *je = (struct journal_entry*)(
         (journal.inmemory
-            ? journal.buffer + journal.sector_info[journal.cur_sector].offset
+            ? (uint8_t*)journal.buffer + journal.sector_info[journal.cur_sector].offset
-            : journal.sector_buf + journal.block_size*journal.cur_sector) + journal.in_sector_pos
+            : (uint8_t*)journal.sector_buf + journal.block_size*journal.cur_sector) + journal.in_sector_pos
     );
     journal.in_sector_pos += size;
     je->magic = JOURNAL_MAGIC;
@@ -168,8 +169,8 @@ void blockstore_impl_t::prepare_journal_sector_write(int cur_sector, blockstore_
         journal.sector_info[cur_sector].flush_count++;
         data->iov = (struct iovec){
             (journal.inmemory
-                ? journal.buffer + journal.sector_info[cur_sector].offset
+                ? (uint8_t*)journal.buffer + journal.sector_info[cur_sector].offset
-                : journal.sector_buf + journal.block_size*cur_sector),
+                : (uint8_t*)journal.sector_buf + journal.block_size*cur_sector),
             journal.block_size
         };
         data->callback = [this, flush_id = journal.submit_id](ring_data_t *data) { handle_journal_write(data, flush_id); };

src/blockstore_read.cpp

@@ -24,7 +24,7 @@ int blockstore_impl_t::fulfill_read_push(blockstore_op_t *op, void *buf, uint64_
     }
     if (journal.inmemory && IS_JOURNAL(item_state))
     {
-        memcpy(buf, journal.buffer + offset, len);
+        memcpy(buf, (uint8_t*)journal.buffer + offset, len);
         return 1;
     }
     BS_SUBMIT_GET_SQE(sqe, data);
@@ -75,7 +75,7 @@ int blockstore_impl_t::fulfill_read(blockstore_op_t *read_op, uint64_t &fulfille
                 };
                 it = PRIV(read_op)->read_vec.insert(it, el);
                 if (!fulfill_read_push(read_op,
-                    read_op->buf + el.offset - read_op->offset,
+                    (uint8_t*)read_op->buf + el.offset - read_op->offset,
                     item_location + el.offset - item_start,
                     el.len, item_state, item_version))
                 {
@@ -102,7 +102,7 @@ uint8_t* blockstore_impl_t::get_clean_entry_bitmap(uint64_t block_loc, int offse
     {
         uint64_t sector = (meta_loc / (meta_block_size / clean_entry_size)) * meta_block_size;
         uint64_t pos = (meta_loc % (meta_block_size / clean_entry_size));
-        clean_entry_bitmap = (uint8_t*)(metadata_buffer + sector + pos*clean_entry_size + sizeof(clean_disk_entry) + offset);
+        clean_entry_bitmap = ((uint8_t*)metadata_buffer + sector + pos*clean_entry_size + sizeof(clean_disk_entry) + offset);
     }
     else
         clean_entry_bitmap = (uint8_t*)(clean_bitmap + meta_loc*2*clean_entry_bitmap_size + offset);
@@ -111,6 +111,7 @@ uint8_t* blockstore_impl_t::get_clean_entry_bitmap(uint64_t block_loc, int offse
 
 int blockstore_impl_t::dequeue_read(blockstore_op_t *read_op)
 {
+    auto & clean_db = clean_db_shard(read_op->oid);
     auto clean_it = clean_db.find(read_op->oid);
     auto dirty_it = dirty_db.upper_bound((obj_ver_id){
         .oid = read_op->oid,
@@ -297,6 +298,7 @@ int blockstore_impl_t::read_bitmap(object_id oid, uint64_t target_version, void
             dirty_it--;
         }
     }
+    auto & clean_db = clean_db_shard(oid);
     auto clean_it = clean_db.find(oid);
     if (clean_it != clean_db.end())
     {

src/blockstore_stable.cpp

@@ -54,6 +54,7 @@ int blockstore_impl_t::dequeue_stable(blockstore_op_t *op)
         auto dirty_it = dirty_db.find(*v);
         if (dirty_it == dirty_db.end())
         {
+            auto & clean_db = clean_db_shard(v->oid);
             auto clean_it = clean_db.find(v->oid);
             if (clean_it == clean_db.end() || clean_it->second.version < v->version)
             {
@@ -188,6 +189,7 @@ void blockstore_impl_t::mark_stable(const obj_ver_id & v, bool forget_dirty)
                     }
                     if (exists == -1)
                     {
+                        auto & clean_db = clean_db_shard(v.oid);
                         auto clean_it = clean_db.find(v.oid);
                         exists = clean_it != clean_db.end() ? 1 : 0;
                     }
@@ -215,6 +217,7 @@ void blockstore_impl_t::mark_stable(const obj_ver_id & v, bool forget_dirty)
                         break;
                     }
                 }
+                auto & clean_db = clean_db_shard(v.oid);
                 auto clean_it = clean_db.find(v.oid);
                 uint64_t clean_loc = clean_it != clean_db.end()
                     ? clean_it->second.location : UINT64_MAX;

src/blockstore_write.cpp

@@ -41,6 +41,7 @@ bool blockstore_impl_t::enqueue_write(blockstore_op_t *op)
     }
     if (!found)
     {
+        auto & clean_db = clean_db_shard(op->oid);
         auto clean_it = clean_db.find(op->oid);
         if (clean_it != clean_db.end())
         {
@@ -102,7 +103,7 @@ bool blockstore_impl_t::enqueue_write(blockstore_op_t *op)
         // Issue an additional sync so that the previous big write can reach 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)
+        sync_op->callback = [](blockstore_op_t *sync_op)
         {
             delete sync_op;
         };
@@ -380,7 +381,7 @@ int blockstore_impl_t::dequeue_write(blockstore_op_t *op)
             if (journal.inmemory)
             {
                 // Copy data
-                memcpy(journal.buffer + journal.next_free, op->buf, op->len);
+                memcpy((uint8_t*)journal.buffer + journal.next_free, op->buf, op->len);
             }
             BS_SUBMIT_GET_SQE(sqe2, data2);
             data2->iov = (struct iovec){ op->buf, op->len };
@@ -543,12 +544,13 @@ resume_4:
             if (ref_us > exec_us + throttle_threshold_us)
             {
                 // Pause reply
+                PRIV(op)->op_state = 5;
+                // Remember that the timer can in theory be called right here
                 tfd->set_timer_us(ref_us-exec_us, false, [this, op](int timer_id)
                 {
                     PRIV(op)->op_state++;
                     ringloop->wakeup();
                 });
-                PRIV(op)->op_state = 5;
                 return 1;
             }
         }
@@ -629,13 +631,17 @@ int blockstore_impl_t::dequeue_del(blockstore_op_t *op)
     {
         return 0;
     }
-    write_iodepth++;
     // Write current journal sector only if it's dirty and full, or in the immediate_commit mode
     BS_SUBMIT_CHECK_SQES(
         (immediate_commit != IMMEDIATE_NONE ||
             (journal_block_size - journal.in_sector_pos) < sizeof(journal_entry_del) &&
             journal.sector_info[journal.cur_sector].dirty) ? 1 : 0
     );
+    if (write_iodepth >= max_write_iodepth)
+    {
+        return 0;
+    }
+    write_iodepth++;
     // Prepare journal sector write
     if (immediate_commit == IMMEDIATE_NONE)
     {

src/cli.cpp

@@ -2,8 +2,7 @@
 // License: VNPL-1.1 (see README.md for details)
 
 /**
- * CLI tool
+ * CLI tool and also a library for administrative tasks
- * Currently can (a) remove inodes and (b) merge snapshot/clone layers
  */
 
 #include <vector>
@@ -17,7 +16,9 @@
 
 static const char *exe_name = NULL;
 
-json11::Json::object cli_tool_t::parse_args(int narg, const char *args[])
+static void help();
+
+static json11::Json::object parse_args(int narg, const char *args[])
 {
     json11::Json::object cfg;
     json11::Json::array cmd;
@@ -79,13 +80,16 @@ json11::Json::object cli_tool_t::parse_args(int narg, const char *args[])
     return cfg;
 }
 
-void cli_tool_t::help()
+static void help()
 {
     printf(
         "Vitastor command-line tool\n"
         "(c) Vitaliy Filippov, 2019+ (VNPL-1.1)\n"
         "\n"
         "USAGE:\n"
+        "%s status\n"
+        "  Show cluster status\n"
+        "\n"
         "%s df\n"
         "  Show pool space statistics\n"
         "\n"
@@ -155,200 +159,177 @@ void cli_tool_t::help()
         "  --no-color          Disable colored output\n"
         "  --json              JSON output\n"
         ,
-        exe_name, exe_name, exe_name, exe_name, exe_name, exe_name,
+        exe_name, exe_name, exe_name, exe_name, exe_name, exe_name, exe_name,
         exe_name, exe_name, exe_name, exe_name, exe_name, exe_name
     );
     exit(0);
 }
 
