Add --restart_interval to tests

Monitor could deceive itself by immediately saving PG configuration changes
which weren't applied to etcd yet in memory, and apply incorrect PG history
changes next time if the first update fails.

This usually only happened under heavy load and was caught in CI. :-)

CMakeLists.txt

@@ -1,7 +1,7 @@
-cmake_minimum_required(VERSION 2.8)
+cmake_minimum_required(VERSION 2.8.12)
 
 project(vitastor)
 
-set(VERSION "0.8.3")
+set(VERSION "0.8.8")
 
 add_subdirectory(src)

VNPL-1.1-RU.txt

@@ -48,9 +48,9 @@ Vitastor, составлены для того, чтобы убедиться,
 интерфейс (прокси), опять же, без открытия в свободный публичный доступ как
 самой программы, так и прокси.
 
-  Сетевая Публичная Лицензия Vitastor разработана специально чтобы
+  Сетевая Публичная Лицензия Vitastor разработана специально, чтобы
 гарантировать, что в таких случаях и модифицированная версия программы, и
-прокси оставались доступными сообществу. Для этого лицензия требует от
+прокси останутся доступными сообществу. Для этого лицензия требует от
 операторов сетевых серверов предоставлять исходный код оригинальной программы,
 а также всех других программ, взаимодействующих с ней на их серверах,
 пользователям этих серверов, на условиях свободных лицензий. Таким образом,

csi/Makefile

@@ -1,4 +1,4 @@
-VERSION ?= v0.8.3
+VERSION ?= v0.8.8
 
 all: build push
 

csi/deploy/004-csi-nodeplugin.yaml

@@ -49,7 +49,7 @@ spec:
             capabilities:
               add: ["SYS_ADMIN"]
             allowPrivilegeEscalation: true
-          image: vitalif/vitastor-csi:v0.8.3
+          image: vitalif/vitastor-csi:v0.8.8
           args:
             - "--node=$(NODE_ID)"
             - "--endpoint=$(CSI_ENDPOINT)"

csi/deploy/007-csi-provisioner.yaml

@@ -116,7 +116,7 @@ spec:
             privileged: true
             capabilities:
               add: ["SYS_ADMIN"]
-          image: vitalif/vitastor-csi:v0.8.3
+          image: vitalif/vitastor-csi:v0.8.8
           args:
             - "--node=$(NODE_ID)"
             - "--endpoint=$(CSI_ENDPOINT)"

csi/src/config.go

@@ -5,7 +5,7 @@ package vitastor
 
 const (
     vitastorCSIDriverName    = "csi.vitastor.io"
-    vitastorCSIDriverVersion = "0.8.3"
+    vitastorCSIDriverVersion = "0.8.8"
 )
 
 // Config struct fills the parameters of request or user input

csi/src/controllerserver.go

@@ -6,11 +6,11 @@ package vitastor
 import (
     "context"
     "encoding/json"
+    "fmt"
     "strings"
     "bytes"
     "strconv"
     "time"
-    "fmt"
     "os"
     "os/exec"
     "io/ioutil"
@@ -21,8 +21,6 @@ import (
     "google.golang.org/grpc/codes"
     "google.golang.org/grpc/status"
 
-    "go.etcd.io/etcd/clientv3"
-
     "github.com/container-storage-interface/spec/lib/go/csi"
 )
 
@@ -114,6 +112,34 @@ func GetConnectionParams(params map[string]string) (map[string]string, []string,
     return ctxVars, etcdUrl, etcdPrefix
 }
 
+func invokeCLI(ctxVars map[string]string, args []string) ([]byte, error)
+{
+    if (ctxVars["etcdUrl"] != "")
+    {
+        args = append(args, "--etcd_address", ctxVars["etcdUrl"])
+    }
+    if (ctxVars["etcdPrefix"] != "")
+    {
+        args = append(args, "--etcd_prefix", ctxVars["etcdPrefix"])
+    }
+    if (ctxVars["configPath"] != "")
+    {
+        args = append(args, "--config_path", ctxVars["configPath"])
+    }
+    c := exec.Command("/usr/bin/vitastor-cli", args...)
+    var stdout, stderr bytes.Buffer
+    c.Stdout = &stdout
+    c.Stderr = &stderr
+    err := c.Run()
+    stderrStr := string(stderr.Bytes())
+    if (err != nil)
+    {
+        klog.Errorf("vitastor-cli %s failed: %s, status %s\n", strings.Join(args, " "), stderrStr, err)
+        return nil, status.Error(codes.Internal, stderrStr+" (status "+err.Error()+")")
+    }
+    return stdout.Bytes(), nil
+}
+
 // Create the volume
 func (cs *ControllerServer) CreateVolume(ctx context.Context, req *csi.CreateVolumeRequest) (*csi.CreateVolumeResponse, error)
 {
@@ -146,128 +172,41 @@ func (cs *ControllerServer) CreateVolume(ctx context.Context, req *csi.CreateVol
         volSize = ((capRange.GetRequiredBytes() + MB - 1) / MB) * MB
     }
 
-    // FIXME: The following should PROBABLY be implemented externally in a management tool
-
-    ctxVars, etcdUrl, etcdPrefix := GetConnectionParams(req.Parameters)
+    ctxVars, etcdUrl, _ := GetConnectionParams(req.Parameters)
     if (len(etcdUrl) == 0)
     {
         return nil, status.Error(codes.InvalidArgument, "no etcdUrl in storage class configuration and no etcd_address in vitastor.conf")
     }
 
-    // Connect to etcd
-    cli, err := clientv3.New(clientv3.Config{
-        DialTimeout: ETCD_TIMEOUT,
-        Endpoints: etcdUrl,
-    })
+    // Create image using vitastor-cli
+    _, err := invokeCLI(ctxVars, []string{ "create", volName, "-s", fmt.Sprintf("%v", volSize), "--pool", fmt.Sprintf("%v", poolId) })
     if (err != nil)
     {
-        return nil, status.Error(codes.Internal, "failed to connect to etcd at "+strings.Join(etcdUrl, ",")+": "+err.Error())
-    }
-    defer cli.Close()
-
-    var imageId uint64 = 0
-    for
-    {
-        // Check if the image exists
-        ctx, cancel := context.WithTimeout(context.Background(), ETCD_TIMEOUT)
-        resp, err := cli.Get(ctx, etcdPrefix+"/index/image/"+volName)
-        cancel()
-        if (err != nil)
+        if (strings.Index(err.Error(), "already exists") > 0)
         {
-            return nil, status.Error(codes.Internal, "failed to read key from etcd: "+err.Error())
-        }
-        if (len(resp.Kvs) > 0)
-        {
-            kv := resp.Kvs[0]
-            var v InodeIndex
-            err := json.Unmarshal(kv.Value, &v)
+            stat, err := invokeCLI(ctxVars, []string{ "ls", "--json", volName })
             if (err != nil)
             {
-                return nil, status.Error(codes.Internal, "invalid /index/image/"+volName+" key in etcd: "+err.Error())
+                return nil, err
             }
-            poolId = v.PoolId
-            imageId = v.Id
-            inodeCfgKey := fmt.Sprintf("/config/inode/%d/%d", poolId, imageId)
-            ctx, cancel := context.WithTimeout(context.Background(), ETCD_TIMEOUT)
-            resp, err := cli.Get(ctx, etcdPrefix+inodeCfgKey)
-            cancel()
+            var inodeCfg []InodeConfig
+            err = json.Unmarshal(stat, &inodeCfg)
             if (err != nil)
             {
-                return nil, status.Error(codes.Internal, "failed to read key from etcd: "+err.Error())
+                return nil, status.Error(codes.Internal, "Invalid JSON in vitastor-cli ls: "+err.Error())
             }
-            if (len(resp.Kvs) == 0)
+            if (len(inodeCfg) == 0)
             {
-                return nil, status.Error(codes.Internal, "missing "+inodeCfgKey+" key in etcd")
+                return nil, status.Error(codes.Internal, "vitastor-cli create said that image already exists, but ls can't find it")
             }
-            var inodeCfg InodeConfig
-            err = json.Unmarshal(resp.Kvs[0].Value, &inodeCfg)
-            if (err != nil)
-            {
-                return nil, status.Error(codes.Internal, "invalid "+inodeCfgKey+" key in etcd: "+err.Error())
-            }
-            if (inodeCfg.Size < uint64(volSize))
+            if (inodeCfg[0].Size < uint64(volSize))
             {
                 return nil, status.Error(codes.Internal, "image "+volName+" is already created, but size is less than expected")
             }
         }
         else
         {
-            // Find a free ID
-            // Create image metadata in a transaction verifying that the image doesn't exist yet AND ID is still free
-            maxIdKey := fmt.Sprintf("%s/index/maxid/%d", etcdPrefix, poolId)
-            ctx, cancel := context.WithTimeout(context.Background(), ETCD_TIMEOUT)
-            resp, err := cli.Get(ctx, maxIdKey)
-            cancel()
-            if (err != nil)
-            {
-                return nil, status.Error(codes.Internal, "failed to read key from etcd: "+err.Error())
-            }
-            var modRev int64
-            var nextId uint64
-            if (len(resp.Kvs) > 0)
-            {
-                var err error
-                nextId, err = strconv.ParseUint(string(resp.Kvs[0].Value), 10, 64)
-                if (err != nil)
-                {
-                    return nil, status.Error(codes.Internal, maxIdKey+" contains invalid ID")
-                }
-                modRev = resp.Kvs[0].ModRevision
-                nextId++
-            }
-            else
-            {
-                nextId = 1
-            }
-            inodeIdxJson, _ := json.Marshal(InodeIndex{
-                Id: nextId,
-                PoolId: poolId,
-            })
-            inodeCfgJson, _ := json.Marshal(InodeConfig{
-                Name: volName,
-                Size: uint64(volSize),
-            })
-            ctx, cancel = context.WithTimeout(context.Background(), ETCD_TIMEOUT)
-            txnResp, err := cli.Txn(ctx).If(
-                clientv3.Compare(clientv3.ModRevision(fmt.Sprintf("%s/index/maxid/%d", etcdPrefix, poolId)), "=", modRev),
-                clientv3.Compare(clientv3.CreateRevision(fmt.Sprintf("%s/index/image/%s", etcdPrefix, volName)), "=", 0),
-                clientv3.Compare(clientv3.CreateRevision(fmt.Sprintf("%s/config/inode/%d/%d", etcdPrefix, poolId, nextId)), "=", 0),
-            ).Then(
-                clientv3.OpPut(fmt.Sprintf("%s/index/maxid/%d", etcdPrefix, poolId), fmt.Sprintf("%d", nextId)),
-                clientv3.OpPut(fmt.Sprintf("%s/index/image/%s", etcdPrefix, volName), string(inodeIdxJson)),
-                clientv3.OpPut(fmt.Sprintf("%s/config/inode/%d/%d", etcdPrefix, poolId, nextId), string(inodeCfgJson)),
-            ).Commit()
-            cancel()
-            if (err != nil)
-            {
-                return nil, status.Error(codes.Internal, "failed to commit transaction in etcd: "+err.Error())
-            }
-            if (txnResp.Succeeded)
-            {
-                imageId = nextId
-                break
-            }
-            // Start over if the transaction fails
+            return nil, err
         }
     }
 
@@ -299,97 +238,12 @@ func (cs *ControllerServer) DeleteVolume(ctx context.Context, req *csi.DeleteVol
     }
     volName := ctxVars["name"]
 
-    _, etcdUrl, etcdPrefix := GetConnectionParams(ctxVars)
-    if (len(etcdUrl) == 0)
-    {
-        return nil, status.Error(codes.InvalidArgument, "no etcdUrl in storage class configuration and no etcd_address in vitastor.conf")
-    }
+    ctxVars, _, _ = GetConnectionParams(ctxVars)
 