-void cli_tool_t::change_parent(inode_t cur, inode_t new_parent)
+static int run(cli_tool_t *p, json11::Json::object cfg)
-{
-    auto cur_cfg_it = cli->st_cli.inode_config.find(cur);
-    if (cur_cfg_it == cli->st_cli.inode_config.end())
-    {
-        fprintf(stderr, "Inode 0x%lx disappeared\n", cur);
-        exit(1);
-    }
-    inode_config_t new_cfg = cur_cfg_it->second;
-    std::string cur_name = new_cfg.name;
-    std::string cur_cfg_key = base64_encode(cli->st_cli.etcd_prefix+
-        "/config/inode/"+std::to_string(INODE_POOL(cur))+
-        "/"+std::to_string(INODE_NO_POOL(cur)));
-    new_cfg.parent_id = new_parent;
-    json11::Json::object cur_cfg_json = cli->st_cli.serialize_inode_cfg(&new_cfg);
-    waiting++;
-    cli->st_cli.etcd_txn(json11::Json::object {
-        { "compare", json11::Json::array {
-            json11::Json::object {
-                { "target", "MOD" },
-                { "key", cur_cfg_key },
-                { "result", "LESS" },
-                { "mod_revision", new_cfg.mod_revision+1 },
-            },
-        } },
-        { "success", json11::Json::array {
-            json11::Json::object {
-                { "request_put", json11::Json::object {
-                    { "key", cur_cfg_key },
-                    { "value", base64_encode(json11::Json(cur_cfg_json).dump()) },
-                } }
-            },
-        } },
-    }, ETCD_SLOW_TIMEOUT, [this, new_parent, cur, cur_name](std::string err, json11::Json res)
-    {
-        if (err != "")
-        {
-            fprintf(stderr, "Error changing parent of %s: %s\n", cur_name.c_str(), err.c_str());
-            exit(1);
-        }
-        if (!res["succeeded"].bool_value())
-        {
-            fprintf(stderr, "Inode %s was modified during snapshot deletion\n", cur_name.c_str());
-            exit(1);
-        }
-        if (new_parent)
-        {
-            auto new_parent_it = cli->st_cli.inode_config.find(new_parent);
-            std::string new_parent_name = new_parent_it != cli->st_cli.inode_config.end()
-                ? new_parent_it->second.name : "<unknown>";
-            printf(
-                "Parent of layer %s (inode %lu in pool %u) changed to %s (inode %lu in pool %u)\n",
-                cur_name.c_str(), INODE_NO_POOL(cur), INODE_POOL(cur),
-                new_parent_name.c_str(), INODE_NO_POOL(new_parent), INODE_POOL(new_parent)
-            );
-        }
-        else
-        {
-            printf(
-                "Parent of layer %s (inode %lu in pool %u) detached\n",
-                cur_name.c_str(), INODE_NO_POOL(cur), INODE_POOL(cur)
-            );
-        }
-        waiting--;
-        ringloop->wakeup();
-    });
-}
-
-inode_config_t* cli_tool_t::get_inode_cfg(const std::string & name)
-{
-    for (auto & ic: cli->st_cli.inode_config)
-    {
-        if (ic.second.name == name)
-        {
-            return &ic.second;
-        }
-    }
-    fprintf(stderr, "Layer %s not found\n", name.c_str());
-    exit(1);
-}
-
-void cli_tool_t::run(json11::Json cfg)
 {
+    cli_result_t result;
+    p->parse_config(cfg);
     json11::Json::array cmd = cfg["command"].array_items();
+    cfg.erase("command");
+    std::function<bool(cli_result_t &)> action_cb;
     if (!cmd.size())
     {
-        fprintf(stderr, "command is missing\n");
+        result = { .err = EINVAL, .text = "command is missing" };
-        exit(1);
+    }
+    else if (cmd[0] == "status")
+    {
+        // Show cluster status
+        action_cb = p->start_status(cfg);
     }
     else if (cmd[0] == "df")
     {
         // Show pool space stats
-        action_cb = start_df(cfg);
+        action_cb = p->start_df(cfg);
     }
     else if (cmd[0] == "ls")
     {
         // List images
-        action_cb = start_ls(cfg);
+        if (cmd.size() > 1)
+        {
+            cmd.erase(cmd.begin(), cmd.begin()+1);
+            cfg["names"] = cmd;
+        }
+        action_cb = p->start_ls(cfg);
     }
-    else if (cmd[0] == "create" || cmd[0] == "snap-create")
+    else if (cmd[0] == "snap-create")
+    {
+        // Create snapshot
+        std::string name = cmd.size() > 1 ? cmd[1].string_value() : "";
+        int pos = name.find('@');
+        if (pos == std::string::npos || pos == name.length()-1)
+        {
+            result = (cli_result_t){ .err = EINVAL, .text = "Please specify new snapshot name after @" };
+        }
+        else
+        {
+            cfg["image"] = name.substr(0, pos);
+            cfg["snapshot"] = name.substr(pos + 1);
+            action_cb = p->start_create(cfg);
+        }
+    }
+    else if (cmd[0] == "create")
     {
         // Create image/snapshot
-        action_cb = start_create(cfg);
+        if (cmd.size() > 1)
+        {
+            cfg["image"] = cmd[1];
+        }
+        action_cb = p->start_create(cfg);
     }
     else if (cmd[0] == "modify")
     {
         // Modify image
-        action_cb = start_modify(cfg);
+        if (cmd.size() > 1)
+        {
+            cfg["image"] = cmd[1];
+        }
+        action_cb = p->start_modify(cfg);
     }
     else if (cmd[0] == "rm-data")
     {
         // Delete inode data
-        action_cb = start_rm(cfg);
+        action_cb = p->start_rm_data(cfg);
     }
     else if (cmd[0] == "merge-data")
     {
         // Merge layer data without affecting metadata
-        action_cb = start_merge(cfg);
+        if (cmd.size() > 1)
+        {
+            cfg["from"] = cmd[1];
+            if (cmd.size() > 2)
+                cfg["to"] = cmd[2];
+        }
+        action_cb = p->start_merge(cfg);
     }
     else if (cmd[0] == "flatten")
     {
         // Merge layer data without affecting metadata
-        action_cb = start_flatten(cfg);
+        if (cmd.size() > 1)
+        {
+            cfg["image"] = cmd[1];
+        }
+        action_cb = p->start_flatten(cfg);
     }
     else if (cmd[0] == "rm")
     {
         // Remove multiple snapshots and rebase their children
-        action_cb = start_snap_rm(cfg);
+        if (cmd.size() > 1)
+        {
+            cfg["from"] = cmd[1];
+            if (cmd.size() > 2)
+                cfg["to"] = cmd[2];
+        }
+        action_cb = p->start_rm(cfg);
     }
     else if (cmd[0] == "alloc-osd")
     {
         // Allocate a new OSD number
-        action_cb = start_alloc_osd(cfg);
+        action_cb = p->start_alloc_osd(cfg);
     }
     else if (cmd[0] == "simple-offsets")
     {
         // Calculate offsets for simple & stupid OSD deployment without superblock
-        action_cb = simple_offsets(cfg);
+        if (cmd.size() > 1)
+        {
+            cfg["device"] = cmd[1];
+        }
+        action_cb = p->simple_offsets(cfg);
     }
     else
     {
-        fprintf(stderr, "unknown command: %s\n", cmd[0].string_value().c_str());
+        result = { .err = EINVAL, .text = "unknown command: "+cmd[0].string_value() };
-        exit(1);
     }
-    if (action_cb == NULL)
+    if (action_cb != NULL)
     {
-        return;
+        // Create client
-    }
+        json11::Json cfg_j = cfg;
-    color = !cfg["no-color"].bool_value();
+        p->ringloop = new ring_loop_t(512);
-    json_output = cfg["json"].bool_value();
+        p->epmgr = new epoll_manager_t(p->ringloop);
-    iodepth = cfg["iodepth"].uint64_value();
+        p->cli = new cluster_client_t(p->ringloop, p->epmgr->tfd, cfg_j);
-    if (!iodepth)
+        // Smaller timeout by default for more interactiveness
-        iodepth = 32;
+        p->cli->st_cli.etcd_slow_timeout = p->cli->st_cli.etcd_quick_timeout;
-    parallel_osds = cfg["parallel_osds"].uint64_value();
+        p->loop_and_wait(action_cb, [&](const cli_result_t & r)
-    if (!parallel_osds)
-        parallel_osds = 4;
-    log_level = cfg["log_level"].int64_value();
-    progress = cfg["progress"].uint64_value() ? true : false;
-    list_first = cfg["wait-list"].uint64_value() ? true : false;
-    // Create client
-    ringloop = new ring_loop_t(512);
-    epmgr = new epoll_manager_t(ringloop);
-    cli = new cluster_client_t(ringloop, epmgr->tfd, cfg);
-    cli->on_ready([this]()
-    {
-        // Initialize job
-        consumer.loop = [this]()
         {
+            result = r;
+            action_cb = NULL;
+        });
+        // Loop until it completes
+        while (action_cb != NULL)
+        {
+            p->ringloop->loop();
             if (action_cb != NULL)
-            {
+                p->ringloop->wait();
-                bool done = action_cb();
+        }
-                if (done)
+        // Destroy the client
-                {
+        delete p->cli;
-                    action_cb = NULL;
+        delete p->epmgr;
-                }
+        delete p->ringloop;
-            }
+        p->cli = NULL;
-            ringloop->submit();
+        p->epmgr = NULL;
-        };
+        p->ringloop = NULL;
-        ringloop->register_consumer(&consumer);
-        consumer.loop();
-    });
-    // Loop until it completes
-    while (action_cb != NULL)
-    {
-        ringloop->loop();
-        if (action_cb != NULL)
-            ringloop->wait();
     }
+    // Print result
+    if (p->json_output && !result.data.is_null())
+    {
+        printf("%s\n", result.data.dump().c_str());
+    }
+    else if (p->json_output && result.err)
+    {
+        printf("%s\n", json11::Json(json11::Json::object {
+            { "error_code", result.err },
+            { "error_text", result.text },
+        }).dump().c_str());
+    }
+    else if (result.text != "")
+    {
+        fprintf(result.err ? stderr : stdout, result.text[result.text.size()-1] == '\n' ? "%s" : "%s\n", result.text.c_str());
+    }
+    return result.err;
 }
 
 int main(int narg, const char *args[])
@@ -357,6 +338,7 @@ int main(int narg, const char *args[])
     setvbuf(stderr, NULL, _IONBF, 0);
     exe_name = args[0];
     cli_tool_t *p = new cli_tool_t();
-    p->run(cli_tool_t::parse_args(narg, args));
+    int r = run(p, parse_args(narg, args));
-    return 0;
+    delete p;
+    return r;
 }

src/cli.h

@@ -19,11 +19,18 @@ class epoll_manager_t;
 class cluster_client_t;
 struct inode_config_t;
 
+struct cli_result_t
+{
+    int err;
+    std::string text;
+    json11::Json data;
+};
+
 class cli_tool_t
 {
 public:
-    uint64_t iodepth = 0, parallel_osds = 0;
+    uint64_t iodepth = 4, parallel_osds = 32;
-    bool progress = true;
+    bool progress = false;
     bool list_first = false;
     bool json_output = false;
     int log_level = 0;
@@ -34,39 +41,42 @@ public:
     cluster_client_t *cli = NULL;
 
     int waiting = 0;
-    ring_consumer_t consumer;
+    cli_result_t etcd_err;
-    std::function<bool(void)> action_cb;
+    json11::Json etcd_result;
 
-    void run(json11::Json cfg);
+    void parse_config(json11::Json cfg);
 
-    void change_parent(inode_t cur, inode_t new_parent);
+    void change_parent(inode_t cur, inode_t new_parent, cli_result_t *result);
     inode_config_t* get_inode_cfg(const std::string & name);
 
-    static json11::Json::object parse_args(int narg, const char *args[]);
-    static void help();
-
     friend struct rm_inode_t;
     friend struct snap_merger_t;
     friend struct snap_flattener_t;
     friend struct snap_remover_t;
 
-    std::function<bool(void)> start_df(json11::Json);
+    std::function<bool(cli_result_t &)> start_status(json11::Json);
-    std::function<bool(void)> start_ls(json11::Json);
+    std::function<bool(cli_result_t &)> start_df(json11::Json);
-    std::function<bool(void)> start_create(json11::Json);
+    std::function<bool(cli_result_t &)> start_ls(json11::Json);
-    std::function<bool(void)> start_modify(json11::Json);
+    std::function<bool(cli_result_t &)> start_create(json11::Json);
-    std::function<bool(void)> start_rm(json11::Json);
+    std::function<bool(cli_result_t &)> start_modify(json11::Json);
-    std::function<bool(void)> start_merge(json11::Json);
+    std::function<bool(cli_result_t &)> start_rm_data(json11::Json);
-    std::function<bool(void)> start_flatten(json11::Json);
+    std::function<bool(cli_result_t &)> start_merge(json11::Json);
-    std::function<bool(void)> start_snap_rm(json11::Json);
+    std::function<bool(cli_result_t &)> start_flatten(json11::Json);
-    std::function<bool(void)> start_alloc_osd(json11::Json cfg, uint64_t *out = NULL);
+    std::function<bool(cli_result_t &)> start_rm(json11::Json);
-    std::function<bool(void)> simple_offsets(json11::Json cfg);
+    std::function<bool(cli_result_t &)> start_alloc_osd(json11::Json cfg);
+    std::function<bool(cli_result_t &)> simple_offsets(json11::Json cfg);
+
+    // Should be called like loop_and_wait(start_status(), <completion callback>)
+    void loop_and_wait(std::function<bool(cli_result_t &)> loop_cb, std::function<void(const cli_result_t &)> complete_cb);
+
+    void etcd_txn(json11::Json txn);
 };
 
 uint64_t parse_size(std::string size_str);
 
 std::string print_table(json11::Json items, json11::Json header, bool use_esc);
 
-std::string format_size(uint64_t size);
+std::string format_size(uint64_t size, bool nobytes = false);
 
 std::string format_lat(uint64_t lat);
 

src/cli_alloc_osd.cpp

@@ -13,10 +13,10 @@ struct alloc_osd_t
 {
     cli_tool_t *parent;
 
-    json11::Json result;
     uint64_t new_id = 1;
 
     int state = 0;
+    cli_result_t result;
 
     bool is_done()
     {
@@ -29,7 +29,7 @@ struct alloc_osd_t
             goto resume_1;
         do
         {
-            etcd_txn(json11::Json::object {
+            parent->etcd_txn(json11::Json::object {
                 { "compare", json11::Json::array {
                     json11::Json::object {
                         { "target", "VERSION" },
@@ -63,10 +63,16 @@ struct alloc_osd_t
             state = 1;
             if (parent->waiting > 0)
                 return;
-            if (!result["succeeded"].bool_value())
+            if (parent->etcd_err.err)
+            {
+                result = parent->etcd_err;
+                state = 100;
+                return;
+            }
+            if (!parent->etcd_result["succeeded"].bool_value())
             {
                 std::vector<osd_num_t> used;
-                for (auto kv: result["responses"][0]["response_range"]["kvs"].array_items())
+                for (auto kv: parent->etcd_result["responses"][0]["response_range"]["kvs"].array_items())
                 {
                     std::string key = base64_decode(kv["key"].string_value());
                     osd_num_t cur_osd;
@@ -98,41 +104,25 @@ struct alloc_osd_t
                     new_id = used[e-1]+1;
                 }
             }
-        } while (!result["succeeded"].bool_value());
+        } while (!parent->etcd_result["succeeded"].bool_value());
         state = 100;
-    }
+        result = (cli_result_t){
-
+            .text = std::to_string(new_id),
-    void etcd_txn(json11::Json txn)
+            .data = json11::Json(new_id),
-    {
+        };
-        parent->waiting++;
-        parent->cli->st_cli.etcd_txn(txn, ETCD_SLOW_TIMEOUT, [this](std::string err, json11::Json res)
-        {
-            parent->waiting--;
-            if (err != "")
-            {
-                fprintf(stderr, "Error reading from etcd: %s\n", err.c_str());
-                exit(1);
-            }
-            this->result = res;
-            parent->ringloop->wakeup();
-        });
     }
 };
 
-std::function<bool(void)> cli_tool_t::start_alloc_osd(json11::Json cfg, uint64_t *out)
+std::function<bool(cli_result_t &)> cli_tool_t::start_alloc_osd(json11::Json cfg)
 {
-    json11::Json::array cmd = cfg["command"].array_items();
     auto alloc_osd = new alloc_osd_t();
     alloc_osd->parent = this;
-    return [alloc_osd, out]()
+    return [alloc_osd](cli_result_t & result)
     {
         alloc_osd->loop();
         if (alloc_osd->is_done())
         {
-            if (out)
+            result = alloc_osd->result;
-                *out = alloc_osd->new_id;
-            else if (alloc_osd->new_id)
-                printf("%lu\n", alloc_osd->new_id);
             delete alloc_osd;
             return true;
         }

src/cli_common.cpp

@@ -0,0 +1,149 @@
+// Copyright (c) Vitaliy Filippov, 2019+
+// License: VNPL-1.1 (see README.md for details)
+
+#include "base64.h"
+#include "cluster_client.h"
+#include "cli.h"
+
+void cli_tool_t::change_parent(inode_t cur, inode_t new_parent, cli_result_t *result)
+{
+    auto cur_cfg_it = cli->st_cli.inode_config.find(cur);
+    if (cur_cfg_it == cli->st_cli.inode_config.end())
+    {
+        char buf[128];
+        snprintf(buf, 128, "Inode 0x%lx disappeared", cur);
+        *result = (cli_result_t){ .err = EIO, .text = buf };
+        return;
+    }
+    inode_config_t new_cfg = cur_cfg_it->second;
+    std::string cur_name = new_cfg.name;
+    std::string cur_cfg_key = base64_encode(cli->st_cli.etcd_prefix+
+        "/config/inode/"+std::to_string(INODE_POOL(cur))+
+        "/"+std::to_string(INODE_NO_POOL(cur)));
+    new_cfg.parent_id = new_parent;
+    json11::Json::object cur_cfg_json = cli->st_cli.serialize_inode_cfg(&new_cfg);
+    waiting++;
+    cli->st_cli.etcd_txn_slow(json11::Json::object {
+        { "compare", json11::Json::array {
+            json11::Json::object {
+                { "target", "MOD" },
+                { "key", cur_cfg_key },
+                { "result", "LESS" },
+                { "mod_revision", new_cfg.mod_revision+1 },
+            },
+        } },
+        { "success", json11::Json::array {
+            json11::Json::object {
+                { "request_put", json11::Json::object {
+                    { "key", cur_cfg_key },
+                    { "value", base64_encode(json11::Json(cur_cfg_json).dump()) },
+                } }
+            },
+        } },
+    }, [this, result, new_parent, cur, cur_name](std::string err, json11::Json res)
+    {
+        if (err != "")
+        {
+            *result = (cli_result_t){ .err = EIO, .text = "Error changing parent of "+cur_name+": "+err };
+        }
+        else if (!res["succeeded"].bool_value())
+        {
+            *result = (cli_result_t){ .err = EAGAIN, .text = "Image "+cur_name+" was modified during change" };
+        }
+        else if (new_parent)
+        {
+            auto new_parent_it = cli->st_cli.inode_config.find(new_parent);
+            std::string new_parent_name = new_parent_it != cli->st_cli.inode_config.end()
+                ? new_parent_it->second.name : "<unknown>";
+            *result = (cli_result_t){
+                .text = "Parent of layer "+cur_name+" (inode "+std::to_string(INODE_NO_POOL(cur))+
+                    " in pool "+std::to_string(INODE_POOL(cur))+") changed to "+new_parent_name+
+                    " (inode "+std::to_string(INODE_NO_POOL(new_parent))+" in pool "+std::to_string(INODE_POOL(new_parent))+")",
+            };
+        }
+        else
+        {
+            *result = (cli_result_t){
+                .text = "Parent of layer "+cur_name+" (inode "+std::to_string(INODE_NO_POOL(cur))+
+                    " in pool "+std::to_string(INODE_POOL(cur))+") detached",
+            };
+        }
+        waiting--;
+        ringloop->wakeup();
+    });
+}
+
+void cli_tool_t::etcd_txn(json11::Json txn)
+{
+    waiting++;
+    cli->st_cli.etcd_txn_slow(txn, [this](std::string err, json11::Json res)
+    {
+        waiting--;
+        if (err != "")
+            etcd_err = (cli_result_t){ .err = EIO, .text = "Error communicating with etcd: "+err };
+        else
+            etcd_err = (cli_result_t){ .err = 0 };
+        etcd_result = res;
+        ringloop->wakeup();
+    });
+}
+
+inode_config_t* cli_tool_t::get_inode_cfg(const std::string & name)
+{
+    for (auto & ic: cli->st_cli.inode_config)
+    {
+        if (ic.second.name == name)
+        {
+            return &ic.second;
+        }
+    }
+    return NULL;
+}
+
+void cli_tool_t::parse_config(json11::Json cfg)
+{
+    color = !cfg["no-color"].bool_value();
+    json_output = cfg["json"].bool_value();
+    iodepth = cfg["iodepth"].uint64_value();
+    if (!iodepth)
+        iodepth = 32;
+    parallel_osds = cfg["parallel_osds"].uint64_value();
+    if (!parallel_osds)
+        parallel_osds = 4;
+    log_level = cfg["log_level"].int64_value();
+    progress = cfg["progress"].uint64_value() ? true : false;
+    list_first = cfg["wait-list"].uint64_value() ? true : false;
+}
+
+struct cli_result_looper_t
+{
+    ring_consumer_t consumer;
+    cli_result_t result;
+    std::function<bool(cli_result_t &)> loop_cb;
+    std::function<void(const cli_result_t &)> complete_cb;
+};
+
+void cli_tool_t::loop_and_wait(std::function<bool(cli_result_t &)> loop_cb, std::function<void(const cli_result_t &)> complete_cb)
+{
+    auto *looper = new cli_result_looper_t();
+    looper->loop_cb = loop_cb;
+    looper->complete_cb = complete_cb;
+    looper->consumer.loop = [this, looper]()
+    {
+        bool done = looper->loop_cb(looper->result);
+        if (done)
+        {
+            ringloop->unregister_consumer(&looper->consumer);
+            looper->loop_cb = NULL;
+            looper->complete_cb(looper->result);
+            delete looper;
+            return;
+        }
+        ringloop->submit();
+    };
+    cli->on_ready([this, looper]()
+    {
+        ringloop->register_consumer(&looper->consumer);
+        ringloop->wakeup();
+    });
+}

src/cli_create.cpp

@@ -25,15 +25,18 @@ struct image_creator_t
     pool_id_t new_pool_id = 0;
     std::string new_pool_name;
     std::string image_name, new_snap, new_parent;
+    json11::Json new_meta;
     uint64_t size;
+    bool force_size = false;
 
     pool_id_t old_pool_id = 0;
     inode_t new_parent_id = 0;
     inode_t new_id = 0, old_id = 0;
     uint64_t max_id_mod_rev = 0, cfg_mod_rev = 0, idx_mod_rev = 0;
-    json11::Json result;
+    inode_config_t new_cfg;
 
     int state = 0;
+    cli_result_t result;
 