-    cli, err := clientv3.New(clientv3.Config{
-        DialTimeout: ETCD_TIMEOUT,
-        Endpoints: etcdUrl,
-    })
+    _, err = invokeCLI(ctxVars, []string{ "rm", volName })
     if (err != nil)
     {
-        return nil, status.Error(codes.Internal, "failed to connect to etcd at "+strings.Join(etcdUrl, ",")+": "+err.Error())
-    }
-    defer cli.Close()
-
-    // Find inode by name
-    ctx, cancel := context.WithTimeout(context.Background(), ETCD_TIMEOUT)
-    resp, err := cli.Get(ctx, etcdPrefix+"/index/image/"+volName)
-    cancel()
-    if (err != nil)
-    {
-        return nil, status.Error(codes.Internal, "failed to read key from etcd: "+err.Error())
-    }
-    if (len(resp.Kvs) == 0)
-    {
-        return nil, status.Error(codes.NotFound, "volume "+volName+" does not exist")
-    }
-    var idx InodeIndex
-    err = json.Unmarshal(resp.Kvs[0].Value, &idx)
-    if (err != nil)
-    {
-        return nil, status.Error(codes.Internal, "invalid /index/image/"+volName+" key in etcd: "+err.Error())
-    }
-
-    // Get inode config
-    inodeCfgKey := fmt.Sprintf("%s/config/inode/%d/%d", etcdPrefix, idx.PoolId, idx.Id)
-    ctx, cancel = context.WithTimeout(context.Background(), ETCD_TIMEOUT)
-    resp, err = cli.Get(ctx, inodeCfgKey)
-    cancel()
-    if (err != nil)
-    {
-        return nil, status.Error(codes.Internal, "failed to read key from etcd: "+err.Error())
-    }
-    if (len(resp.Kvs) == 0)
-    {
-        return nil, status.Error(codes.NotFound, "volume "+volName+" does not exist")
-    }
-    var inodeCfg InodeConfig
-    err = json.Unmarshal(resp.Kvs[0].Value, &inodeCfg)
-    if (err != nil)
-    {
-        return nil, status.Error(codes.Internal, "invalid "+inodeCfgKey+" key in etcd: "+err.Error())
-    }
-
-    // Delete inode data by invoking vitastor-cli
-    args := []string{
-        "rm-data", "--etcd_address", strings.Join(etcdUrl, ","),
-        "--pool", fmt.Sprintf("%d", idx.PoolId),
-        "--inode", fmt.Sprintf("%d", idx.Id),
-    }
-    if (ctxVars["configPath"] != "")
-    {
-        args = append(args, "--config_path", ctxVars["configPath"])
-    }
-    c := exec.Command("/usr/bin/vitastor-cli", args...)
-    var stderr bytes.Buffer
-    c.Stdout = nil
-    c.Stderr = &stderr
-    err = c.Run()
-    stderrStr := string(stderr.Bytes())
-    if (err != nil)
-    {
-        klog.Errorf("vitastor-cli rm-data failed: %s, status %s\n", stderrStr, err)
-        return nil, status.Error(codes.Internal, stderrStr+" (status "+err.Error()+")")
-    }
-
-    // Delete inode config in etcd
-    ctx, cancel = context.WithTimeout(context.Background(), ETCD_TIMEOUT)
-    txnResp, err := cli.Txn(ctx).Then(
-        clientv3.OpDelete(fmt.Sprintf("%s/index/image/%s", etcdPrefix, volName)),
-        clientv3.OpDelete(fmt.Sprintf("%s/config/inode/%d/%d", etcdPrefix, idx.PoolId, idx.Id)),
-    ).Commit()
-    cancel()
-    if (err != nil)
-    {
-        return nil, status.Error(codes.Internal, "failed to delete keys in etcd: "+err.Error())
-    }
-    if (!txnResp.Succeeded)
-    {
-        return nil, status.Error(codes.Internal, "failed to delete keys in etcd: transaction failed")
+        return nil, err
     }
 
     return &csi.DeleteVolumeResponse{}, nil

debian/changelog

@@ -1,10 +1,10 @@
-vitastor (0.8.3-1) unstable; urgency=medium
+vitastor (0.8.8-1) unstable; urgency=medium
 
   * Bugfixes
 
  -- Vitaliy Filippov <vitalif@yourcmc.ru>  Fri, 03 Jun 2022 02:09:44 +0300
 
-vitastor (0.8.3-1) unstable; urgency=medium
+vitastor (0.8.8-1) unstable; urgency=medium
 
   * Implement NFS proxy
   * Add documentation

debian/vitastor.Dockerfile

@@ -34,8 +34,8 @@ RUN set -e -x; \
     mkdir -p /root/packages/vitastor-$REL; \
     rm -rf /root/packages/vitastor-$REL/*; \
     cd /root/packages/vitastor-$REL; \
-    cp -r /root/vitastor vitastor-0.8.3; \
-    cd vitastor-0.8.3; \
+    cp -r /root/vitastor vitastor-0.8.8; \
+    cd vitastor-0.8.8; \
     ln -s /root/fio-build/fio-*/ ./fio; \
     FIO=$(head -n1 fio/debian/changelog | perl -pe 's/^.*\((.*?)\).*$/$1/'); \
     ls /usr/include/linux/raw.h || cp ./debian/raw.h /usr/include/linux/raw.h; \
@@ -48,8 +48,8 @@ RUN set -e -x; \
     rm -rf a b; \
     echo "dep:fio=$FIO" > debian/fio_version; \
     cd /root/packages/vitastor-$REL; \
-    tar --sort=name --mtime='2020-01-01' --owner=0 --group=0 --exclude=debian -cJf vitastor_0.8.3.orig.tar.xz vitastor-0.8.3; \
-    cd vitastor-0.8.3; \
+    tar --sort=name --mtime='2020-01-01' --owner=0 --group=0 --exclude=debian -cJf vitastor_0.8.8.orig.tar.xz vitastor-0.8.8; \
+    cd vitastor-0.8.8; \
     V=$(head -n1 debian/changelog | perl -pe 's/^.*\((.*?)\).*$/$1/'); \
     DEBFULLNAME="Vitaliy Filippov <vitalif@yourcmc.ru>" dch -D $REL -v "$V""$REL" "Rebuild for $REL"; \
     DEB_BUILD_OPTIONS=nocheck dpkg-buildpackage --jobs=auto -sa; \

docs/config.en.md

@@ -17,14 +17,16 @@ 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.
+- Command line of Vitastor components: OSD (when you run it without vitastor-disk),
+  mon, fio and QEMU options, OpenStack/Proxmox/etc configuration. The latter
+  doesn't allow to set all variables directly, but it allows to override the
+  configuration file and set everything you need inside it.
+- OSD superblocks created by [vitastor-disk](../usage/disk.en.md) contain
+  primarily disk layout parameters of specific OSDs. In fact, these parameters
+  are automatically passed into the command line of vitastor-osd process, so
+  they have the same "status" as command-line parameters.
 
 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>`.
 
 ## Parameter Reference

docs/config.ru.md

@@ -19,14 +19,17 @@
 - 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 командной строке компонентов Vitastor: OSD (при ручном запуске без vitastor-disk),
+  монитора, опциях fio и QEMU, настроек OpenStack, Proxmox и т.п. Последние,
+  как правило, не включают полный набор параметров напрямую, но позволяют
+  определить путь к файлу конфигурации и задать любые параметры в нём.
+- В суперблоке OSD, записываемом [vitastor-disk](../usage/disk.ru.md) - параметры,
+  связанные с дисковым форматом и с этим конкретным OSD. На самом деле,
+  при запуске OSD эти параметры автоматически передаются в командную строку
+  процесса vitastor-osd, то есть по "статусу" они эквивалентны параметрам
+  командной строки OSD.
 
 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>`.
 
 ## Список параметров

docs/config/network.en.md

@@ -19,6 +19,7 @@ between clients, OSDs and etcd.
 - [rdma_max_sge](#rdma_max_sge)
 - [rdma_max_msg](#rdma_max_msg)
 - [rdma_max_recv](#rdma_max_recv)
+- [rdma_max_send](#rdma_max_send)
 - [peer_connect_interval](#peer_connect_interval)
 - [peer_connect_timeout](#peer_connect_timeout)
 - [osd_idle_timeout](#osd_idle_timeout)
@@ -74,6 +75,12 @@ 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.
 
+Remember that you also have to configure your network switches if you use
+RoCE/RoCEv2, otherwise you may experience unstable performance. Refer to
+the manual of your network vendor for details about setting up the switch
+for RoCEv2 correctly. Usually it means setting up Lossless Ethernet with
+PFC (Priority Flow Control) and ECN (Explicit Congestion Notification).
+
 ## rdma_port_num
 
 - Type: integer
@@ -116,20 +123,30 @@ required to change this parameter.
 ## rdma_max_msg
 
 - Type: integer
-- Default: 1048576
+- Default: 132096
 
 Maximum size of a single RDMA send or receive operation in bytes.
 
 ## rdma_max_recv
 
+- Type: integer
+- Default: 16
+
+Maximum number of RDMA receive buffers per connection (RDMA requires
+preallocated buffers to receive data). Each buffer is `rdma_max_msg` bytes
+in size. So this setting directly affects memory usage: a single Vitastor
+RDMA client uses `rdma_max_recv * rdma_max_msg * OSD_COUNT` bytes of memory.
+Default is roughly 2 MB * number of OSDs.
+
+## rdma_max_send
+
 - Type: integer
 - Default: 8
 
-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.
+Maximum number of outstanding RDMA send operations per connection. Should be
+less than `rdma_max_recv` so the receiving side doesn't run out of buffers.
+Doesn't affect memory usage - additional memory isn't allocated for send
+operations.
 
 ## peer_connect_interval
 

docs/config/network.ru.md

@@ -19,6 +19,7 @@
 - [rdma_max_sge](#rdma_max_sge)
 - [rdma_max_msg](#rdma_max_msg)
 - [rdma_max_recv](#rdma_max_recv)
+- [rdma_max_send](#rdma_max_send)
 - [peer_connect_interval](#peer_connect_interval)
 - [peer_connect_timeout](#peer_connect_timeout)
 - [osd_idle_timeout](#osd_idle_timeout)
@@ -78,6 +79,13 @@ Implicit On-Demand Paging (Implicit ODP) и Scatter/Gather (SG). Наприме
 суперпользователя, чтобы посмотреть список доступных RDMA-устройств, их
 параметры и возможности.
 
+Обратите внимание, что если вы используете RoCE/RoCEv2, вам также необходимо
+правильно настроить для него коммутаторы, иначе вы можете столкнуться с
+нестабильной производительностью. Подробную информацию о настройке
+коммутатора для RoCEv2 ищите в документации производителя. Обычно это
+подразумевает настройку сети без потерь на основе PFC (Priority Flow
+Control) и ECN (Explicit Congestion Notification).
+
 ## rdma_port_num
 
 - Тип: целое число
@@ -121,22 +129,32 @@ OSD в любом случае согласовывают реальное зн
 ## rdma_max_msg
 
 - Тип: целое число
-- Значение по умолчанию: 1048576
+- Значение по умолчанию: 132096
 
 Максимальный размер одной RDMA-операции отправки или приёма.
 
 ## rdma_max_recv
 
+- Тип: целое число
+- Значение по умолчанию: 16
+
+Максимальное число буферов для RDMA-приёма данных на одно соединение
+(RDMA требует заранее выделенных буферов для приёма данных). Каждый буфер
+имеет размер `rdma_max_msg` байт. Таким образом, настройка прямо влияет на
+потребление памяти - один Vitastor-клиент с RDMA использует
+`rdma_max_recv * rdma_max_msg * ЧИСЛО_OSD` байт памяти, по умолчанию -
+примерно 2 МБ * число OSD.
+
+## rdma_max_send
+
 - Тип: целое число
 - Значение по умолчанию: 8
 
-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.
+Максимальное число RDMA-операций отправки, отправляемых в очередь одного
+соединения. Желательно, чтобы оно было меньше `rdma_max_recv`, чтобы
+у принимающей стороны в процессе работы не заканчивались буферы на приём.
+Не влияет на потребление памяти - дополнительная память на операции отправки
+не выделяется.
 
 ## peer_connect_interval
 

docs/config/osd.en.md

@@ -17,6 +17,7 @@ initialization and can be changed with an OSD restart.
 - [autosync_interval](#autosync_interval)
 - [autosync_writes](#autosync_writes)
 - [recovery_queue_depth](#recovery_queue_depth)
+- [recovery_pg_switch](#recovery_pg_switch)
 - [recovery_sync_batch](#recovery_sync_batch)
 - [readonly](#readonly)
 - [no_recovery](#no_recovery)
@@ -115,6 +116,16 @@ 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.
 
+## recovery_pg_switch
+
+- Type: integer
+- Default: 128
+
+Number of recovery operations before switching to recovery of the next PG.
+The idea is to mix all PGs during recovery for more even space and load
+distribution but still benefit from recovery queue depth greater than 1.
+Degraded PGs are anyway scanned first.
+
 ## recovery_sync_batch
 
 - Type: integer

docs/config/osd.ru.md

@@ -18,6 +18,7 @@
 - [autosync_interval](#autosync_interval)
 - [autosync_writes](#autosync_writes)
 - [recovery_queue_depth](#recovery_queue_depth)
+- [recovery_pg_switch](#recovery_pg_switch)
 - [recovery_sync_batch](#recovery_sync_batch)
 - [readonly](#readonly)
 - [no_recovery](#no_recovery)
@@ -119,6 +120,17 @@ OSD, чтобы успевать очищать журнал - без них OSD
 для ускорения или замедления восстановления и перебалансировки данных, но
 в планах реализация других параметров.
 
+## recovery_pg_switch
+
+- Тип: целое число
+- Значение по умолчанию: 128
+
+Число операций восстановления перед переключением на восстановление другой PG.
+Идея заключается в том, чтобы восстанавливать все PG одновременно для более
+равномерного распределения места и нагрузки, но при этом всё равно выигрывать
+от глубины очереди восстановления, большей, чем 1. Деградированные PG в любом
+случае сканируются первыми.
+
 ## recovery_sync_batch
 
 - Тип: целое число

docs/config/src/network.yml

@@ -53,6 +53,12 @@
     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.
+
+    Remember that you also have to configure your network switches if you use
+    RoCE/RoCEv2, otherwise you may experience unstable performance. Refer to
+    the manual of your network vendor for details about setting up the switch
+    for RoCEv2 correctly. Usually it means setting up Lossless Ethernet with
+    PFC (Priority Flow Control) and ECN (Explicit Congestion Notification).
   info_ru: |
     Название RDMA-устройства для связи с Vitastor OSD (например, "rocep5s0f0").
     Имейте в виду, что поддержка RDMA в Vitastor требует функций устройства
@@ -61,6 +67,13 @@
     потому не поддерживаются в Vitastor. Запустите `ibv_devinfo -v` от имени
     суперпользователя, чтобы посмотреть список доступных RDMA-устройств, их
     параметры и возможности.
+
+    Обратите внимание, что если вы используете RoCE/RoCEv2, вам также необходимо
+    правильно настроить для него коммутаторы, иначе вы можете столкнуться с
+    нестабильной производительностью. Подробную информацию о настройке
+    коммутатора для RoCEv2 ищите в документации производителя. Обычно это
+    подразумевает настройку сети без потерь на основе PFC (Priority Flow
+    Control) и ECN (Explicit Congestion Notification).
 - name: rdma_port_num
   type: int
   default: 1
@@ -114,26 +127,39 @@
     так что менять этот параметр обычно не нужно.
 - name: rdma_max_msg
   type: int
-  default: 1048576
+  default: 132096
   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: 16
+  info: |
+    Maximum number of RDMA receive buffers per connection (RDMA requires
+    preallocated buffers to receive data). Each buffer is `rdma_max_msg` bytes
+    in size. So this setting directly affects memory usage: a single Vitastor
+    RDMA client uses `rdma_max_recv * rdma_max_msg * OSD_COUNT` bytes of memory.
+    Default is roughly 2 MB * number of OSDs.
+  info_ru: |
+    Максимальное число буферов для RDMA-приёма данных на одно соединение
+    (RDMA требует заранее выделенных буферов для приёма данных). Каждый буфер
+    имеет размер `rdma_max_msg` байт. Таким образом, настройка прямо влияет на
+    потребление памяти - один Vitastor-клиент с RDMA использует
+    `rdma_max_recv * rdma_max_msg * ЧИСЛО_OSD` байт памяти, по умолчанию -
+    примерно 2 МБ * число OSD.
+- name: rdma_max_send
   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.
+    Maximum number of outstanding RDMA send operations per connection. Should be
+    less than `rdma_max_recv` so the receiving side doesn't run out of buffers.
+    Doesn't affect memory usage - additional memory isn't allocated for send
+    operations.
   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.
+    Максимальное число RDMA-операций отправки, отправляемых в очередь одного
+    соединения. Желательно, чтобы оно было меньше `rdma_max_recv`, чтобы
+    у принимающей стороны в процессе работы не заканчивались буферы на приём.
+    Не влияет на потребление памяти - дополнительная память на операции отправки
+    не выделяется.
 - name: peer_connect_interval
   type: sec
   min: 1

docs/config/src/osd.yml

@@ -102,6 +102,20 @@
     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_pg_switch
+  type: int
+  default: 128
+  info: |
+    Number of recovery operations before switching to recovery of the next PG.
+    The idea is to mix all PGs during recovery for more even space and load
+    distribution but still benefit from recovery queue depth greater than 1.
+    Degraded PGs are anyway scanned first.
+  info_ru: |
+    Число операций восстановления перед переключением на восстановление другой PG.
+    Идея заключается в том, чтобы восстанавливать все PG одновременно для более
+    равномерного распределения места и нагрузки, но при этом всё равно выигрывать
+    от глубины очереди восстановления, большей, чем 1. Деградированные PG в любом
+    случае сканируются первыми.
 - name: recovery_sync_batch
   type: int
   default: 16

docs/installation/packages.en.md

@@ -9,7 +9,7 @@
 ## Debian
 
 - Trust Vitastor package signing key:
-  `wget -q -O - https://vitastor.io/debian/pubkey | sudo apt-key add -`
+  `wget https://vitastor.io/debian/pubkey.gpg -O /etc/apt/trusted.gpg.d/vitastor.gpg`
 - Add Vitastor package repository to your /etc/apt/sources.list:
   - Debian 11 (Bullseye/Sid): `deb https://vitastor.io/debian bullseye main`
   - Debian 10 (Buster): `deb https://vitastor.io/debian buster main`
@@ -20,16 +20,19 @@
 ## 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`
+  - CentOS 7: `yum install https://vitastor.io/rpms/centos/7/vitastor-release.rpm`
+  - CentOS 8: `dnf install https://vitastor.io/rpms/centos/8/vitastor-release.rpm`
+  - AlmaLinux 9 and other RHEL 9 clones (Rocky, Oracle...): `dnf install https://vitastor.io/rpms/centos/9/vitastor-release.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`
+  - RHEL 9 clones: not required
 - 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`
+  - RHEL 9 clones: optional, not required: `dnf install https://www.elrepo.org/elrepo-release-9.el9.elrepo.noarch.rpm`
+- Install packages: `yum/dnf install vitastor lpsolve etcd qemu-kvm` and optionally `kernel-ml` if you use elrepo-kernel
 
 ## Installation requirements
 

docs/installation/packages.ru.md

@@ -9,7 +9,7 @@
 ## Debian
 
 - Добавьте ключ репозитория Vitastor:
-  `wget -q -O - https://vitastor.io/debian/pubkey | sudo apt-key add -`
+  `wget https://vitastor.io/debian/pubkey.gpg -O /etc/apt/trusted.gpg.d/vitastor.gpg`
 - Добавьте репозиторий Vitastor в /etc/apt/sources.list:
   - Debian 11 (Bullseye/Sid): `deb https://vitastor.io/debian bullseye main`
   - Debian 10 (Buster): `deb https://vitastor.io/debian buster main`
@@ -20,8 +20,8 @@
 ## CentOS
 
 - Добавьте в систему репозиторий Vitastor:
-  - 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`
+  - CentOS 7: `yum install https://vitastor.io/rpms/centos/7/vitastor-release.rpm`
+  - CentOS 8: `dnf install https://vitastor.io/rpms/centos/8/vitastor-release.rpm`
 - Включите EPEL: `yum/dnf install epel-release`
 - Включите дополнительные репозитории CentOS:
   - CentOS 7: `yum install centos-release-scl`

docs/installation/proxmox.en.md

@@ -6,10 +6,10 @@
 
 # Proxmox VE
 
-To enable Vitastor support in Proxmox Virtual Environment (6.4-7.3 are supported):
+To enable Vitastor support in Proxmox Virtual Environment (6.4-7.4 are supported):
 
 - Add the corresponding Vitastor Debian repository into sources.list on Proxmox hosts:
-  buster for 6.4, bullseye for 7.3, pve7.1 for 7.1, pve7.2 for 7.2
+  buster for 6.4, bullseye for 7.4, pve7.1 for 7.1, pve7.2 for 7.2, pve7.3 for 7.3
 - Install vitastor-client, pve-qemu-kvm, pve-storage-vitastor (* or see note) packages from Vitastor repository
 - Define storage in `/etc/pve/storage.cfg` (see below)
 - Block network access from VMs to Vitastor network (to OSDs and etcd),

docs/installation/proxmox.ru.md

@@ -6,10 +6,10 @@
 
 # Proxmox
 
-Чтобы подключить Vitastor к Proxmox Virtual Environment (поддерживаются версии 6.4-7.3):
+Чтобы подключить Vitastor к Proxmox Virtual Environment (поддерживаются версии 6.4-7.4):
 
 - Добавьте соответствующий Debian-репозиторий Vitastor в sources.list на хостах Proxmox:
-  buster для 6.4, bullseye для 7.3, pve7.1 для 7.1, pve7.2 для 7.2
+  buster для 6.4, bullseye для 7.4, pve7.1 для 7.1, pve7.2 для 7.2, pve7.3 для 7.3
 - Установите пакеты vitastor-client, pve-qemu-kvm, pve-storage-vitastor (* или см. сноску) из репозитория Vitastor
 - Определите тип хранилища в `/etc/pve/storage.cfg` (см. ниже)
 - Обязательно заблокируйте доступ от виртуальных машин к сети Vitastor (OSD и etcd), т.к. Vitastor (пока) не поддерживает аутентификацию

docs/intro/quickstart.en.md

@@ -45,7 +45,9 @@ On the monitor hosts:
   }
   ```
 - Initialize OSDs:
-  - SSD-only: `vitastor-disk prepare /dev/sdXXX [/dev/sdYYY ...]`
+  - SSD-only: `vitastor-disk prepare /dev/sdXXX [/dev/sdYYY ...]`. You can add
+    `--disable_data_fsync off` to leave disk cache enabled if you use desktop
+    SSDs without capacitors.
   - Hybrid, SSD+HDD: `vitastor-disk prepare --hybrid /dev/sdXXX [/dev/sdYYY ...]`.
     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
@@ -53,7 +55,9 @@ On the monitor hosts:
     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](../config.en.md) for parameter descriptions.
-- If all your drives have capacitors, create global configuration in etcd: \
+- If all your drives have capacitors, and even if not, but if you ran `vitastor-disk`
+  without `--disable_data_fsync off` at the first step, then put the following
+  setting into etcd: \
   `etcdctl --endpoints=... put /vitastor/config/global '{"immediate_commit":"all"}'`
 - Start all OSDs: `systemctl start vitastor.target`
 
@@ -70,11 +74,15 @@ For EC pools the configuration should look like the following:
 
 ```
 etcdctl --endpoints=... put /vitastor/config/pools '{"2":{"name":"ecpool",
-  "scheme":"ec","pg_size":4,"parity_chunks":2,"pg_minsize":2,"pg_count":256,"failure_domain":"host"}`
+  "scheme":"ec","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.
 
+If you use HDDs you should also add `"block_size": 1048576` to pool configuration.
+The other option is to add it into /vitastor/config/global, in this case it will
+apply to all pools by default.
+
 ## Check cluster status
 
 `vitastor-cli status`

docs/intro/quickstart.ru.md

@@ -45,7 +45,9 @@
   }
   ```
 - Инициализуйте OSD:
-  - SSD: `vitastor-disk prepare /dev/sdXXX [/dev/sdYYY ...]`
+  - SSD: `vitastor-disk prepare /dev/sdXXX [/dev/sdYYY ...]`. Если вы используете
+    десктопные SSD без конденсаторов, можете оставить кэш включённым, добавив
+    опцию `--disable_data_fsync off`.
   - Гибридные, SSD+HDD: `vitastor-disk prepare --hybrid /dev/sdXXX [/dev/sdYYY ...]`.
     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
@@ -54,8 +56,11 @@
     для журналов, на SSD должно быть доступно свободное нераспределённое место.
 - Вы можете менять параметры OSD в юнитах systemd или в `vitastor.conf`. Описания параметров
   смотрите в [справке по конфигурации](../config.ru.md).
-- Если все ваши диски - серверные с конденсаторами, пропишите это в глобальную конфигурацию в etcd: \
-  `etcdctl --endpoints=... put /vitastor/config/global '{"immediate_commit":"all"}'`
+- Если все ваши диски - серверные с конденсаторами, и даже если нет, но при этом
+  вы не добавляли опцию `--disable_data_fsync off` на первом шаге, а `vitastor-disk`
+  не ругался на невозможность отключения кэша дисков, пропишите следующую настройку
+  в глобальную конфигурацию в etcd: \
+  `etcdctl --endpoints=... put /vitastor/config/global '{"immediate_commit":"all"}'`.
 - Запустите все OSD: `systemctl start vitastor.target`
 
 ## Создайте пул
@@ -71,11 +76,15 @@ etcdctl --endpoints=... put /vitastor/config/pools '{"1":{"name":"testpool",
 