     bool is_done()
     {
@@ -44,13 +47,27 @@ struct image_creator_t
     {
         if (state >= 1)
             goto resume_1;
+        if (image_name == "")
+        {
+            // FIXME: EINVAL -> specific codes for every error
+            result = (cli_result_t){ .err = EINVAL, .text = "Image name is missing" };
+            state = 100;
+            return;
+        }
+        if (image_name.find('@') != std::string::npos)
+        {
+            result = (cli_result_t){ .err = EINVAL, .text = "Image name can't contain @ character" };
+            state = 100;
+            return;
+        }
         if (new_pool_id)
         {
             auto & pools = parent->cli->st_cli.pool_config;
             if (pools.find(new_pool_id) == pools.end())
             {
-                fprintf(stderr, "Pool %u does not exist\n", new_pool_id);
+                result = (cli_result_t){ .err = ENOENT, .text = "Pool "+std::to_string(new_pool_id)+" does not exist" };
-                exit(1);
+                state = 100;
+                return;
             }
         }
         else if (new_pool_name != "")
@@ -65,8 +82,9 @@ struct image_creator_t
             }
             if (!new_pool_id)
             {
-                fprintf(stderr, "Pool %s does not exist\n", new_pool_name.c_str());
+                result = (cli_result_t){ .err = ENOENT, .text = "Pool "+new_pool_name+" does not exist" };
-                exit(1);
+                state = 100;
+                return;
             }
         }
         else if (parent->cli->st_cli.pool_config.size() == 1)
@@ -92,8 +110,9 @@ struct image_creator_t
         {
             if (ic.second.name == image_name)
             {
-                fprintf(stderr, "Image %s already exists\n", image_name.c_str());
+                result = (cli_result_t){ .err = EEXIST, .text = "Image "+image_name+" already exists" };
-                exit(1);
+                state = 100;
+                return;
             }
             if (ic.second.name == new_parent)
             {
@@ -110,45 +129,61 @@ struct image_creator_t
         }
         if (new_parent != "" && !new_parent_id)
         {
-            fprintf(stderr, "Parent image not found\n");
+            result = (cli_result_t){ .err = ENOENT, .text = "Parent image "+new_parent+" not found" };
-            exit(1);
+            state = 100;
+            return;
         }
         if (!new_pool_id)
         {
-            fprintf(stderr, "Pool name or ID is missing\n");
+            result = (cli_result_t){ .err = EINVAL, .text = "Pool name or ID is missing" };
-            exit(1);
+            state = 100;
+            return;
         }
-        if (!size)
+        if (!size && !force_size)
         {
-            fprintf(stderr, "Image size is missing\n");
+            result = (cli_result_t){ .err = EINVAL, .text = "Image size is missing" };
-            exit(1);
+            state = 100;
+            return;
         }
         do
         {
-            etcd_txn(json11::Json::object {
+            parent->etcd_txn(json11::Json::object {
                 { "success", json11::Json::array { get_next_id() } }
             });
             state = 2;
 resume_2:
             if (parent->waiting > 0)
                 return;
-            extract_next_id(result["responses"][0]);
+            if (parent->etcd_err.err)
+            {
+                result = parent->etcd_err;
+                state = 100;
+                return;
+            }
+            extract_next_id(parent->etcd_result["responses"][0]);
             attempt_create();
             state = 3;
 resume_3:
             if (parent->waiting > 0)
                 return;
-            if (!result["succeeded"].bool_value() &&
+            if (parent->etcd_err.err)
-                result["responses"][0]["response_range"]["kvs"].array_items().size() > 0)
             {
-                fprintf(stderr, "Image %s already exists\n", image_name.c_str());
+                result = parent->etcd_err;
-                exit(1);
+                state = 100;
+                return;
             }
-        } while (!result["succeeded"].bool_value());
+            if (!parent->etcd_result["succeeded"].bool_value() &&
-        if (parent->progress)
+                parent->etcd_result["responses"][0]["response_range"]["kvs"].array_items().size() > 0)
-        {
+            {
-            printf("Image %s created\n", image_name.c_str());
+                result = (cli_result_t){ .err = EEXIST, .text = "Image "+image_name+" already exists" };
-        }
+                state = 100;
+                return;
+            }
+        } while (!parent->etcd_result["succeeded"].bool_value());
+        // Save into inode_config for library users to be able to take it from there immediately
+        new_cfg.mod_revision = parent->etcd_result["responses"][0]["response_put"]["header"]["revision"].uint64_value();
+        parent->cli->st_cli.insert_inode_config(new_cfg);
+        result = (cli_result_t){ .err = 0, .text = "Image "+image_name+" created" };
         state = 100;
     }
 
@@ -164,14 +199,16 @@ resume_3:
         {
             if (ic.second.name == image_name+"@"+new_snap)
             {
-                fprintf(stderr, "Snapshot %s@%s already exists\n", image_name.c_str(), new_snap.c_str());
+                result = (cli_result_t){ .err = EEXIST, .text = "Snapshot "+image_name+"@"+new_snap+" already exists" };
-                exit(1);
+                state = 100;
+                return;
             }
         }
         if (new_parent != "")
         {
-            fprintf(stderr, "--parent can't be used with snapshots\n");
+            result = (cli_result_t){ .err = EINVAL, .text = "Parent can't be specified for snapshots" };
-            exit(1);
+            state = 100;
+            return;
         }
         do
         {
@@ -183,8 +220,9 @@ resume_3:
                 return;
             if (!old_id)
             {
-                fprintf(stderr, "Image %s does not exist\n", image_name.c_str());
+                result = (cli_result_t){ .err = ENOENT, .text = "Image "+image_name+" does not exist" };
-                exit(1);
+                state = 100;
+                return;
             }
             if (!new_pool_id)
             {
@@ -196,17 +234,24 @@ resume_3:
 resume_4:
             if (parent->waiting > 0)
                 return;
-            if (!result["succeeded"].bool_value() &&
+            if (parent->etcd_err.err)
-                result["responses"][0]["response_range"]["kvs"].array_items().size() > 0)
             {
-                fprintf(stderr, "Snapshot %s@%s already exists\n", image_name.c_str(), new_snap.c_str());
+                result = parent->etcd_err;
-                exit(1);
+                state = 100;
+                return;
             }
-        } while (!result["succeeded"].bool_value());
+            if (!parent->etcd_result["succeeded"].bool_value() &&
-        if (parent->progress)
+                parent->etcd_result["responses"][0]["response_range"]["kvs"].array_items().size() > 0)
-        {
+            {
-            printf("Snapshot %s@%s created\n", image_name.c_str(), new_snap.c_str());
+                result = (cli_result_t){ .err = EEXIST, .text = "Snapshot "+image_name+"@"+new_snap+" already exists" };
-        }
+                state = 100;
+                return;
+            }
+        } while (!parent->etcd_result["succeeded"].bool_value());
+        // Save into inode_config for library users to be able to take it from there immediately
+        new_cfg.mod_revision = parent->etcd_result["responses"][0]["response_put"]["header"]["revision"].uint64_value();
+        parent->cli->st_cli.insert_inode_config(new_cfg);
+        result = (cli_result_t){ .err = 0, .text = "Snapshot "+image_name+"@"+new_snap+" created" };
         state = 100;
     }
 
@@ -246,7 +291,7 @@ resume_4:
             goto resume_2;
         else if (state == 3)
             goto resume_3;
-        etcd_txn(json11::Json::object { { "success", json11::Json::array {
+        parent->etcd_txn(json11::Json::object { { "success", json11::Json::array {
             get_next_id(),
             json11::Json::object {
                 { "request_range", json11::Json::object {
@@ -260,11 +305,17 @@ resume_4:
 resume_2:
         if (parent->waiting > 0)
             return;
-        extract_next_id(result["responses"][0]);
+        if (parent->etcd_err.err)
+        {
+            result = parent->etcd_err;
+            state = 100;
+            return;
+        }
+        extract_next_id(parent->etcd_result["responses"][0]);
         old_id = 0;
         old_pool_id = 0;
         cfg_mod_rev = idx_mod_rev = 0;
-        if (result["responses"][1]["response_range"]["kvs"].array_items().size() == 0)
+        if (parent->etcd_result["responses"][1]["response_range"]["kvs"].array_items().size() == 0)
         {
             for (auto & ic: parent->cli->st_cli.inode_config)
             {
@@ -283,17 +334,18 @@ resume_2:
         {
             // FIXME: Parse kvs in etcd_state_client automatically
             {
-                auto kv = parent->cli->st_cli.parse_etcd_kv(result["responses"][1]["response_range"]["kvs"][0]);
+                auto kv = parent->cli->st_cli.parse_etcd_kv(parent->etcd_result["responses"][1]["response_range"]["kvs"][0]);
                 old_id = INODE_NO_POOL(kv.value["id"].uint64_value());
                 old_pool_id = (pool_id_t)kv.value["pool_id"].uint64_value();
                 idx_mod_rev = kv.mod_revision;
                 if (!old_id || !old_pool_id || old_pool_id >= POOL_ID_MAX)
                 {
-                    fprintf(stderr, "Invalid pool or inode ID in etcd key %s\n", kv.key.c_str());
+                    result = (cli_result_t){ .err = ENOENT, .text = "Invalid pool or inode ID in etcd key "+kv.key };
-                    exit(1);
+                    state = 100;
+                    return;
                 }
             }
-            etcd_txn(json11::Json::object {
+            parent->etcd_txn(json11::Json::object {
                 { "success", json11::Json::array {
                     json11::Json::object {
                         { "request_range", json11::Json::object {
@@ -309,8 +361,14 @@ resume_2:
 resume_3:
             if (parent->waiting > 0)
                 return;
+            if (parent->etcd_err.err)
             {
-                auto kv = parent->cli->st_cli.parse_etcd_kv(result["responses"][0]["response_range"]["kvs"][0]);
+                result = parent->etcd_err;
+                state = 100;
+                return;
+            }
+            {
+                auto kv = parent->cli->st_cli.parse_etcd_kv(parent->etcd_result["responses"][0]["response_range"]["kvs"][0]);
                 size = kv.value["size"].uint64_value();
                 new_parent_id = kv.value["parent_id"].uint64_value();
                 uint64_t parent_pool_id = kv.value["parent_pool_id"].uint64_value();
@@ -325,12 +383,13 @@ resume_3:
 
     void attempt_create()
     {
-        inode_config_t new_cfg = {
+        new_cfg = {
             .num = INODE_WITH_POOL(new_pool_id, new_id),
             .name = image_name,
             .size = size,
             .parent_id = (new_snap != "" ? INODE_WITH_POOL(old_pool_id, old_id) : new_parent_id),
             .readonly = false,
+            .meta = new_meta,
         };
         json11::Json::array checks = json11::Json::array {
             json11::Json::object {
@@ -439,28 +498,12 @@ resume_3:
                 } },
             });
         };
-        etcd_txn(json11::Json::object {
+        parent->etcd_txn(json11::Json::object {
             { "compare", checks },
             { "success", success },
             { "failure", failure },
         });
     }
-
-    void etcd_txn(json11::Json txn)
-    {
-        parent->waiting++;
-        parent->cli->st_cli.etcd_txn(txn, ETCD_SLOW_TIMEOUT, [this](std::string err, json11::Json res)
-        {
-            parent->waiting--;
-            if (err != "")
-            {
-                fprintf(stderr, "Error reading from etcd: %s\n", err.c_str());
-                exit(1);
-            }
-            this->result = res;
-            parent->ringloop->wakeup();
-        });
-    }
 };
 
 uint64_t parse_size(std::string size_str)
@@ -474,77 +517,76 @@ uint64_t parse_size(std::string size_str)
     if (type_char == 'k' || type_char == 'm' || type_char == 'g' || type_char == 't')
     {
         if (type_char == 'k')
-            mul = 1l<<10;
+            mul = (uint64_t)1<<10;
         else if (type_char == 'm')
-            mul = 1l<<20;
+            mul = (uint64_t)1<<20;
         else if (type_char == 'g')
-            mul = 1l<<30;
+            mul = (uint64_t)1<<30;
         else /*if (type_char == 't')*/
-            mul = 1l<<40;
+            mul = (uint64_t)1<<40;
         size_str = size_str.substr(0, size_str.length()-1);
     }
     uint64_t size = json11::Json(size_str).uint64_value() * mul;
     if (size == 0 && size_str != "0" && (size_str != "" || mul != 1))
     {
-        fprintf(stderr, "Invalid syntax for size: %s\n", size_str.c_str());
+        return UINT64_MAX;
-        exit(1);
     }
     return size;
 }
 