 ```
 etcdctl --endpoints=... put /vitastor/config/pools '{"2":{"name":"ecpool",
-  "scheme":"ec","pg_size":4,"parity_chunks":2,"pg_minsize":2,"pg_count":256,"failure_domain":"host"}`
+  "scheme":"ec","pg_size":4,"parity_chunks":2,"pg_minsize":2,"pg_count":256,"failure_domain":"host"}}'
 ```
 
 После этого один из мониторов должен сконфигурировать PG, а OSD должны запустить их.
 
+Если вы используете HDD-диски, то добавьте в конфигурацию пулов опцию `"block_size": 1048576`.
+Также эту опцию можно добавить в /vitastor/config/global, в этом случае она будет
+применяться ко всем пулам по умолчанию.
+
 ## Проверьте состояние кластера
 
 `vitastor-cli status`

docs/performance/theoretical.en.md

@@ -35,15 +35,24 @@ Write amplification for 4 KB blocks is usually 3-5 in Vitastor:
 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.
 
+Implemented NVDIMM support can basically eliminate WA at all - all extra writes will
+go to DRAM memory. But this requires a test cluster with NVDIMM - please contact me
+if you want to provide me with such cluster for tests.
+
 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](understanding.en.md)
-and good server-grade SSD/NVMe drives, you should head for:
+In practice, using tests from [Understanding Performance](understanding.en.md), decent TCP network,
+good server-grade SSD/NVMe drives and disabled CPU power saving, you should head for:
 - At least 5000 T1Q1 replicated read and write iops (maximum 0.2ms latency)
+- At least 5000 T1Q1 EC read IOPS and at least 2200 EC write IOPS (maximum 0.45ms 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
 
 Lower results may mean that you have bad drives, bad network or some kind of misconfiguration.
+
+Current latency records:
+- 9668 T1Q1 replicated write iops (0.103 ms latency) with TCP and NVMe
+- 9143 T1Q1 replicated read iops (0.109 ms latency) with TCP and NVMe

docs/performance/theoretical.ru.md

@@ -36,6 +36,25 @@ WA (мультипликатор записи) для 4 КБ блоков в Vit
 Если вы найдёте SSD, хорошо работающий с 512-байтными блоками данных (Optane?),
 то 1, 3 и 4 можно снизить до 512 байт (1/8 от размера данных) и получить WA всего 2.375.
 
+Если реализовать поддержку NVDIMM, то WA можно, условно говоря, ликвидировать вообще - все
+дополнительные операции записи смогут обслуживаться DRAM памятью. Но для этого необходим
+тестовый кластер с NVDIMM - пишите, если готовы предоставить такой для тестов.
+
 Кроме того, WA снижается при использовании отложенного/ленивого сброса при параллельной
 нагрузке, т.к. блоки журнала записываются на диск только когда они заполняются или явным
 образом запрашивается fsync.
+
+## На практике
+
+На практике, используя тесты fio со страницы [Понимание сути производительности систем хранения](understanding.ru.md),
+нормальную TCP-сеть, хорошие серверные SSD/NVMe, при отключённом энергосбережении процессоров вы можете рассчитывать на:
+- От 5000 IOPS в 1 поток (T1Q1) и на чтение, и на запись при использовании репликации (задержка до 0.2мс)
+- От 5000 IOPS в 1 поток (T1Q1) на чтение и 2200 IOPS в 1 поток на запись при использовании EC (задержка до 0.45мс)
+- От 80000 IOPS на чтение в параллельном режиме на 1 ядро, от 30000 IOPS на запись на 1 ядро (на 1 OSD)
+- Скорость параллельного линейного чтения и записи, равная меньшему значению из скорости дисков или сети
+
+Худшие результаты означают, что у вас либо медленные диски, либо медленная сеть, либо что-то неправильно настроено.
+
+Зафиксированный на данный момент рекорд задержки:
+- 9668 IOPS (0.103 мс задержка) в 1 поток (T1Q1) на запись с TCP и NVMe при использовании репликации
+- 9143 IOPS (0.109 мс задержка) в 1 поток (T1Q1) на чтение с TCP и NVMe при использовании репликации

docs/usage/cli.en.md

@@ -14,6 +14,7 @@ It supports the following commands:
 - [df](#df)
 - [ls](#ls)
 - [create](#create)
+- [snap-create](#create)
 - [modify](#modify)
 - [rm](#rm)
 - [flatten](#flatten)
@@ -123,6 +124,8 @@ 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.
 
+See also about [how to export snapshots](qemu.en.md#exporting-snapshots).
+
 ## modify
 
 `vitastor-cli modify <name> [--rename <new-name>] [--resize <size>] [--readonly | --readwrite] [-f|--force]`

docs/usage/cli.ru.md

@@ -15,6 +15,7 @@ vitastor-cli - интерфейс командной строки для адм
 - [df](#df)
 - [ls](#ls)
 - [create](#create)
+- [snap-create](#create)
 - [modify](#modify)
 - [rm](#rm)
 - [flatten](#flatten)
@@ -126,6 +127,8 @@ vitastor-cli snap-create [-p|--pool <id|name>] <image>@<snapshot>
 Создать снимок образа `<name>` (можно использовать любую форму команды). Снимок можно создавать без остановки
 клиентов, если пишущий клиент максимум 1.
 
+Смотрите также информацию о том, [как экспортировать снимки](qemu.ru.md#экспорт-снимков).
+
 ## modify
 
 `vitastor-cli modify <name> [--rename <new-name>] [--resize <size>] [--readonly | --readwrite] [-f|--force]`

docs/usage/disk.en.md

@@ -1,4 +1,4 @@
-[Documentation](../../README.md#documentation) → Usage → Disk Tool
+[Documentation](../../README.md#documentation) → Usage → Disk management tool
 
 -----
 

docs/usage/disk.ru.md

@@ -1,4 +1,4 @@
-[Документация](../../README-ru.md#документация) → Использование → Управление дисками
+[Документация](../../README-ru.md#документация) → Использование → Инструмент управления дисками
 
 -----
 

docs/usage/qemu.en.md

@@ -46,3 +46,40 @@ qemu-img convert -f qcow2 debian10.qcow2 -p -O raw 'vitastor:etcd_host=192.168.7
 
 You can also specify `:pool=<POOL>:inode=<INODE>:size=<SIZE>` instead of `:image=<IMAGE>`
 if you don't want to use inode metadata.
+
+### Exporting snapshots
+
+Starting with 0.8.4, you can also export individual layers (snapshot diffs) using `qemu-img`.
+
+Suppose you have an image `testimg` and a snapshot `testimg@0` created with `vitastor-cli snap-create testimg@0`.
+
+Then you can export the `testimg@0` snapshot and the data written to `testimg` after creating
+the snapshot separately using the following commands (key points are using `skip-parents=1` and
+`-B backing_file` option):
+
+```
+qemu-img convert -f raw 'vitastor:etcd_host=192.168.7.2\:2379/v3:image=testimg@0' \
+    -O qcow2 testimg_0.qcow2
+
+qemu-img convert -f raw 'vitastor:etcd_host=192.168.7.2\:2379/v3:image=testimg:skip-parents=1' \
+    -O qcow2 -o 'cluster_size=4k' -B testimg_0.qcow2 testimg.qcow2
+```
+
+In fact, with `cluster_size=4k` any QCOW2 file can be used instead `-B testimg_0.qcow2`, even an empty one.
+
+QCOW2 `cluster_size=4k` option is required if you want `testimg.qcow2` to contain only the data
+overwritten  **exactly** in the child layer. With the default 64 KB QCOW2 cluster size you'll
+get a bit of extra data from parent layers, i.e. a 4 KB overwrite will result in `testimg.qcow2`
+containing 64 KB of data. And this extra data will be taken by `qemu-img` from the file passed
+in `-B` option, so you really need 4 KB cluster if you use an empty image in `-B`.
+
+After this procedure you'll get two chained QCOW2 images. To detach `testimg.qcow2` from
+its parent, run:
+
+```
+qemu-img rebase -u -b '' testimg.qcow2
+```
+
+This can be used for backups. Just note that exporting an image that is currently being written to
+is of course unsafe and doesn't produce a consistent result, so only export snapshots if you do this
+on a live VM.

docs/usage/qemu.ru.md

@@ -50,3 +50,40 @@ qemu-img convert -f qcow2 debian10.qcow2 -p -O raw 'vitastor:etcd_host=10.115.0.
 
 Если вы не хотите обращаться к образу по имени, вместо `:image=<IMAGE>` можно указать номер пула, номер инода и размер:
 `:pool=<POOL>:inode=<INODE>:size=<SIZE>`.
+
+### Экспорт снимков
+
+Начиная с 0.8.4 вы можете экспортировать отдельные слои (изменения в снимках) с помощью `qemu-img`.
+
+Допустим, что у вас есть образ `testimg` и его снимок `testimg@0`, созданный с помощью `vitastor-cli snap-create testimg@0`.
+
+Тогда вы можете выгрузить снимок `testimg@0` и данные, изменённые в `testimg` после создания снимка, отдельно,
+с помощью следующих команд (ключевые моменты - использование `skip-parents=1` и опции `-B backing_file.qcow2`):
+
+```
+qemu-img convert -f raw 'vitastor:etcd_host=192.168.7.2\:2379/v3:image=testimg@0' \
+    -O qcow2 testimg_0.qcow2
+
+qemu-img convert -f raw 'vitastor:etcd_host=192.168.7.2\:2379/v3:image=testimg:skip-parents=1' \
+    -O qcow2 -o 'cluster_size=4k' -B testimg_0.qcow2 testimg.qcow2
+```
+
+На самом деле, с `cluster_size=4k` вместо `-B testimg_0.qcow2` можно использовать любой qcow2-файл,
+даже пустой.
+
+Опция QCOW2 `cluster_size=4k` нужна, если вы хотите, чтобы `testimg.qcow2` содержал **в точности**
+данные, перезаписанные в дочернем слое. С размером кластера QCOW2 по умолчанию, составляющим 64 КБ,
+вы получите немного "лишних" данных из родительских слоёв - перезапись 4 КБ будет приводить к тому,
+что в `testimg.qcow2` будет появляться 64 КБ данных. Причём "лишние" данные qemu-img будет брать
+как раз из файла, указанного в опции `-B`, так что если там указан пустой образ, кластер обязан быть 4 КБ.
+
+После данной процедуры вы получите два QCOW2-образа, связанных в цепочку. Чтобы "отцепить" образ
+`testimg.qcow2` от базового, выполните:
+
+```
+qemu-img rebase -u -b '' testimg.qcow2
+```
+
+Это можно использовать для резервного копирования. Только помните, что экспортировать образ, в который
+в то же время идёт запись, небезопасно - результат чтения не будет целостным. Так что если вы работаете
+с активными виртуальными машинами, экспортируйте только их снимки, но не сам образ.

mon/PGUtil.js

@@ -21,7 +21,7 @@ function add_pg_history(new_pg_history, new_pg, prev_pgs, prev_pg_history, old_p
     {
         for (const pg of oh.osd_sets)
         {
-            nh.osd_sets[pg.join(' ')] = pg;
+            nh.osd_sets[pg.join(' ')] = pg.map(osd_num => Number(osd_num));
         }
     }
     if (oh && oh.all_peers && oh.all_peers.length)
@@ -43,16 +43,16 @@ function finish_pg_history(merged_history)
     merged_history.all_peers = Object.values(merged_history.all_peers);
 }
 
-function scale_pg_count(prev_pgs, prev_pg_history, new_pg_history, new_pg_count)
+function scale_pg_count(prev_pgs, real_prev_pgs, prev_pg_history, new_pg_history, new_pg_count)
 {
-    const old_pg_count = prev_pgs.length;
+    const old_pg_count = real_prev_pgs.length;
     // Add all possibly intersecting PGs to the history of new PGs
     if (!(new_pg_count % old_pg_count))
     {
         // New PG count is a multiple of old PG count
         for (let i = 0; i < new_pg_count; i++)
         {
-            add_pg_history(new_pg_history, i, prev_pgs, prev_pg_history, i % old_pg_count);
+            add_pg_history(new_pg_history, i, real_prev_pgs, prev_pg_history, i % old_pg_count);
             finish_pg_history(new_pg_history[i]);
         }
     }
@@ -64,7 +64,7 @@ function scale_pg_count(prev_pgs, prev_pg_history, new_pg_history, new_pg_count)
         {
             for (let j = 0; j < mul; j++)
             {
-                add_pg_history(new_pg_history, i, prev_pgs, prev_pg_history, i+j*new_pg_count);
+                add_pg_history(new_pg_history, i, real_prev_pgs, prev_pg_history, i+j*new_pg_count);
             }
             finish_pg_history(new_pg_history[i]);
         }
@@ -76,7 +76,7 @@ function scale_pg_count(prev_pgs, prev_pg_history, new_pg_history, new_pg_count)
         let merged_history = {};
         for (let i = 0; i < old_pg_count; i++)
         {
-            add_pg_history(merged_history, 1, prev_pgs, prev_pg_history, i);
+            add_pg_history(merged_history, 1, real_prev_pgs, prev_pg_history, i);
         }
         finish_pg_history(merged_history[1]);
         for (let i = 0; i < new_pg_count; i++)
@@ -90,15 +90,15 @@ function scale_pg_count(prev_pgs, prev_pg_history, new_pg_history, new_pg_count)
         new_pg_history[i] = null;
     }