-std::function<bool(void)> cli_tool_t::start_create(json11::Json cfg)
+std::function<bool(cli_result_t &)> cli_tool_t::start_create(json11::Json cfg)
 {
-    json11::Json::array cmd = cfg["command"].array_items();
     auto image_creator = new image_creator_t();
     image_creator->parent = this;
-    image_creator->image_name = cmd.size() > 1 ? cmd[1].string_value() : "";
+    image_creator->image_name = cfg["image"].string_value();
     image_creator->new_pool_id = cfg["pool"].uint64_value();
     image_creator->new_pool_name = cfg["pool"].string_value();
+    image_creator->force_size = cfg["force_size"].bool_value();
+    if (cfg["image_meta"].is_object())
+    {
+        image_creator->new_meta = cfg["image-meta"];
+    }
     if (cfg["snapshot"].string_value() != "")
     {
         image_creator->new_snap = cfg["snapshot"].string_value();
     }
-    else if (cmd[0] == "snap-create")
-    {
-        int p = image_creator->image_name.find('@');
-        if (p == std::string::npos || p == image_creator->image_name.length()-1)
-        {
-            fprintf(stderr, "Please specify new snapshot name after @\n");
-            exit(1);
-        }
-        image_creator->new_snap = image_creator->image_name.substr(p + 1);
-        image_creator->image_name = image_creator->image_name.substr(0, p);
-    }
     image_creator->new_parent = cfg["parent"].string_value();
     if (cfg["size"].string_value() != "")
     {
         image_creator->size = parse_size(cfg["size"].string_value());
-        if (image_creator->size % 4096)
+        if (image_creator->size == UINT64_MAX)
         {
-            fprintf(stderr, "Size should be a multiple of 4096\n");
+            return [size = cfg["size"].string_value()](cli_result_t & result)
-            exit(1);
+            {
+                result = (cli_result_t){ .err = EINVAL, .text = "Invalid syntax for size: "+size };
+                return true;
+            };
+        }
+        if ((image_creator->size % 4096) && !cfg["force_size"].bool_value())
+        {
+            delete image_creator;
+            return [](cli_result_t & result)
+            {
+                result = (cli_result_t){ .err = EINVAL, .text = "Size should be a multiple of 4096" };
+                return true;
+            };
         }
         if (image_creator->new_snap != "")
         {
-            fprintf(stderr, "--size can't be specified for snapshots\n");
+            delete image_creator;
-            exit(1);
+            return [](cli_result_t & result)
+            {
+                result = (cli_result_t){ .err = EINVAL, .text = "Size can't be specified for snapshots" };
+                return true;
+            };
         }
     }
-    if (image_creator->image_name == "")
+    return [image_creator](cli_result_t & result)
-    {
-        fprintf(stderr, "Image name is missing\n");
-        exit(1);
-    }
-    if (image_creator->image_name.find('@') != std::string::npos)
-    {
-        fprintf(stderr, "Image name can't contain @ character\n");
-        exit(1);
-    }
-    return [image_creator]()
     {
         image_creator->loop();
         if (image_creator->is_done())
         {
+            result = image_creator->result;
             delete image_creator;
             return true;
         }

src/cli_df.cpp

@@ -12,6 +12,7 @@ struct pool_lister_t
 
     int state = 0;
     json11::Json space_info;
+    cli_result_t result;
     std::map<pool_id_t, json11::Json::object> pool_stats;
 
     bool is_done()
@@ -24,8 +25,7 @@ struct pool_lister_t
         if (state == 1)
             goto resume_1;
         // Space statistics - pool/stats/<pool>
-        parent->waiting++;
+        parent->etcd_txn(json11::Json::object {
-        parent->cli->st_cli.etcd_txn(json11::Json::object {
             { "success", json11::Json::array {
                 json11::Json::object {
                     { "request_range", json11::Json::object {
@@ -48,21 +48,18 @@ struct pool_lister_t
                     } },
                 },
             } },
-        }, ETCD_SLOW_TIMEOUT, [this](std::string err, json11::Json res)
-        {
-            parent->waiting--;
-            if (err != "")
-            {
-                fprintf(stderr, "Error reading from etcd: %s\n", err.c_str());
-                exit(1);
-            }
-            space_info = res;
-            parent->ringloop->wakeup();
         });
         state = 1;
 resume_1:
         if (parent->waiting > 0)
             return;
+        if (parent->etcd_err.err)
+        {
+            result = parent->etcd_err;
+            state = 100;
+            return;
+        }
+        space_info = parent->etcd_result;
         std::map<pool_id_t, uint64_t> osd_free;
         for (auto & kv_item: space_info["responses"][0]["response_range"]["kvs"].array_items())
         {
@@ -118,9 +115,14 @@ resume_1:
                     pool_avail = pg_free;
                 }
             }
+            if (pool_avail == UINT64_MAX)
+            {
+                pool_avail = 0;
+            }
             if (pool_cfg.scheme != POOL_SCHEME_REPLICATED)
             {
-                pool_avail = pool_avail * (pool_cfg.pg_size - pool_cfg.parity_chunks) / pool_stats[pool_cfg.id]["pg_real_size"].uint64_value();
+                uint64_t pg_real_size = pool_stats[pool_cfg.id]["pg_real_size"].uint64_value();
+                pool_avail = pg_real_size > 0 ? pool_avail * (pool_cfg.pg_size - pool_cfg.parity_chunks) / pg_real_size : 0;
             }
             pool_stats[pool_cfg.id] = json11::Json::object {
                 { "name", pool_cfg.name },
@@ -129,8 +131,8 @@ resume_1:
                 { "scheme_name", pool_cfg.scheme == POOL_SCHEME_REPLICATED
                     ? std::to_string(pool_cfg.pg_size)+"/"+std::to_string(pool_cfg.pg_minsize)
                     : "EC "+std::to_string(pool_cfg.pg_size-pool_cfg.parity_chunks)+"+"+std::to_string(pool_cfg.parity_chunks) },
-                { "used_raw", (uint64_t)(pool_stats[pool_cfg.id]["used_raw_tb"].number_value() * (1l<<40)) },
+                { "used_raw", (uint64_t)(pool_stats[pool_cfg.id]["used_raw_tb"].number_value() * ((uint64_t)1<<40)) },
-                { "total_raw", (uint64_t)(pool_stats[pool_cfg.id]["total_raw_tb"].number_value() * (1l<<40)) },
+                { "total_raw", (uint64_t)(pool_stats[pool_cfg.id]["total_raw_tb"].number_value() * ((uint64_t)1<<40)) },
                 { "max_available", pool_avail },
                 { "raw_to_usable", pool_stats[pool_cfg.id]["raw_to_usable"].number_value() },
                 { "space_efficiency", pool_stats[pool_cfg.id]["space_efficiency"].number_value() },
@@ -155,10 +157,12 @@ resume_1:
         get_stats();
         if (parent->waiting > 0)
             return;
+        if (state == 100)
+            return;
         if (parent->json_output)
         {
             // JSON output
-            printf("%s\n", json11::Json(to_list()).dump().c_str());
+            result.data = to_list();
             state = 100;
             return;
         }
@@ -199,28 +203,34 @@ resume_1:
         json11::Json::array list;
         for (auto & kv: pool_stats)
         {
-            kv.second["total_fmt"] = format_size(kv.second["total_raw"].uint64_value() / kv.second["raw_to_usable"].number_value());
+            double raw_to = kv.second["raw_to_usable"].number_value();
-            kv.second["used_fmt"] = format_size(kv.second["used_raw"].uint64_value() / kv.second["raw_to_usable"].number_value());
+            if (raw_to < 0.000001 && raw_to > -0.000001)
+                raw_to = 1;
+            kv.second["total_fmt"] = format_size(kv.second["total_raw"].uint64_value() / raw_to);
+            kv.second["used_fmt"] = format_size(kv.second["used_raw"].uint64_value() / raw_to);
             kv.second["max_avail_fmt"] = format_size(kv.second["max_available"].uint64_value());
-            kv.second["used_pct"] = format_q(100 - 100*kv.second["max_available"].uint64_value() *
+            kv.second["used_pct"] = format_q(kv.second["total_raw"].uint64_value()
-                kv.second["raw_to_usable"].number_value() / kv.second["total_raw"].uint64_value())+"%";
+                ? (100 - 100*kv.second["max_available"].uint64_value() *
+                    kv.second["raw_to_usable"].number_value() / kv.second["total_raw"].uint64_value())
+                : 100)+"%";
             kv.second["eff_fmt"] = format_q(kv.second["space_efficiency"].number_value()*100)+"%";
         }
-        printf("%s", print_table(to_list(), cols, parent->color).c_str());
+        result.data = to_list();
+        result.text = print_table(result.data, cols, parent->color);
         state = 100;
     }
 };
 
-std::function<bool(void)> cli_tool_t::start_df(json11::Json cfg)
+std::function<bool(cli_result_t &)> cli_tool_t::start_df(json11::Json cfg)
 {
-    json11::Json::array cmd = cfg["command"].array_items();
     auto lister = new pool_lister_t();
     lister->parent = this;
-    return [lister]()
+    return [lister](cli_result_t & result)
     {
         lister->loop();
         if (lister->is_done())
         {
+            result = lister->result;
             delete lister;
             return true;
         }

src/cli_flatten.cpp

@@ -3,6 +3,7 @@
 
 #include "cli.h"
 #include "cluster_client.h"
+#include <sys/stat.h>
 
 // Flatten a layer: merge all parents into a layer and break the connection completely
 struct snap_flattener_t
@@ -21,12 +22,19 @@ struct snap_flattener_t
     std::string top_parent_name;
     inode_t target_id = 0;
     int state = 0;
-    std::function<bool(void)> merger_cb;
+    std::function<bool(cli_result_t &)> merger_cb;
+    cli_result_t result;
 
     void get_merge_parents()
     {
         // Get all parents of target
         inode_config_t *target_cfg = parent->get_inode_cfg(target_name);
+        if (!target_cfg)
+        {
+            result = (cli_result_t){ .err = ENOENT, .text = "Layer "+target_name+" not found" };
+            state = 100;
+            return;
+        }
         target_id = target_cfg->num;
         std::vector<inode_t> chain_list;
         inode_config_t *cur = target_cfg;
@@ -36,23 +44,34 @@ struct snap_flattener_t
             auto it = parent->cli->st_cli.inode_config.find(cur->parent_id);
             if (it == parent->cli->st_cli.inode_config.end())
             {
-                fprintf(stderr, "Parent inode of layer %s (id %ld) not found\n", cur->name.c_str(), cur->parent_id);
+                result = (cli_result_t){
-                exit(1);
+                    .err = ENOENT,
+                    .text = "Parent inode of layer "+cur->name+" (id "+std::to_string(cur->parent_id)+") does not exist",
+                    .data = json11::Json::object {
+                        { "error", "parent-not-found" },
+                        { "inode_id", cur->num },
+                        { "inode_name", cur->name },
+                        { "parent_id", cur->parent_id },
+                    },
+                };
+                state = 100;
+                return;
             }
             cur = &it->second;
             chain_list.push_back(cur->num);
         }
         if (cur->parent_id != 0)
         {
-            fprintf(stderr, "Layer %s has a loop in parents\n", target_name.c_str());
+            result = (cli_result_t){ .err = EBADF, .text = "Layer "+target_name+" has a loop in parents" };
-            exit(1);
+            state = 100;
+            return;
         }
         top_parent_name = cur->name;
     }
 
     bool is_done()
     {
-        return state == 5;
+        return state == 100;
     }
 
     void loop()
@@ -63,11 +82,20 @@ struct snap_flattener_t
             goto resume_2;
         else if (state == 3)
             goto resume_3;
+        if (target_name == "")
+        {
+            result = (cli_result_t){ .err = EINVAL, .text = "Layer to flatten not specified" };
+            state = 100;
+            return;
+        }
         // Get parent layers
         get_merge_parents();
+        if (state == 100)
+            return;
         // Start merger
         merger_cb = parent->start_merge(json11::Json::object {
-            { "command", json11::Json::array{ "merge-data", top_parent_name, target_name } },
+            { "from", top_parent_name },
+            { "to", target_name },
             { "target", target_name },
             { "delete-source", false },
             { "cas", use_cas },
@@ -75,14 +103,19 @@ struct snap_flattener_t
         });
         // Wait for it
 resume_1:
-        while (!merger_cb())
+        while (!merger_cb(result))
         {
             state = 1;
             return;
         }
         merger_cb = NULL;
+        if (result.err)
+        {
+            state = 100;
+            return;
+        }
         // Change parent
-        parent->change_parent(target_id, 0);
+        parent->change_parent(target_id, 0, &result);
         // Wait for it to complete
         state = 2;
 resume_2:
@@ -91,31 +124,26 @@ resume_2:
         state = 3;
 resume_3:
         // Done
-        return;
+        state = 100;
     }
 };
 
-std::function<bool(void)> cli_tool_t::start_flatten(json11::Json cfg)
+std::function<bool(cli_result_t &)> cli_tool_t::start_flatten(json11::Json cfg)
 {
-    json11::Json::array cmd = cfg["command"].array_items();
     auto flattener = new snap_flattener_t();
     flattener->parent = this;
-    flattener->target_name = cmd.size() > 1 ? cmd[1].string_value() : "";
+    flattener->target_name = cfg["image"].string_value();
-    if (flattener->target_name == "")
-    {
-        fprintf(stderr, "Layer to flatten argument is missing\n");
-        exit(1);
-    }
     flattener->fsync_interval = cfg["fsync-interval"].uint64_value();
     if (!flattener->fsync_interval)
         flattener->fsync_interval = 128;
     if (!cfg["cas"].is_null())
         flattener->use_cas = cfg["cas"].uint64_value() ? 2 : 0;
-    return [flattener]()
+    return [flattener](cli_result_t & result)
     {
         flattener->loop();
         if (flattener->is_done())
         {
+            result = flattener->result;
             delete flattener;
             return true;
         }

src/cli_ls.cpp

@@ -24,6 +24,7 @@ struct image_lister_t
     int state = 0;
     std::map<inode_t, json11::Json::object> stats;
     json11::Json space_info;
+    cli_result_t result;
 
     bool is_done()
     {
@@ -44,8 +45,9 @@ struct image_lister_t
             }
             if (!list_pool_id)
             {
-                fprintf(stderr, "Pool %s does not exist\n", list_pool_name.c_str());
+                result = (cli_result_t){ .err = ENOENT, .text = "Pool "+list_pool_name+" does not exist" };
-                exit(1);
+                state = 100;
+                return;
             }
         }
         for (auto & ic: parent->cli->st_cli.inode_config)
@@ -84,8 +86,7 @@ struct image_lister_t
         // Space statistics
         // inode/stats/<pool>/<inode>::raw_used divided by pool/stats/<pool>::pg_real_size
         // multiplied by 1 or number of data drives
-        parent->waiting++;
+        parent->etcd_txn(json11::Json::object {
-        parent->cli->st_cli.etcd_txn(json11::Json::object {
             { "success", json11::Json::array {
                 json11::Json::object {
                     { "request_range", json11::Json::object {
@@ -112,21 +113,18 @@ struct image_lister_t
                     } },
                 },
             } },
-        }, ETCD_SLOW_TIMEOUT, [this](std::string err, json11::Json res)
-        {
-            parent->waiting--;
-            if (err != "")
-            {
-                fprintf(stderr, "Error reading from etcd: %s\n", err.c_str());
-                exit(1);
-            }
-            space_info = res;
-            parent->ringloop->wakeup();
         });
         state = 1;
 resume_1:
         if (parent->waiting > 0)
             return;
+        if (parent->etcd_err.err)
+        {
+            result = parent->etcd_err;
+            state = 100;
+            return;
+        }
+        space_info = parent->etcd_result;
         std::map<pool_id_t, uint64_t> pool_pg_real_size;
         for (auto & kv_item: space_info["responses"][0]["response_range"]["kvs"].array_items())
         {
@@ -164,7 +162,7 @@ resume_1:
             if (pool_it != parent->cli->st_cli.pool_config.end())
             {
                 auto & pool_cfg = pool_it->second;
-                used_size = used_size / pool_pg_real_size[pool_id]
+                used_size = used_size / (pool_pg_real_size[pool_id] ? pool_pg_real_size[pool_id] : 1)
                     * (pool_cfg.scheme == POOL_SCHEME_REPLICATED ? 1 : pool_cfg.pg_size-pool_cfg.parity_chunks);
             }
             auto stat_it = stats.find(inode_num);
@@ -255,11 +253,13 @@ resume_1:
             get_stats();
             if (parent->waiting > 0)
                 return;
+            if (state == 100)
+                return;
         }
+        result.data = to_list();
         if (parent->json_output)
         {
             // JSON output
-            printf("%s\n", json11::Json(to_list()).dump().c_str());
             state = 100;
             return;
         }
@@ -369,7 +369,7 @@ resume_1:
             kv.second["size_fmt"] = format_size(kv.second["size"].uint64_value());
             kv.second["ro"] = kv.second["readonly"].bool_value() ? "RO" : "-";
         }
-        printf("%s", print_table(to_list(), cols, parent->color).c_str());
+        result.text = print_table(to_list(), cols, parent->color);
         state = 100;
     }
 };
@@ -446,23 +446,26 @@ std::string print_table(json11::Json items, json11::Json header, bool use_esc)
     return str;
 }
 
-static uint64_t size_thresh[] = { 1024l*1024*1024*1024, 1024l*1024*1024, 1024l*1024, 1024, 0 };
+static uint64_t size_thresh[] = { (uint64_t)1024*1024*1024*1024, (uint64_t)1024*1024*1024, (uint64_t)1024*1024, 1024, 0 };
+static uint64_t size_thresh_d[] = { (uint64_t)1000000000000, (uint64_t)1000000000, (uint64_t)1000000, (uint64_t)1000, 0 };
+static const int size_thresh_n = sizeof(size_thresh)/sizeof(size_thresh[0]);
 static const char *size_unit = "TGMKB";
 
-std::string format_size(uint64_t size)
+std::string format_size(uint64_t size, bool nobytes)
 {
+    uint64_t *thr = nobytes ? size_thresh_d : size_thresh;
     char buf[256];
-    for (int i = 0; i < sizeof(size_thresh)/sizeof(size_thresh[0]); i++)
+    for (int i = 0; i < size_thresh_n; i++)
     {
-        if (size >= size_thresh[i] || i >= sizeof(size_thresh)/sizeof(size_thresh[0])-1)
+        if (size >= thr[i] || i >= size_thresh_n-1)
         {
-            double value = size_thresh[i] ? (double)size/size_thresh[i] : size;
+            double value = thr[i] ? (double)size/thr[i] : size;
             int l = snprintf(buf, sizeof(buf), "%.1f", value);
             assert(l < sizeof(buf)-2);
             if (buf[l-1] == '0')
                 l -= 2;
-            buf[l] = ' ';
+            buf[l] = i == size_thresh_n-1 && nobytes ? 0 : ' ';
-            buf[l+1] = size_unit[i];
+            buf[l+1] = i == size_thresh_n-1 && nobytes ? 0 : size_unit[i];
             buf[l+2] = 0;
             break;
         }
@@ -553,9 +556,8 @@ back:
     return true;
 }
 
-std::function<bool(void)> cli_tool_t::start_ls(json11::Json cfg)
+std::function<bool(cli_result_t &)> cli_tool_t::start_ls(json11::Json cfg)
 {
-    json11::Json::array cmd = cfg["command"].array_items();
     auto lister = new image_lister_t();
     lister->parent = this;
     lister->list_pool_id = cfg["pool"].uint64_value();
@@ -565,15 +567,16 @@ std::function<bool(void)> cli_tool_t::start_ls(json11::Json cfg)
     lister->sort_field = cfg["sort"].string_value();
     lister->reverse = cfg["reverse"].bool_value();
     lister->max_count = cfg["count"].uint64_value();
-    for (int i = 1; i < cmd.size(); i++)
+    for (auto & item: cfg["names"].array_items())
     {
-        lister->only_names.insert(cmd[i].string_value());
+        lister->only_names.insert(item.string_value());
     }
-    return [lister]()
+    return [lister](cli_result_t & result)
     {
         lister->loop();
         if (lister->is_done())
         {
+            result = lister->result;
             delete lister;
             return true;
         }

src/cli_merge.cpp

@@ -12,6 +12,9 @@ struct snap_rw_op_t
     cluster_op_t op;
     int todo = 0;
     uint32_t start = 0, end = 0;
+    int error_code = 0;
+    uint64_t error_offset = 0;
+    bool error_read = false;
 };
 