     // Just for the lp_solve optimizer - pick a "previous" PG for each "new" one
-    if (old_pg_count < new_pg_count)
+    if (prev_pgs.length < new_pg_count)
     {
-        for (let i = old_pg_count; i < new_pg_count; i++)
+        for (let i = prev_pgs.length; i < new_pg_count; i++)
         {
-            prev_pgs[i] = prev_pgs[i % old_pg_count];
+            prev_pgs[i] = prev_pgs[i % prev_pgs.length];
         }
     }
-    else if (old_pg_count > new_pg_count)
+    else if (prev_pgs.length > new_pg_count)
     {
-        prev_pgs.splice(new_pg_count, old_pg_count-new_pg_count);
+        prev_pgs.splice(new_pg_count, prev_pgs.length-new_pg_count);
     }
 }

mon/lp-optimizer.js

@@ -550,8 +550,8 @@ function random_combinations(osd_tree, pg_size, count, ordered)
         seed ^= seed << 5;
         return seed + 2147483648;
     };
-    const hosts = Object.keys(osd_tree).sort();
     const osds = Object.keys(osd_tree).reduce((a, c) => { a[c] = Object.keys(osd_tree[c]).sort(); return a; }, {});
+    const hosts = Object.keys(osd_tree).sort().filter(h => osds[h].length > 0);
     const r = {};
     // Generate random combinations including each OSD at least once
     for (let h = 0; h < hosts.length; h++)

mon/make-etcd

@@ -79,7 +79,7 @@ StartLimitInterval=0
 RestartSec=10
 
 [Install]
-WantedBy=local.target
+WantedBy=multi-user.target
 `);
     await system(`useradd etcd`);
     await system(`systemctl daemon-reload`);

mon/mon-main.js

@@ -13,7 +13,7 @@ for (let i = 2; i < process.argv.length; i++)
     {
         console.error('USAGE: '+process.argv[0]+' '+process.argv[1]+' [--verbose 1]'+
             ' [--etcd_address "http://127.0.0.1:2379,..."] [--config_path /etc/vitastor/vitastor.conf]'+
-            ' [--etcd_prefix "/vitastor"] [--etcd_start_timeout 5]');
+            ' [--etcd_prefix "/vitastor"] [--etcd_start_timeout 5] [--restart_interval 5]');
         process.exit();
     }
     else if (process.argv[i].substr(0, 2) == '--')

mon/mon.js

@@ -51,8 +51,9 @@ const etcd_tree = {
             // THIS IS JUST A POOR MAN'S CONFIG DOCUMENTATION
             // etcd connection
             config_path: "/etc/vitastor/vitastor.conf",
-            etcd_address: "10.0.115.10:2379/v3",
             etcd_prefix: "/vitastor",
+            // etcd connection - configurable online
+            etcd_address: "10.0.115.10:2379/v3",
             // mon
             etcd_mon_ttl: 30, // min: 10
             etcd_mon_timeout: 1000, // ms. min: 0
@@ -70,14 +71,15 @@ const etcd_tree = {
             rdma_gid_index: 0,
             rdma_mtu: 4096,
             rdma_max_sge: 128,
-            rdma_max_send: 32,
-            rdma_max_recv: 8,
-            rdma_max_msg: 1048576,
-            log_level: 0,
+            rdma_max_send: 8,
+            rdma_max_recv: 16,
+            rdma_max_msg: 132096,
             block_size: 131072,
             disk_alignment: 4096,
             bitmap_granularity: 4096,
             immediate_commit: false, // 'all' or 'small'
+            // client and osd - configurable online
+            log_level: 0,
             client_dirty_limit: 33554432,
             peer_connect_interval: 5, // seconds. min: 1
             peer_connect_timeout: 5, // seconds. min: 1
@@ -95,18 +97,19 @@ const etcd_tree = {
             osd_network: null, // "192.168.7.0/24" or an array of masks
             bind_address: "0.0.0.0",
             bind_port: 0,
+            readonly: false,
+            osd_memlock: false,
+            // osd - configurable online
             autosync_interval: 5,
             autosync_writes: 128,
             client_queue_depth: 128, // unused
             recovery_queue_depth: 4,
             recovery_sync_batch: 16,
-            readonly: false,
             no_recovery: false,
             no_rebalance: false,
             print_stats_interval: 3,
             slow_log_interval: 10,
             inode_vanish_time: 60,
-            osd_memlock: false,
             // blockstore - fixed in superblock
             block_size,
             disk_alignment,
@@ -125,14 +128,15 @@ const etcd_tree = {
             meta_offset,
             disable_meta_fsync,
             disable_device_lock,
-            // blockstore - configurable
-            max_write_iodepth,
-            min_flusher_count: 1,
-            max_flusher_count: 256,
+            // blockstore - configurable offline
             inmemory_metadata,
             inmemory_journal,
             journal_sector_buffer_count,
             journal_no_same_sector_overwrites,
+            // blockstore - configurable online
+            max_write_iodepth,
+            min_flusher_count: 1,
+            max_flusher_count: 256,
             throttle_small_writes: false,
             throttle_target_iops: 100,
             throttle_target_mbs: 100,
@@ -261,7 +265,7 @@ const etcd_tree = {
             /* <pool_id>: {
                 <pg_id>: {
                     primary: osd_num_t,
-                    state: ("starting"|"peering"|"peered"|"incomplete"|"active"|"repeering"|"stopping"|"offline"|
+                    state: ("starting"|"peering"|"incomplete"|"active"|"repeering"|"stopping"|"offline"|
                         "degraded"|"has_incomplete"|"has_degraded"|"has_misplaced"|"has_unclean"|
                         "has_invalid"|"left_on_dead")[],
                 }
@@ -557,7 +561,7 @@ class Mon
         }
         if (!this.ws)
         {
-            this.die('Failed to open etcd watch websocket');
+            await this.die('Failed to open etcd watch websocket');
         }
         const cur_addr = this.selected_etcd_url;
         this.ws_alive = true;
@@ -663,12 +667,15 @@ class Mon
     async save_last_clean()
     {
         // last_clean_pgs is used to avoid extra data move when observing a series of changes in the cluster
+        const new_clean_pgs = { items: {} };
+    next_pool:
         for (const pool_id in this.state.config.pools)
         {
+            new_clean_pgs.items[pool_id] = (this.state.history.last_clean_pgs.items||{})[pool_id];
             const pool_cfg = this.state.config.pools[pool_id];
             if (!this.validate_pool_cfg(pool_id, pool_cfg, false))
             {
-                continue;
+                continue next_pool;
             }
             for (let pg_num = 1; pg_num <= pool_cfg.pg_count; pg_num++)
             {
@@ -677,17 +684,18 @@ class Mon
                     !(this.state.pg.state[pool_id][pg_num].state instanceof Array))
                 {
                     // Unclean
-                    return;
+                    continue next_pool;
                 }
                 let st = this.state.pg.state[pool_id][pg_num].state.join(',');
                 if (st != 'active' && st != 'active,left_on_dead' && st != 'left_on_dead,active')
                 {
                     // Unclean
-                    return;
+                    continue next_pool;
                 }
             }
+            new_clean_pgs.items[pool_id] = this.state.config.pgs.items[pool_id];
         }
-        this.state.history.last_clean_pgs = JSON.parse(JSON.stringify(this.state.config.pgs));
+        this.state.history.last_clean_pgs = new_clean_pgs;
         await this.etcd_call('/kv/txn', {
             success: [ { requestPut: {
                 key: b64(this.etcd_prefix+'/history/last_clean_pgs'),
@@ -720,7 +728,7 @@ class Mon
             const res = await this.etcd_call('/lease/keepalive', { ID: this.etcd_lease_id }, this.config.etcd_mon_timeout, this.config.etcd_mon_retries);
             if (!res.result.TTL)
             {
-                this.die('Lease expired');
+                await this.die('Lease expired');
             }
         }, this.config.etcd_mon_timeout);
         if (!this.signals_set)
@@ -733,9 +741,32 @@ class Mon
 
     async on_stop(status)
     {
-        clearInterval(this.lease_timer);
-        await this.etcd_call('/lease/revoke', { ID: this.etcd_lease_id }, this.config.etcd_mon_timeout, this.config.etcd_mon_retries);
-        process.exit(status);
+        if (this.ws_keepalive_timer)
+        {
+            clearInterval(this.ws_keepalive_timer);
+            this.ws_keepalive_timer = null;
+        }
+        if (this.lease_timer)
+        {
+            clearInterval(this.lease_timer);
+            this.lease_timer = null;
+        }
+        if (this.etcd_lease_id)
+        {
+            const lease_id = this.etcd_lease_id;
+            this.etcd_lease_id = null;
+            await this.etcd_call('/lease/revoke', { ID: lease_id }, this.config.etcd_mon_timeout, this.config.etcd_mon_retries);
+        }
+        if (!status || !this.initConfig.restart_interval)
+        {
+            process.exit(status);
+        }
+        else
+        {
+            console.log('Restarting after '+this.initConfig.restart_interval+' seconds');
+            await new Promise(ok => setTimeout(ok, this.initConfig.restart_interval*1000));
+            await this.start();
+        }
     }
 
     async become_master()
@@ -948,7 +979,7 @@ class Mon
         return alive_set[this.rng() % alive_set.length];
     }
 
-    save_new_pgs_txn(request, pool_id, up_osds, osd_tree, prev_pgs, new_pgs, pg_history)
+    save_new_pgs_txn(save_to, 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 = {};
@@ -1001,14 +1032,14 @@ class Mon
                 });
             }
         }
-        this.state.config.pgs.items = this.state.config.pgs.items || {};
+        save_to.items = save_to.items || {};
         if (!new_pgs.length)
         {
-            delete this.state.config.pgs.items[pool_id];
+            delete save_to.items[pool_id];
         }
         else
         {
-            this.state.config.pgs.items[pool_id] = pg_items;
+            save_to.items[pool_id] = pg_items;
         }
     }
 
@@ -1152,6 +1183,7 @@ class Mon
         if (this.state.config.pgs.hash != tree_hash)
         {
             // Something has changed
+            const new_config_pgs = JSON.parse(JSON.stringify(this.state.config.pgs));
             const etcd_request = { compare: [], success: [] };
             for (const pool_id in (this.state.config.pgs||{}).items||{})
             {
@@ -1172,7 +1204,7 @@ class Mon
                     etcd_request.success.push({ requestDeleteRange: {
                         key: b64(this.etcd_prefix+'/pool/stats/'+pool_id),
                     } });
-                    this.save_new_pgs_txn(etcd_request, pool_id, up_osds, osd_tree, prev_pgs, [], []);
+                    this.save_new_pgs_txn(new_config_pgs, etcd_request, pool_id, up_osds, osd_tree, prev_pgs, [], []);
                 }
             }
             for (const pool_id in this.state.config.pools)
@@ -1226,7 +1258,7 @@ class Mon
                             return;
                         }
                         const new_pg_history = [];
-                        PGUtil.scale_pg_count(prev_pgs, pg_history, new_pg_history, pool_cfg.pg_count);
+                        PGUtil.scale_pg_count(prev_pgs, real_prev_pgs, pg_history, new_pg_history, pool_cfg.pg_count);
                         pg_history = new_pg_history;
                     }
                     for (const pg of prev_pgs)
@@ -1279,14 +1311,15 @@ 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, osd_tree, real_prev_pgs, optimize_result.int_pgs, pg_history);
+                this.save_new_pgs_txn(new_config_pgs, 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);
+            new_config_pgs.hash = tree_hash;
+            await this.save_pg_config(new_config_pgs, etcd_request);
         }
         else
         {
             // Nothing changed, but we still want to recheck the distribution of primaries
+            let new_config_pgs;
             let changed = false;
             for (const pool_id in this.state.config.pools)
             {
@@ -1306,31 +1339,35 @@ class Mon
                         const new_primary = this.pick_primary(pool_id, pg_cfg.osd_set, up_osds, aff_osds);
                         if (pg_cfg.primary != new_primary)
                         {
+                            if (!new_config_pgs)
+                            {
+                                new_config_pgs = JSON.parse(JSON.stringify(this.state.config.pgs));
+                            }
                             console.log(
                                 `Moving pool ${pool_id} (${pool_cfg.name || 'unnamed'}) PG ${pg_num}`+
                                 ` primary OSD from ${pg_cfg.primary} to ${new_primary}`
                             );
                             changed = true;
-                            pg_cfg.primary = new_primary;
+                            new_config_pgs.items[pool_id][pg_num].primary = new_primary;
                         }
                     }
                 }
             }
             if (changed)
             {
-                await this.save_pg_config();
+                await this.save_pg_config(new_config_pgs);
             }
         }
     }
 
-    async save_pg_config(etcd_request = { compare: [], success: [] })
+    async save_pg_config(new_config_pgs, etcd_request = { compare: [], success: [] })
     {
         etcd_request.compare.push(
             { key: b64(this.etcd_prefix+'/mon/master'), target: 'LEASE', lease: ''+this.etcd_lease_id },
             { key: b64(this.etcd_prefix+'/config/pgs'), target: 'MOD', mod_revision: ''+this.etcd_watch_revision, result: 'LESS' },
         );
         etcd_request.success.push(
-            { requestPut: { key: b64(this.etcd_prefix+'/config/pgs'), value: b64(JSON.stringify(this.state.config.pgs)) } },
+            { requestPut: { key: b64(this.etcd_prefix+'/config/pgs'), value: b64(JSON.stringify(new_config_pgs)) } },
         );
         const res = await this.etcd_call('/kv/txn', etcd_request, this.config.etcd_mon_timeout, 0);
         if (!res.succeeded)
@@ -1374,16 +1411,14 @@ class Mon
     // This is required for multiple change events to trigger at most 1 recheck in 1s
     schedule_recheck()
     {
-        if (this.recheck_timer)
+        if (!this.recheck_timer)
         {
-            clearTimeout(this.recheck_timer);
-            this.recheck_timer = null;
+            this.recheck_timer = setTimeout(() =>
+            {
+                this.recheck_timer = null;
+                this.recheck_pgs().catch(this.die);
+            }, this.config.mon_change_timeout || 1000);
         }
-        this.recheck_timer = setTimeout(() =>
-        {
-            this.recheck_timer = null;
-            this.recheck_pgs().catch(this.die);
-        }, this.config.mon_change_timeout || 1000);
     }
 
     sum_op_stats(timestamp, prev_stats)
@@ -1719,11 +1754,11 @@ class Mon
         else if (key_parts[0] === 'osd' && key_parts[1] === 'stats')
         {
             // Recheck OSD tree on OSD addition/deletion
-            if ((!old) != (!kv.value) || old && kv.value && (old.size != kv.value.size || old.time != kv.value.time))
+            if ((!old) != (!kv.value) || old && kv.value && old.size != kv.value.size)
             {
                 this.schedule_recheck();
             }
-            // Recheck PGs <osd_out_time> later
+            // Recheck PGs <osd_out_time> after last OSD statistics report
             this.schedule_next_recheck_at(
                 !this.state.osd.stats[key[2]] ? 0 : this.state.osd.stats[key[2]].time+this.config.osd_out_time
             );
@@ -1759,14 +1794,13 @@ class Mon
                 return res.json;
             }
         }
-        this.die();
+        await this.die();
     }
 
-    _die(err)
+    async _die(err)
     {
-        // In fact we can just try to rejoin
         console.error(new Error(err || 'Cluster connection failed'));
-        process.exit(1);
+        await this.on_stop(1);
     }
 
     local_ips(all)
@@ -1811,6 +1845,7 @@ function POST(url, body, timeout)
             clearTimeout(timer_id);
             let res_body = '';
             res.setEncoding('utf8');
+            res.on('error', no);
             res.on('data', chunk => { res_body += chunk; });
             res.on('end', () =>
             {
@@ -1830,6 +1865,8 @@ function POST(url, body, timeout)
                 }
             });
         });
+        req.on('error', no);
+        req.on('close', () => no(new Error('Connection closed prematurely')));
         req.write(body_text);
         req.end();
     });

mon/vitastor-mon.service

@@ -15,4 +15,4 @@ StartLimitInterval=0
 RestartSec=10
 
 [Install]
-WantedBy=vitastor.target
+WantedBy=multi-user.target

patches/VitastorPlugin.pm

@@ -16,6 +16,11 @@ use PVE::Tools qw(run_command);
 
 use base qw(PVE::Storage::Plugin);
 
+if (@PVE::Storage::Plugin::SHARED_STORAGE)
+{
+    push @PVE::Storage::Plugin::SHARED_STORAGE, 'vitastor';
+}
+
 sub api
 {
     # Trick it :)
@@ -133,9 +138,11 @@ sub properties
 sub options
 {
     return {
+        shared => { optional => 1 },
+        content => { optional => 1 },
         nodes => { optional => 1 },
         disable => { optional => 1 },
-        vitastor_etcd_address => { optional => 1},
+        vitastor_etcd_address => { optional => 1 },
         vitastor_etcd_prefix => { optional => 1 },
         vitastor_config_path => { optional => 1 },
         vitastor_prefix => { optional => 1 },

patches/cinder-vitastor.py

@@ -50,7 +50,7 @@ from cinder.volume import configuration
 from cinder.volume import driver
 from cinder.volume import volume_utils
 
-VERSION = '0.8.3'
+VERSION = '0.8.8'
 
 LOG = logging.getLogger(__name__)
 

patches/pve-qemu-7.2-vitastor.patch

@@ -0,0 +1,169 @@
+Index: pve-qemu-kvm-7.2.0/block/meson.build
+===================================================================
+--- pve-qemu-kvm-7.2.0.orig/block/meson.build
++++ pve-qemu-kvm-7.2.0/block/meson.build
+@@ -113,6 +113,7 @@ foreach m : [
+   [libnfs, 'nfs', files('nfs.c')],
+   [libssh, 'ssh', files('ssh.c')],
+   [rbd, 'rbd', files('rbd.c')],
++  [vitastor, 'vitastor', files('vitastor.c')],
+ ]
+   if m[0].found()
+     module_ss = ss.source_set()
+Index: pve-qemu-kvm-7.2.0/meson.build
+===================================================================
+--- pve-qemu-kvm-7.2.0.orig/meson.build
++++ pve-qemu-kvm-7.2.0/meson.build
+@@ -1026,6 +1026,26 @@ if not get_option('rbd').auto() or have_
+   endif
+ endif
+ 
++vitastor = not_found
++if not get_option('vitastor').auto() or have_block
++  libvitastor_client = cc.find_library('vitastor_client', has_headers: ['vitastor_c.h'],
++    required: get_option('vitastor'), kwargs: static_kwargs)
++  if libvitastor_client.found()
++    if cc.links('''
++      #include <vitastor_c.h>
++      int main(void) {
++        vitastor_c_create_qemu(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0);
++        return 0;
++      }''', dependencies: libvitastor_client)
++      vitastor = declare_dependency(dependencies: libvitastor_client)
++    elif get_option('vitastor').enabled()
++      error('could not link libvitastor_client')
++    else
++      warning('could not link libvitastor_client, disabling')
++    endif
++  endif
++endif
++
+ glusterfs = not_found
+ glusterfs_ftruncate_has_stat = false
+ glusterfs_iocb_has_stat = false
+@@ -1865,6 +1885,7 @@ config_host_data.set('CONFIG_NUMA', numa
+ config_host_data.set('CONFIG_OPENGL', opengl.found())
+ config_host_data.set('CONFIG_PROFILER', get_option('profiler'))
+ config_host_data.set('CONFIG_RBD', rbd.found())
++config_host_data.set('CONFIG_VITASTOR', vitastor.found())
+ config_host_data.set('CONFIG_RDMA', rdma.found())
+ config_host_data.set('CONFIG_SDL', sdl.found())
+ config_host_data.set('CONFIG_SDL_IMAGE', sdl_image.found())
+@@ -3957,6 +3978,7 @@ if spice_protocol.found()
+   summary_info += {'  spice server support': spice}
+ endif
+ summary_info += {'rbd support':       rbd}
++summary_info += {'vitastor support':  vitastor}
+ summary_info += {'smartcard support': cacard}
+ summary_info += {'U2F support':       u2f}
+ summary_info += {'libusb':            libusb}
+Index: pve-qemu-kvm-7.2.0/meson_options.txt
+===================================================================
+--- pve-qemu-kvm-7.2.0.orig/meson_options.txt
++++ pve-qemu-kvm-7.2.0/meson_options.txt
+@@ -169,6 +169,8 @@ option('lzo', type : 'feature', value :
+        description: 'lzo compression support')
+ option('rbd', type : 'feature', value : 'auto',
+        description: 'Ceph block device driver')
++option('vitastor', type : 'feature', value : 'auto',
++       description: 'Vitastor block device driver')
+ option('opengl', type : 'feature', value : 'auto',
+        description: 'OpenGL support')
+ option('rdma', type : 'feature', value : 'auto',
+Index: pve-qemu-kvm-7.2.0/qapi/block-core.json
+===================================================================
+--- pve-qemu-kvm-7.2.0.orig/qapi/block-core.json
++++ pve-qemu-kvm-7.2.0/qapi/block-core.json
+@@ -3213,7 +3213,7 @@
+             'raw', 'rbd',
+             { 'name': 'replication', 'if': 'CONFIG_REPLICATION' },
+             'pbs',
+-            'ssh', 'throttle', 'vdi', 'vhdx',
++            'ssh', 'throttle', 'vdi', 'vhdx', 'vitastor',
+             { 'name': 'virtio-blk-vfio-pci', 'if': 'CONFIG_BLKIO' },
+             { 'name': 'virtio-blk-vhost-user', 'if': 'CONFIG_BLKIO' },
+             { 'name': 'virtio-blk-vhost-vdpa', 'if': 'CONFIG_BLKIO' },
+@@ -4223,6 +4223,28 @@
+             '*server': ['InetSocketAddressBase'] } }
+ 
+ ##
++# @BlockdevOptionsVitastor:
++#
++# Driver specific block device options for vitastor
++#
++# @image:       Image name
++# @inode:       Inode number
++# @pool:        Pool ID
++# @size:        Desired image size in bytes
++# @config-path: Path to Vitastor configuration
++# @etcd-host:   etcd connection address(es)
++# @etcd-prefix: etcd key/value prefix
++##
++{ 'struct': 'BlockdevOptionsVitastor',
++  'data': { '*inode': 'uint64',
++            '*pool': 'uint64',
++            '*size': 'uint64',
++            '*image': 'str',
++            '*config-path': 'str',
++            '*etcd-host': 'str',
++            '*etcd-prefix': 'str' } }
++
++##
+ # @ReplicationMode:
+ #
+ # An enumeration of replication modes.
+@@ -4671,6 +4693,7 @@
+       'throttle':   'BlockdevOptionsThrottle',
+       'vdi':        'BlockdevOptionsGenericFormat',
+       'vhdx':       'BlockdevOptionsGenericFormat',
++      'vitastor':   'BlockdevOptionsVitastor',
+       'virtio-blk-vfio-pci':
+                     { 'type': 'BlockdevOptionsVirtioBlkVfioPci',
+                       'if': 'CONFIG_BLKIO' },
+@@ -5072,6 +5095,17 @@
+             '*encrypt' :        'RbdEncryptionCreateOptions' } }
+ 
+ ##
++# @BlockdevCreateOptionsVitastor:
++#
++# Driver specific image creation options for Vitastor.
++#
++# @size: Size of the virtual disk in bytes
++##
++{ 'struct': 'BlockdevCreateOptionsVitastor',
++  'data': { 'location':         'BlockdevOptionsVitastor',
++            'size':             'size' } }
++
++##
+ # @BlockdevVmdkSubformat:
+ #
+ # Subformat options for VMDK images
+@@ -5269,6 +5303,7 @@
+       'ssh':            'BlockdevCreateOptionsSsh',
+       'vdi':            'BlockdevCreateOptionsVdi',
+       'vhdx':           'BlockdevCreateOptionsVhdx',
++      'vitastor':       'BlockdevCreateOptionsVitastor',
+       'vmdk':           'BlockdevCreateOptionsVmdk',
+       'vpc':            'BlockdevCreateOptionsVpc'
+   } }
+Index: pve-qemu-kvm-7.2.0/scripts/ci/org.centos/stream/8/x86_64/configure
+===================================================================
+--- pve-qemu-kvm-7.2.0.orig/scripts/ci/org.centos/stream/8/x86_64/configure