 // Layer merge is the base for multiple operations:
@@ -54,17 +57,45 @@ struct snap_merger_t
     uint64_t last_written_offset = 0;
     int deleted_unsynced = 0;
     uint64_t processed = 0, to_process = 0;
+    std::string rwo_error;
+
+    cli_result_t result;
 
     void start_merge()
     {
+        if (from_name == "" || to_name == "")
+        {
+            result = (cli_result_t){ .err = EINVAL, .text = "Beginning or end of the merge sequence is missing" };
+            state = 100;
+            return;
+        }
         check_delete_source = delete_source || check_delete_source;
         inode_config_t *from_cfg = parent->get_inode_cfg(from_name);
+        if (!from_cfg)
+        {
+            result = (cli_result_t){ .err = ENOENT, .text = "Layer "+from_name+" not found" };
+            state = 100;
+            return;
+        }
         inode_config_t *to_cfg = parent->get_inode_cfg(to_name);
+        if (!to_cfg)
+        {
+            result = (cli_result_t){ .err = ENOENT, .text = "Layer "+to_name+" not found" };
+            state = 100;
+            return;
+        }
         inode_config_t *target_cfg = target_name == "" ? from_cfg : parent->get_inode_cfg(target_name);
+        if (!target_cfg)
+        {
+            result = (cli_result_t){ .err = ENOENT, .text = "Layer "+target_name+" not found" };
+            state = 100;
+            return;
+        }
         if (to_cfg->num == from_cfg->num)
         {
-            fprintf(stderr, "Only one layer specified, nothing to merge\n");
+            result = (cli_result_t){ .err = EINVAL, .text = "Only one layer specified, nothing to merge" };
-            exit(1);
+            state = 100;
+            return;
         }
         // Check that to_cfg is actually a child of from_cfg and target_cfg is somewhere between them
         std::vector<inode_t> chain_list;
@@ -78,8 +109,18 @@ struct snap_merger_t
             auto it = parent->cli->st_cli.inode_config.find(cur->parent_id);
             if (it == parent->cli->st_cli.inode_config.end())
             {
-                fprintf(stderr, "Parent inode of layer %s (id %ld) not found\n", cur->name.c_str(), cur->parent_id);
+                result = (cli_result_t){
-                exit(1);
+                    .err = ENOENT,
+                    .text = "Parent inode of layer "+cur->name+" (id "+std::to_string(cur->parent_id)+") does not exist",
+                    .data = json11::Json::object {
+                        { "error", "parent-not-found" },
+                        { "inode_id", cur->num },
+                        { "inode_name", cur->name },
+                        { "parent_id", cur->parent_id },
+                    },
+                };
+                state = 100;
+                return;
             }
             cur = &it->second;
             chain_list.push_back(cur->num);
@@ -87,8 +128,9 @@ struct snap_merger_t
         }
         if (cur->parent_id != from_cfg->num)
         {
-            fprintf(stderr, "Layer %s is not a child of %s\n", to_name.c_str(), from_name.c_str());
+            result = (cli_result_t){ .err = EINVAL, .text = "Layer "+to_name+" is not a child of "+from_name };
-            exit(1);
+            state = 100;
+            return;
         }
         chain_list.push_back(from_cfg->num);
         layer_block_size[from_cfg->num] = get_block_size(from_cfg->num);
@@ -99,8 +141,9 @@ struct snap_merger_t
         }
         if (sources.find(target_cfg->num) == sources.end())
         {
-            fprintf(stderr, "Layer %s is not between %s and %s\n", target_name.c_str(), to_name.c_str(), from_name.c_str());
+            result = (cli_result_t){ .err = EINVAL, .text = "Layer "+target_name+" is not between "+to_name+" and "+from_name };
-            exit(1);
+            state = 100;
+            return;
         }
         target = target_cfg->num;
         target_rank = sources.at(target);
@@ -130,14 +173,15 @@ struct snap_merger_t
                     int parent_rank = it->second;
                     if (parent_rank < to_rank && (parent_rank >= target_rank || check_delete_source))
                     {
-                        fprintf(
+                        result = (cli_result_t){
-                            stderr, "Layers at or above %s, but below %s are not allowed"
+                            .err = EINVAL,
-                                " to have other children, but %s is a child of %s\n",
+                            .text = "Layers at or above "+(check_delete_source ? from_name : target_name)+
-                            (check_delete_source ? from_name.c_str() : target_name.c_str()),
+                                ", but below "+to_name+" are not allowed to have other children, but "+
-                            to_name.c_str(), ic.second.name.c_str(),
+                                ic.second.name+" is a child of "+
-                            parent->cli->st_cli.inode_config.at(ic.second.parent_id).name.c_str()
+                                parent->cli->st_cli.inode_config.at(ic.second.parent_id).name,
-                        );
+                        };
-                        exit(1);
+                        state = 100;
+                        return;
                     }
                     if (parent_rank >= to_rank)
                     {
@@ -152,11 +196,14 @@ struct snap_merger_t
             use_cas = 0;
         }
         sources.erase(target);
-        printf(
+        if (parent->progress)
-            "Merging %ld layer(s) into target %s%s (inode %lu in pool %u)\n",
+        {
-            sources.size(), target_cfg->name.c_str(),
+            printf(
-            use_cas ? " online (with CAS)" : "", INODE_NO_POOL(target), INODE_POOL(target)
+                "Merging %ld layer(s) into target %s%s (inode %lu in pool %u)\n",
-        );
+                sources.size(), target_cfg->name.c_str(),
+                use_cas ? " online (with CAS)" : "", INODE_NO_POOL(target), INODE_POOL(target)
+            );
+        }
         target_block_size = get_block_size(target);
     }
 
@@ -179,7 +226,7 @@ struct snap_merger_t
 
     bool is_done()
     {
-        return state == 6;
+        return state == 100;
     }
 
     void continue_merge()
@@ -194,8 +241,8 @@ struct snap_merger_t
             goto resume_4;
         else if (state == 5)
             goto resume_5;
-        else if (state == 6)
+        else if (state == 100)
-            goto resume_6;
+            goto resume_100;
         // Get parents and so on
         start_merge();
         // First list lower layers
@@ -253,7 +300,8 @@ struct snap_merger_t
         oit = merge_offsets.begin();
     resume_5:
         // Now read, overwrite and optionally delete offsets one by one
-        while (in_flight < parent->iodepth*parent->parallel_osds && oit != merge_offsets.end())
+        while (in_flight < parent->iodepth*parent->parallel_osds &&
+            oit != merge_offsets.end() && !rwo_error.size())
         {
             in_flight++;
             read_and_write(*oit);
@@ -264,6 +312,15 @@ struct snap_merger_t
                 printf("\rOverwriting blocks: %lu/%lu", processed, to_process);
             }
         }
+        if (in_flight == 0 && rwo_error.size())
+        {
+            result = (cli_result_t){
+                .err = EIO,
+                .text = rwo_error,
+            };
+            state = 100;
+            return;
+        }
         if (in_flight > 0 || oit != merge_offsets.end())
         {
             // Wait until overwrites finish
@@ -274,9 +331,9 @@ struct snap_merger_t
             printf("\rOverwriting blocks: %lu/%lu\n", to_process, to_process);
         }
         // Done
-        printf("Done, layers from %s to %s merged into %s\n", from_name.c_str(), to_name.c_str(), target_name.c_str());
+        result = (cli_result_t){ .text = "Done, layers from "+from_name+" to "+to_name+" merged into "+target_name };
-        state = 6;
+        state = 100;
-    resume_6:
+    resume_100:
         return;
     }
 
@@ -314,7 +371,10 @@ struct snap_merger_t
                     if (status & INODE_LIST_DONE)
                     {
                         auto & name = parent->cli->st_cli.inode_config.at(src).name;
-                        printf("Got listing of layer %s (inode %lu in pool %u)\n", name.c_str(), INODE_NO_POOL(src), INODE_POOL(src));
+                        if (parent->progress)
+                        {
+                            printf("Got listing of layer %s (inode %lu in pool %u)\n", name.c_str(), INODE_NO_POOL(src), INODE_POOL(src));
+                        }
                         if (delete_source)
                         {
                             // Sort the inode listing
@@ -396,8 +456,9 @@ struct snap_merger_t
         {
             if (op->retval != op->len)
             {
-                fprintf(stderr, "error reading target at offset %lx: %s\n", op->offset, strerror(-op->retval));
+                rwo->error_code = -op->retval;
-                exit(1);
+                rwo->error_offset = op->offset;
+                rwo->error_read = true;
             }
             next_write(rwo);
         };
@@ -410,9 +471,9 @@ struct snap_merger_t
         // FIXME: Allow to use single write with "holes" (OSDs don't allow it yet)
         uint32_t gran = parent->cli->get_bs_bitmap_granularity();
         uint64_t bitmap_size = target_block_size / gran;
-        while (rwo->end < bitmap_size)
+        while (rwo->end < bitmap_size && !rwo->error_code)
         {
-            auto bit = ((*(uint8_t*)(rwo->op.bitmap_buf + (rwo->end >> 3))) & (1 << (rwo->end & 0x7)));
+            auto bit = ((*((uint8_t*)rwo->op.bitmap_buf + (rwo->end >> 3))) & (1 << (rwo->end & 0x7)));
             if (!bit)
             {
                 if (rwo->end > rwo->start)
@@ -434,7 +495,7 @@ struct snap_merger_t
                 rwo->end++;
             }
         }
-        if (rwo->end > rwo->start)
+        if (rwo->end > rwo->start && !rwo->error_code)
         {
             // write start->end
             rwo->todo++;
@@ -459,7 +520,7 @@ struct snap_merger_t
         subop->len = end-start;
         subop->version = version;
         subop->flags = OSD_OP_IGNORE_READONLY;
-        subop->iov.push_back(rwo->buf+start, end-start);
+        subop->iov.push_back((uint8_t*)rwo->buf+start, end-start);
         subop->callback = [this, rwo](cluster_op_t *subop)
         {
             rwo->todo--;
@@ -473,8 +534,9 @@ struct snap_merger_t
                     delete subop;
                     return;
                 }
-                fprintf(stderr, "error writing target at offset %lx: %s\n", subop->offset, strerror(-subop->retval));
+                rwo->error_code = -subop->retval;
-                exit(1);
+                rwo->error_offset = subop->offset;
+                rwo->error_read = false;
             }
             // Increment CAS version
             rwo->op.version++;
@@ -495,7 +557,7 @@ struct snap_merger_t
         subop->offset = offset;
         subop->len = 0;
         subop->flags = OSD_OP_IGNORE_READONLY;
-        subop->callback = [this](cluster_op_t *subop)
+        subop->callback = [](cluster_op_t *subop)
         {
             if (subop->retval != 0)
             {
@@ -510,19 +572,20 @@ struct snap_merger_t
     {
         if (!rwo->todo)
         {
-            if (last_written_offset < rwo->op.offset+target_block_size)
+            if (!rwo->error_code &&
+                last_written_offset < rwo->op.offset+target_block_size)
             {
                 last_written_offset = rwo->op.offset+target_block_size;
             }
-            if (delete_source)
+            if (!rwo->error_code && delete_source)
             {
                 deleted_unsynced++;
                 if (deleted_unsynced >= fsync_interval)
                 {
-                    uint64_t from = last_fsync_offset, to = last_written_offset;
+                    uint64_t to = last_written_offset;
                     cluster_op_t *subop = new cluster_op_t;
                     subop->opcode = OSD_OP_SYNC;
-                    subop->callback = [this, from, to](cluster_op_t *subop)
+                    subop->callback = [this, to](cluster_op_t *subop)
                     {
                         delete subop;
                         // We can now delete source data between <from> and <to>
@@ -544,6 +607,13 @@ struct snap_merger_t
                 }
             }
             free(rwo->buf);
+            if (rwo->error_code)
+            {
+                char buf[1024];
+                snprintf(buf, 1024, "Error %s target at offset %lx: %s",
+                    rwo->error_read ? "reading" : "writing", rwo->error_offset, strerror(rwo->error_code));
+                rwo_error = std::string(buf);
+            }
             delete rwo;
             in_flight--;
             continue_merge_reent();
@@ -551,30 +621,25 @@ struct snap_merger_t
     }
 };
 