++++ pve-qemu-kvm-7.2.0/scripts/ci/org.centos/stream/8/x86_64/configure
+@@ -31,7 +31,7 @@
+ --with-git=meson \
+ --with-git-submodules=update \
+ --target-list="x86_64-softmmu" \
+---block-drv-rw-whitelist="qcow2,raw,file,host_device,nbd,iscsi,rbd,blkdebug,luks,null-co,nvme,copy-on-read,throttle,gluster" \
++--block-drv-rw-whitelist="qcow2,raw,file,host_device,nbd,iscsi,rbd,vitastor,blkdebug,luks,null-co,nvme,copy-on-read,throttle,gluster" \
+ --audio-drv-list="" \
+ --block-drv-ro-whitelist="vmdk,vhdx,vpc,https,ssh" \
+ --with-coroutine=ucontext \
+@@ -179,6 +179,7 @@
+ --enable-opengl \
+ --enable-pie \
+ --enable-rbd \
++--enable-vitastor \
+ --enable-rdma \
+ --enable-seccomp \
+ --enable-snappy \

patches/qemu-6.2-vitastor.patch

@@ -0,0 +1,169 @@
+diff --git a/block/meson.build b/block/meson.build
+index deb73ca389..e269f599a1 100644
+--- a/block/meson.build
++++ b/block/meson.build
+@@ -78,6 +78,7 @@ foreach m : [
+   [libnfs, 'nfs', files('nfs.c')],
+   [libssh, 'ssh', files('ssh.c')],
+   [rbd, 'rbd', files('rbd.c')],
++  [vitastor, 'vitastor', files('vitastor.c')],
+ ]
+   if m[0].found()
+     module_ss = ss.source_set()
+diff --git a/meson.build b/meson.build
+index 96de1a6ef9..2e3994777d 100644
+--- a/meson.build
++++ b/meson.build
+@@ -838,6 +838,26 @@ if not get_option('rbd').auto() or have_block
+   endif
+ endif
+ 
++vitastor = not_found
++if not get_option('vitastor').auto() or have_block
++  libvitastor_client = cc.find_library('vitastor_client', has_headers: ['vitastor_c.h'],
++    required: get_option('vitastor'), kwargs: static_kwargs)
++  if libvitastor_client.found()
++    if cc.links('''
++      #include <vitastor_c.h>
++      int main(void) {
++        vitastor_c_create_qemu(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0);
++        return 0;
++      }''', dependencies: libvitastor_client)
++      vitastor = declare_dependency(dependencies: libvitastor_client)
++    elif get_option('vitastor').enabled()
++      error('could not link libvitastor_client')
++    else
++      warning('could not link libvitastor_client, disabling')
++    endif
++  endif
++endif
++
+ glusterfs = not_found
+ glusterfs_ftruncate_has_stat = false
+ glusterfs_iocb_has_stat = false
+@@ -1455,6 +1475,7 @@ config_host_data.set('CONFIG_LINUX_AIO', libaio.found())
+ config_host_data.set('CONFIG_LINUX_IO_URING', linux_io_uring.found())
+ config_host_data.set('CONFIG_LIBPMEM', libpmem.found())
+ config_host_data.set('CONFIG_RBD', rbd.found())
++config_host_data.set('CONFIG_VITASTOR', vitastor.found())
+ config_host_data.set('CONFIG_SDL', sdl.found())
+ config_host_data.set('CONFIG_SDL_IMAGE', sdl_image.found())
+ config_host_data.set('CONFIG_SECCOMP', seccomp.found())
+@@ -3412,6 +3433,7 @@ if spice_protocol.found()
+   summary_info += {'  spice server support': spice}
+ endif
+ summary_info += {'rbd support':       rbd}
++summary_info += {'vitastor support':  vitastor}
+ summary_info += {'xfsctl support':    config_host.has_key('CONFIG_XFS')}
+ summary_info += {'smartcard support': cacard}
+ summary_info += {'U2F support':       u2f}
+diff --git a/meson_options.txt b/meson_options.txt
+index e392323732..5b56007475 100644
+--- a/meson_options.txt
++++ b/meson_options.txt
+@@ -121,6 +121,8 @@ option('lzo', type : 'feature', value : 'auto',
+        description: 'lzo compression support')
+ option('rbd', type : 'feature', value : 'auto',
+        description: 'Ceph block device driver')
++option('vitastor', type : 'feature', value : 'auto',
++       description: 'Vitastor block device driver')
+ option('gtk', type : 'feature', value : 'auto',
+        description: 'GTK+ user interface')
+ option('sdl', type : 'feature', value : 'auto',
+diff --git a/qapi/block-core.json b/qapi/block-core.json
+index 1d3dd9cb48..88453405e5 100644
+--- a/qapi/block-core.json
++++ b/qapi/block-core.json
+@@ -2930,7 +2930,7 @@
+             'luks', 'nbd', 'nfs', 'null-aio', 'null-co', 'nvme', 'parallels',
+             'preallocate', 'qcow', 'qcow2', 'qed', 'quorum', 'raw', 'rbd',
+             { 'name': 'replication', 'if': 'CONFIG_REPLICATION' },
+-            'ssh', 'throttle', 'vdi', 'vhdx', 'vmdk', 'vpc', 'vvfat' ] }
++            'ssh', 'throttle', 'vdi', 'vhdx', 'vitastor', 'vmdk', 'vpc', 'vvfat' ] }
+ 
+ ##
+ # @BlockdevOptionsFile:
+@@ -3864,6 +3864,28 @@
+             '*key-secret': 'str',
+             '*server': ['InetSocketAddressBase'] } }
+ 
++##
++# @BlockdevOptionsVitastor:
++#
++# Driver specific block device options for vitastor
++#
++# @image:       Image name
++# @inode:       Inode number
++# @pool:        Pool ID
++# @size:        Desired image size in bytes
++# @config-path: Path to Vitastor configuration
++# @etcd-host:   etcd connection address(es)
++# @etcd-prefix: etcd key/value prefix
++##
++{ 'struct': 'BlockdevOptionsVitastor',
++  'data': { '*inode': 'uint64',
++            '*pool': 'uint64',
++            '*size': 'uint64',
++            '*image': 'str',
++            '*config-path': 'str',
++            '*etcd-host': 'str',
++            '*etcd-prefix': 'str' } }
++
+ ##
+ # @ReplicationMode:
+ #
+@@ -4259,6 +4281,7 @@
+       'throttle':   'BlockdevOptionsThrottle',
+       'vdi':        'BlockdevOptionsGenericFormat',
+       'vhdx':       'BlockdevOptionsGenericFormat',
++      'vitastor':   'BlockdevOptionsVitastor',
+       'vmdk':       'BlockdevOptionsGenericCOWFormat',
+       'vpc':        'BlockdevOptionsGenericFormat',
+       'vvfat':      'BlockdevOptionsVVFAT'
+@@ -4647,6 +4670,17 @@
+             '*cluster-size' :   'size',
+             '*encrypt' :        'RbdEncryptionCreateOptions' } }
+ 
++##
++# @BlockdevCreateOptionsVitastor:
++#
++# Driver specific image creation options for Vitastor.
++#
++# @size: Size of the virtual disk in bytes
++##
++{ 'struct': 'BlockdevCreateOptionsVitastor',
++  'data': { 'location':         'BlockdevOptionsVitastor',
++            'size':             'size' } }
++
+ ##
+ # @BlockdevVmdkSubformat:
+ #
+@@ -4846,6 +4880,7 @@
+       'ssh':            'BlockdevCreateOptionsSsh',
+       'vdi':            'BlockdevCreateOptionsVdi',
+       'vhdx':           'BlockdevCreateOptionsVhdx',
++      'vitastor':       'BlockdevCreateOptionsVitastor',
+       'vmdk':           'BlockdevCreateOptionsVmdk',
+       'vpc':            'BlockdevCreateOptionsVpc'
+   } }
+diff --git a/scripts/meson-buildoptions.sh b/scripts/meson-buildoptions.sh
+index 7a17ff4218..cdddbf32aa 100644
+--- a/scripts/meson-buildoptions.sh
++++ b/scripts/meson-buildoptions.sh
+@@ -69,6 +69,7 @@ meson_options_help() {
+   printf "%s\n" '  oss             OSS sound support'
+   printf "%s\n" '  pa              PulseAudio sound support'
+   printf "%s\n" '  rbd             Ceph block device driver'
++  printf "%s\n" '  vitastor        Vitastor block device driver'
+   printf "%s\n" '  sdl             SDL user interface'
+   printf "%s\n" '  sdl-image       SDL Image support for icons'
+   printf "%s\n" '  seccomp         seccomp support'
+@@ -210,6 +211,8 @@ _meson_option_parse() {
+     --disable-pa) printf "%s" -Dpa=disabled ;;
+     --enable-rbd) printf "%s" -Drbd=enabled ;;
+     --disable-rbd) printf "%s" -Drbd=disabled ;;
++    --enable-vitastor) printf "%s" -Dvitastor=enabled ;;
++    --disable-vitastor) printf "%s" -Dvitastor=disabled ;;
+     --enable-sdl) printf "%s" -Dsdl=enabled ;;
+     --disable-sdl) printf "%s" -Dsdl=disabled ;;
+     --enable-sdl-image) printf "%s" -Dsdl_image=enabled ;;

patches/qemu-7.0-vitastor.patch

@@ -0,0 +1,190 @@
+diff --git a/block/meson.build b/block/meson.build
+index 0b2a60c99b..d923713804 100644
+--- a/block/meson.build
++++ b/block/meson.build
+@@ -98,6 +98,7 @@ foreach m : [
+   [libnfs, 'nfs', files('nfs.c')],
+   [libssh, 'ssh', files('ssh.c')],
+   [rbd, 'rbd', files('rbd.c')],
++  [vitastor, 'vitastor', files('vitastor.c')],
+ ]
+   if m[0].found()
+     module_ss = ss.source_set()
+diff --git a/meson.build b/meson.build
+index 861de93c4f..272f72af11 100644
+--- a/meson.build
++++ b/meson.build
+@@ -884,6 +884,26 @@ if not get_option('rbd').auto() or have_block
+   endif
+ endif
+ 
++vitastor = not_found
++if not get_option('vitastor').auto() or have_block
++  libvitastor_client = cc.find_library('vitastor_client', has_headers: ['vitastor_c.h'],
++    required: get_option('vitastor'), kwargs: static_kwargs)
++  if libvitastor_client.found()
++    if cc.links('''
++      #include <vitastor_c.h>
++      int main(void) {
++        vitastor_c_create_qemu(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0);
++        return 0;
++      }''', dependencies: libvitastor_client)
++      vitastor = declare_dependency(dependencies: libvitastor_client)
++    elif get_option('vitastor').enabled()
++      error('could not link libvitastor_client')
++    else
++      warning('could not link libvitastor_client, disabling')
++    endif
++  endif
++endif
++
+ glusterfs = not_found
+ glusterfs_ftruncate_has_stat = false
+ glusterfs_iocb_has_stat = false
+@@ -1546,6 +1566,7 @@ config_host_data.set('CONFIG_LIBPMEM', libpmem.found())
+ config_host_data.set('CONFIG_NUMA', numa.found())
+ config_host_data.set('CONFIG_PROFILER', get_option('profiler'))
+ config_host_data.set('CONFIG_RBD', rbd.found())
++config_host_data.set('CONFIG_VITASTOR', vitastor.found())
+ config_host_data.set('CONFIG_SDL', sdl.found())
+ config_host_data.set('CONFIG_SDL_IMAGE', sdl_image.found())
+ config_host_data.set('CONFIG_SECCOMP', seccomp.found())
+@@ -3709,6 +3730,7 @@ if spice_protocol.found()
+   summary_info += {'  spice server support': spice}
+ endif
+ summary_info += {'rbd support':       rbd}
++summary_info += {'vitastor support':  vitastor}
+ summary_info += {'smartcard support': cacard}
+ summary_info += {'U2F support':       u2f}
+ summary_info += {'libusb':            libusb}
+diff --git a/meson_options.txt b/meson_options.txt
+index 52b11cead4..d8d0868174 100644
+--- a/meson_options.txt
++++ b/meson_options.txt
+@@ -149,6 +149,8 @@ option('lzo', type : 'feature', value : 'auto',
+        description: 'lzo compression support')
+ option('rbd', type : 'feature', value : 'auto',
+        description: 'Ceph block device driver')
++option('vitastor', type : 'feature', value : 'auto',
++       description: 'Vitastor block device driver')
+ option('gtk', type : 'feature', value : 'auto',
+        description: 'GTK+ user interface')
+ option('sdl', type : 'feature', value : 'auto',
+diff --git a/qapi/block-core.json b/qapi/block-core.json
+index beeb91952a..1c98dc0e12 100644
+--- a/qapi/block-core.json
++++ b/qapi/block-core.json
+@@ -2929,7 +2929,7 @@
+             'luks', 'nbd', 'nfs', 'null-aio', 'null-co', 'nvme', 'parallels',
+             'preallocate', 'qcow', 'qcow2', 'qed', 'quorum', 'raw', 'rbd',
+             { 'name': 'replication', 'if': 'CONFIG_REPLICATION' },
+-            'ssh', 'throttle', 'vdi', 'vhdx', 'vmdk', 'vpc', 'vvfat' ] }
++            'ssh', 'throttle', 'vdi', 'vhdx', 'vitastor', 'vmdk', 'vpc', 'vvfat' ] }
+ 
+ ##
+ # @BlockdevOptionsFile:
+@@ -3863,6 +3863,28 @@
+             '*key-secret': 'str',
+             '*server': ['InetSocketAddressBase'] } }
+ 
++##
++# @BlockdevOptionsVitastor:
++#
++# Driver specific block device options for vitastor
++#
++# @image:       Image name
++# @inode:       Inode number
++# @pool:        Pool ID
++# @size:        Desired image size in bytes
++# @config-path: Path to Vitastor configuration
++# @etcd-host:   etcd connection address(es)
++# @etcd-prefix: etcd key/value prefix
++##
++{ 'struct': 'BlockdevOptionsVitastor',
++  'data': { '*inode': 'uint64',
++            '*pool': 'uint64',
++            '*size': 'uint64',
++            '*image': 'str',
++            '*config-path': 'str',
++            '*etcd-host': 'str',
++            '*etcd-prefix': 'str' } }