-std::function<bool(void)> cli_tool_t::start_merge(json11::Json cfg)
+std::function<bool(cli_result_t &)> cli_tool_t::start_merge(json11::Json cfg)
 {
-    json11::Json::array cmd = cfg["command"].array_items();
     auto merger = new snap_merger_t();
     merger->parent = this;
-    merger->from_name = cmd.size() > 1 ? cmd[1].string_value() : "";
+    merger->from_name = cfg["from"].string_value();
-    merger->to_name = cmd.size() > 2 ? cmd[2].string_value() : "";
+    merger->to_name = cfg["to"].string_value();
     merger->target_name = cfg["target"].string_value();
-    if (merger->from_name == "" || merger->to_name == "")
-    {
-        fprintf(stderr, "Beginning or end of the merge sequence is missing\n");
-        exit(1);
-    }
     merger->delete_source = cfg["delete-source"].string_value() != "";
     merger->fsync_interval = cfg["fsync-interval"].uint64_value();
     if (!merger->fsync_interval)
         merger->fsync_interval = 128;
     if (!cfg["cas"].is_null())
         merger->use_cas = cfg["cas"].uint64_value() ? 2 : 0;
-    return [merger]()
+    return [merger](cli_result_t & result)
     {
         merger->continue_merge_reent();
         if (merger->is_done())
         {
+            result = merger->result;
             delete merger;
             return true;
         }

src/cli_modify.cpp

@@ -13,6 +13,7 @@ struct image_changer_t
     std::string image_name;
     std::string new_name;
     uint64_t new_size = 0;
+    bool force_size = false;
     bool set_readonly = false, set_readwrite = false, force = false;
     // interval between fsyncs
     int fsync_interval = 128;
@@ -23,7 +24,8 @@ struct image_changer_t
     bool has_children = false;
 
     int state = 0;
-    std::function<bool(void)> cb;
+    std::function<bool(cli_result_t &)> cb;
+    cli_result_t result;
 
     bool is_done()
     {
@@ -36,6 +38,18 @@ struct image_changer_t
             goto resume_1;
         else if (state == 2)
             goto resume_2;
+        if (image_name == "")
+        {
+            result = (cli_result_t){ .err = EINVAL, .text = "Image name is missing" };
+            state = 100;
+            return;
+        }
+        if (new_size != 0 && (new_size % 4096) && !force_size)
+        {
+            result = (cli_result_t){ .err = EINVAL, .text = "Image size should be a multiple of 4096" };
+            state = 100;
+            return;
+        }
         for (auto & ic: parent->cli->st_cli.inode_config)
         {
             if (ic.second.name == image_name)
@@ -46,14 +60,16 @@ struct image_changer_t
             }
             if (new_name != "" && ic.second.name == new_name)
             {
-                fprintf(stderr, "Image %s already exists\n", new_name.c_str());
+                result = (cli_result_t){ .err = EEXIST, .text = "Image "+new_name+" already exists" };
-                exit(1);
+                state = 100;
+                return;
             }
         }
         if (!inode_num)
         {
-            fprintf(stderr, "Image %s does not exist\n", image_name.c_str());
+            result = (cli_result_t){ .err = ENOENT, .text = "Image "+image_name+" does not exist" };
-            exit(1);
+            state = 100;
+            return;
         }
         for (auto & ic: parent->cli->st_cli.inode_config)
         {
@@ -65,37 +81,43 @@ struct image_changer_t
         }
         if ((!set_readwrite || !cfg.readonly) &&
             (!set_readonly || cfg.readonly) &&
-            (!new_size || cfg.size == new_size) &&
+            (!new_size && !force_size || cfg.size == new_size) &&
             (new_name == "" || new_name == image_name))
         {
-            printf("No change\n");
+            result = (cli_result_t){ .text = "No change" };
             state = 100;
             return;
         }
-        if (new_size != 0)
+        if (new_size != 0 || force_size)
         {
             if (cfg.size >= new_size)
             {
                 // Check confirmation when trimming an image with children
                 if (has_children && !force)
                 {
-                    fprintf(stderr, "Image %s has children. Refusing to shrink it without --force\n", image_name.c_str());
+                    result = (cli_result_t){ .err = EINVAL, .text = "Image "+image_name+" has children. Refusing to shrink it without --force" };
-                    exit(1);
+                    state = 100;
+                    return;
                 }
                 // Shrink the image first
-                cb = parent->start_rm(json11::Json::object {
+                cb = parent->start_rm_data(json11::Json::object {
                     { "inode", INODE_NO_POOL(inode_num) },
                     { "pool", (uint64_t)INODE_POOL(inode_num) },
                     { "fsync-interval", fsync_interval },
-                    { "min-offset", new_size },
+                    { "min-offset", ((new_size+4095)/4096)*4096 },
                 });
 resume_1:
-                while (!cb())
+                while (!cb(result))
                 {
                     state = 1;
                     return;
                 }
                 cb = NULL;
+                if (result.err)
+                {
+                    state = 100;
+                    return;
+                }
             }
             cfg.size = new_size;
         }
@@ -109,8 +131,9 @@ resume_1:
             // Check confirmation when making an image with children read-write
             if (has_children && !force)
             {
-                fprintf(stderr, "Image %s has children. Refusing to make it read-write without --force\n", image_name.c_str());
+                result = (cli_result_t){ .err = EINVAL, .text = "Image "+image_name+" has children. Refusing to make it read-write without --force" };
-                exit(1);
+                state = 100;
+                return;
             }
         }
         if (new_name != "")
@@ -170,52 +193,46 @@ resume_1:
                 } }
             });
         }
-        parent->waiting++;
+        parent->etcd_txn(json11::Json::object {
-        parent->cli->st_cli.etcd_txn(json11::Json::object {
             { "compare", checks },
             { "success", success },
-        }, ETCD_SLOW_TIMEOUT, [this](std::string err, json11::Json res)
-        {
-            if (err != "")
-            {
-                fprintf(stderr, "Error changing %s: %s\n", image_name.c_str(), err.c_str());
-                exit(1);
-            }
-            if (!res["succeeded"].bool_value())
-            {
-                fprintf(stderr, "Image %s was modified by someone else, please repeat your request\n", image_name.c_str());
-                exit(1);
-            }
-            parent->waiting--;
-            parent->ringloop->wakeup();
         });
         state = 2;
 resume_2:
         if (parent->waiting > 0)
             return;
-        printf("Image %s modified\n", image_name.c_str());
+        if (parent->etcd_err.err)
+        {
+            result = parent->etcd_err;
+            state = 100;
+            return;
+        }
+        if (!parent->etcd_result["succeeded"].bool_value())
+        {
+            result = (cli_result_t){ .err = EAGAIN, .text = "Image "+image_name+" was modified by someone else, please repeat your request" };
+            state = 100;
+            return;
+        }
+        // Save into inode_config for library users to be able to take it from there immediately
+        cfg.mod_revision = parent->etcd_result["responses"][0]["response_put"]["header"]["revision"].uint64_value();
+        if (new_name != "")
+        {
+            parent->cli->st_cli.inode_by_name.erase(image_name);
+        }
+        parent->cli->st_cli.insert_inode_config(cfg);
+        result = (cli_result_t){ .err = 0, .text = "Image "+image_name+" modified" };
         state = 100;
     }
 };
 
-std::function<bool(void)> cli_tool_t::start_modify(json11::Json cfg)
+std::function<bool(cli_result_t &)> cli_tool_t::start_modify(json11::Json cfg)
 {
-    json11::Json::array cmd = cfg["command"].array_items();
     auto changer = new image_changer_t();
     changer->parent = this;
-    changer->image_name = cmd.size() > 1 ? cmd[1].string_value() : "";
+    changer->image_name = cfg["image"].string_value();
-    if (changer->image_name == "")
-    {
-        fprintf(stderr, "Image name is missing\n");
-        exit(1);
-    }
     changer->new_name = cfg["rename"].string_value();
-    changer->new_size = parse_size(cfg["resize"].string_value());
+    changer->new_size = parse_size(cfg["resize"].as_string());
-    if (changer->new_size != 0 && (changer->new_size % 4096))
+    changer->force_size = cfg["force_size"].bool_value();
-    {
-        fprintf(stderr, "Image size should be a multiple of 4096\n");
-        exit(1);
-    }
     changer->force = cfg["force"].bool_value();
     changer->set_readonly = cfg["readonly"].bool_value();
     changer->set_readwrite = cfg["readwrite"].bool_value();
@@ -223,11 +240,12 @@ std::function<bool(void)> cli_tool_t::start_modify(json11::Json cfg)
     if (!changer->fsync_interval)
         changer->fsync_interval = 128;
     // FIXME Check that the image doesn't have children when shrinking
-    return [changer]()
+    return [changer](cli_result_t & result)
     {
         changer->loop();
         if (changer->is_done())
         {
+            result = changer->result;
             delete changer;
             return true;
         }