++
+ ##
+ # @ReplicationMode:
+ #
+@@ -4277,6 +4299,7 @@
+       'throttle':   'BlockdevOptionsThrottle',
+       'vdi':        'BlockdevOptionsGenericFormat',
+       'vhdx':       'BlockdevOptionsGenericFormat',
++      'vitastor':   'BlockdevOptionsVitastor',
+       'vmdk':       'BlockdevOptionsGenericCOWFormat',
+       'vpc':        'BlockdevOptionsGenericFormat',
+       'vvfat':      'BlockdevOptionsVVFAT'
+@@ -4665,6 +4688,17 @@
+             '*cluster-size' :   'size',
+             '*encrypt' :        'RbdEncryptionCreateOptions' } }
+ 
++##
++# @BlockdevCreateOptionsVitastor:
++#
++# Driver specific image creation options for Vitastor.
++#
++# @size: Size of the virtual disk in bytes
++##
++{ 'struct': 'BlockdevCreateOptionsVitastor',
++  'data': { 'location':         'BlockdevOptionsVitastor',
++            'size':             'size' } }
++
+ ##
+ # @BlockdevVmdkSubformat:
+ #
+@@ -4864,6 +4898,7 @@
+       'ssh':            'BlockdevCreateOptionsSsh',
+       'vdi':            'BlockdevCreateOptionsVdi',
+       'vhdx':           'BlockdevCreateOptionsVhdx',
++      'vitastor':       'BlockdevCreateOptionsVitastor',
+       'vmdk':           'BlockdevCreateOptionsVmdk',
+       'vpc':            'BlockdevCreateOptionsVpc'
+   } }
+diff --git a/scripts/ci/org.centos/stream/8/x86_64/configure b/scripts/ci/org.centos/stream/8/x86_64/configure
+index 9850dd4444..72b1287520 100755
+--- a/scripts/ci/org.centos/stream/8/x86_64/configure
++++ b/scripts/ci/org.centos/stream/8/x86_64/configure
+@@ -31,7 +31,7 @@
+ --with-git=meson \
+ --with-git-submodules=update \
+ --target-list="x86_64-softmmu" \
+---block-drv-rw-whitelist="qcow2,raw,file,host_device,nbd,iscsi,rbd,blkdebug,luks,null-co,nvme,copy-on-read,throttle,gluster" \
++--block-drv-rw-whitelist="qcow2,raw,file,host_device,nbd,iscsi,rbd,vitastor,blkdebug,luks,null-co,nvme,copy-on-read,throttle,gluster" \
+ --audio-drv-list="" \
+ --block-drv-ro-whitelist="vmdk,vhdx,vpc,https,ssh" \
+ --with-coroutine=ucontext \
+@@ -181,6 +181,7 @@
+ --enable-opengl \
+ --enable-pie \
+ --enable-rbd \
++--enable-vitastor \
+ --enable-rdma \
+ --enable-seccomp \
+ --enable-snappy \
+diff --git a/scripts/meson-buildoptions.sh b/scripts/meson-buildoptions.sh
+index 1e26f4571e..370898d48c 100644
+--- a/scripts/meson-buildoptions.sh
++++ b/scripts/meson-buildoptions.sh
+@@ -98,6 +98,7 @@ meson_options_help() {
+   printf "%s\n" '  qed             qed image format support'
+   printf "%s\n" '  qga-vss         build QGA VSS support (broken with MinGW)'
+   printf "%s\n" '  rbd             Ceph block device driver'
++  printf "%s\n" '  vitastor        Vitastor block device driver'
+   printf "%s\n" '  replication     replication support'
+   printf "%s\n" '  sdl             SDL user interface'
+   printf "%s\n" '  sdl-image       SDL Image support for icons'
+@@ -289,6 +290,8 @@ _meson_option_parse() {
+     --disable-qom-cast-debug) printf "%s" -Dqom_cast_debug=false ;;
+     --enable-rbd) printf "%s" -Drbd=enabled ;;
+     --disable-rbd) printf "%s" -Drbd=disabled ;;
++    --enable-vitastor) printf "%s" -Dvitastor=enabled ;;
++    --disable-vitastor) printf "%s" -Dvitastor=disabled ;;
+     --enable-replication) printf "%s" -Dreplication=enabled ;;
+     --disable-replication) printf "%s" -Dreplication=disabled ;;
+     --enable-rng-none) printf "%s" -Drng_none=true ;;

patches/qemu-7.1-vitastor.patch

@@ -0,0 +1,190 @@
+diff --git a/block/meson.build b/block/meson.build
+index 60bc305597..89a042216f 100644
+--- a/block/meson.build
++++ b/block/meson.build
+@@ -98,6 +98,7 @@ foreach m : [
+   [libnfs, 'nfs', files('nfs.c')],
+   [libssh, 'ssh', files('ssh.c')],
+   [rbd, 'rbd', files('rbd.c')],
++  [vitastor, 'vitastor', files('vitastor.c')],
+ ]
+   if m[0].found()
+     module_ss = ss.source_set()
+diff --git a/meson.build b/meson.build
+index 20fddbd707..600db4e2fb 100644
+--- a/meson.build
++++ b/meson.build
+@@ -967,6 +967,26 @@ if not get_option('rbd').auto() or have_block
+   endif
+ endif
+ 
++vitastor = not_found
++if not get_option('vitastor').auto() or have_block
++  libvitastor_client = cc.find_library('vitastor_client', has_headers: ['vitastor_c.h'],
++    required: get_option('vitastor'), kwargs: static_kwargs)
++  if libvitastor_client.found()
++    if cc.links('''
++      #include <vitastor_c.h>
++      int main(void) {
++        vitastor_c_create_qemu(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0);
++        return 0;
++      }''', dependencies: libvitastor_client)
++      vitastor = declare_dependency(dependencies: libvitastor_client)
++    elif get_option('vitastor').enabled()
++      error('could not link libvitastor_client')
++    else
++      warning('could not link libvitastor_client, disabling')
++    endif
++  endif
++endif
++
+ glusterfs = not_found
+ glusterfs_ftruncate_has_stat = false
+ glusterfs_iocb_has_stat = false
+@@ -1799,6 +1819,7 @@ config_host_data.set('CONFIG_NUMA', numa.found())
+ config_host_data.set('CONFIG_OPENGL', opengl.found())
+ config_host_data.set('CONFIG_PROFILER', get_option('profiler'))
+ config_host_data.set('CONFIG_RBD', rbd.found())
++config_host_data.set('CONFIG_VITASTOR', vitastor.found())
+ config_host_data.set('CONFIG_RDMA', rdma.found())
+ config_host_data.set('CONFIG_SDL', sdl.found())
+ config_host_data.set('CONFIG_SDL_IMAGE', sdl_image.found())
+@@ -3954,6 +3975,7 @@ if spice_protocol.found()
+   summary_info += {'  spice server support': spice}
+ endif
+ summary_info += {'rbd support':       rbd}
++summary_info += {'vitastor support':  vitastor}
+ summary_info += {'smartcard support': cacard}
+ summary_info += {'U2F support':       u2f}
+ summary_info += {'libusb':            libusb}
+diff --git a/meson_options.txt b/meson_options.txt
+index e58e158396..9747b38fd0 100644
+--- a/meson_options.txt
++++ b/meson_options.txt
+@@ -167,6 +167,8 @@ option('lzo', type : 'feature', value : 'auto',
+        description: 'lzo compression support')
+ option('rbd', type : 'feature', value : 'auto',
+        description: 'Ceph block device driver')
++option('vitastor', type : 'feature', value : 'auto',
++       description: 'Vitastor block device driver')
+ option('opengl', type : 'feature', value : 'auto',
+        description: 'OpenGL support')
+ option('rdma', type : 'feature', value : 'auto',
+diff --git a/qapi/block-core.json b/qapi/block-core.json
+index 2173e7734a..5a4900b322 100644
+--- a/qapi/block-core.json
++++ b/qapi/block-core.json
+@@ -2955,7 +2955,7 @@
+             'luks', 'nbd', 'nfs', 'null-aio', 'null-co', 'nvme', 'parallels',
+             'preallocate', 'qcow', 'qcow2', 'qed', 'quorum', 'raw', 'rbd',
+             { 'name': 'replication', 'if': 'CONFIG_REPLICATION' },
+-            'ssh', 'throttle', 'vdi', 'vhdx', 'vmdk', 'vpc', 'vvfat' ] }
++            'ssh', 'throttle', 'vdi', 'vhdx', 'vitastor', 'vmdk', 'vpc', 'vvfat' ] }
+ 
+ ##
+ # @BlockdevOptionsFile:
+@@ -3883,6 +3883,28 @@
+             '*key-secret': 'str',
+             '*server': ['InetSocketAddressBase'] } }
+ 
++##
++# @BlockdevOptionsVitastor:
++#
++# Driver specific block device options for vitastor
++#
++# @image:       Image name
++# @inode:       Inode number
++# @pool:        Pool ID
++# @size:        Desired image size in bytes
++# @config-path: Path to Vitastor configuration
++# @etcd-host:   etcd connection address(es)
++# @etcd-prefix: etcd key/value prefix
++##
++{ 'struct': 'BlockdevOptionsVitastor',
++  'data': { '*inode': 'uint64',
++            '*pool': 'uint64',
++            '*size': 'uint64',
++            '*image': 'str',
++            '*config-path': 'str',
++            '*etcd-host': 'str',
++            '*etcd-prefix': 'str' } }
++
+ ##
+ # @ReplicationMode:
+ #
+@@ -4327,6 +4349,7 @@
+       'throttle':   'BlockdevOptionsThrottle',
+       'vdi':        'BlockdevOptionsGenericFormat',
+       'vhdx':       'BlockdevOptionsGenericFormat',
++      'vitastor':   'BlockdevOptionsVitastor',
+       'vmdk':       'BlockdevOptionsGenericCOWFormat',
+       'vpc':        'BlockdevOptionsGenericFormat',
+       'vvfat':      'BlockdevOptionsVVFAT'
+@@ -4717,6 +4740,17 @@
+             '*cluster-size' :   'size',
+             '*encrypt' :        'RbdEncryptionCreateOptions' } }
+ 
++##
++# @BlockdevCreateOptionsVitastor:
++#
++# Driver specific image creation options for Vitastor.
++#
++# @size: Size of the virtual disk in bytes
++##
++{ 'struct': 'BlockdevCreateOptionsVitastor',
++  'data': { 'location':         'BlockdevOptionsVitastor',
++            'size':             'size' } }
++
+ ##
+ # @BlockdevVmdkSubformat:
+ #
+@@ -4915,6 +4949,7 @@
+       'ssh':            'BlockdevCreateOptionsSsh',
+       'vdi':            'BlockdevCreateOptionsVdi',
+       'vhdx':           'BlockdevCreateOptionsVhdx',
++      'vitastor':       'BlockdevCreateOptionsVitastor',
+       'vmdk':           'BlockdevCreateOptionsVmdk',
+       'vpc':            'BlockdevCreateOptionsVpc'
+   } }
+diff --git a/scripts/ci/org.centos/stream/8/x86_64/configure b/scripts/ci/org.centos/stream/8/x86_64/configure
+index a7f92aff90..53dc55be2e 100755
+--- a/scripts/ci/org.centos/stream/8/x86_64/configure
++++ b/scripts/ci/org.centos/stream/8/x86_64/configure
+@@ -31,7 +31,7 @@
+ --with-git=meson \
+ --with-git-submodules=update \
+ --target-list="x86_64-softmmu" \
+---block-drv-rw-whitelist="qcow2,raw,file,host_device,nbd,iscsi,rbd,blkdebug,luks,null-co,nvme,copy-on-read,throttle,gluster" \
++--block-drv-rw-whitelist="qcow2,raw,file,host_device,nbd,iscsi,rbd,vitastor,blkdebug,luks,null-co,nvme,copy-on-read,throttle,gluster" \
+ --audio-drv-list="" \
+ --block-drv-ro-whitelist="vmdk,vhdx,vpc,https,ssh" \
+ --with-coroutine=ucontext \
+@@ -179,6 +179,7 @@
+ --enable-opengl \
+ --enable-pie \
+ --enable-rbd \
++--enable-vitastor \
+ --enable-rdma \
+ --enable-seccomp \
+ --enable-snappy \
+diff --git a/scripts/meson-buildoptions.sh b/scripts/meson-buildoptions.sh
+index 359b04e0e6..f5b85ba78c 100644
+--- a/scripts/meson-buildoptions.sh
++++ b/scripts/meson-buildoptions.sh
+@@ -135,6 +135,7 @@ meson_options_help() {
+   printf "%s\n" '  qed             qed image format support'
+   printf "%s\n" '  qga-vss         build QGA VSS support (broken with MinGW)'
+   printf "%s\n" '  rbd             Ceph block device driver'
++  printf "%s\n" '  vitastor        Vitastor block device driver'
+   printf "%s\n" '  rdma            Enable RDMA-based migration'
+   printf "%s\n" '  replication     replication support'
+   printf "%s\n" '  sdl             SDL user interface'
+@@ -370,6 +371,8 @@ _meson_option_parse() {
+     --disable-qom-cast-debug) printf "%s" -Dqom_cast_debug=false ;;
+     --enable-rbd) printf "%s" -Drbd=enabled ;;
+     --disable-rbd) printf "%s" -Drbd=disabled ;;
++    --enable-vitastor) printf "%s" -Dvitastor=enabled ;;
++    --disable-vitastor) printf "%s" -Dvitastor=disabled ;;
+     --enable-rdma) printf "%s" -Drdma=enabled ;;
+     --disable-rdma) printf "%s" -Drdma=disabled ;;
+     --enable-replication) printf "%s" -Dreplication=enabled ;;

patches/qemu-7.2-vitastor.patch

@@ -0,0 +1,190 @@
+diff --git a/block/meson.build b/block/meson.build
+index b7c68b83a3..95d8a6f15d 100644
+--- a/block/meson.build
++++ b/block/meson.build
+@@ -100,6 +100,7 @@ foreach m : [
+   [libnfs, 'nfs', files('nfs.c')],
+   [libssh, 'ssh', files('ssh.c')],
+   [rbd, 'rbd', files('rbd.c')],
++  [vitastor, 'vitastor', files('vitastor.c')],
+ ]
+   if m[0].found()
+     module_ss = ss.source_set()
+diff --git a/meson.build b/meson.build
+index 5c6b5a1c75..f31f73612e 100644
+--- a/meson.build
++++ b/meson.build
+@@ -1026,6 +1026,26 @@ if not get_option('rbd').auto() or have_block
+   endif
+ endif
+ 
++vitastor = not_found
++if not get_option('vitastor').auto() or have_block
++  libvitastor_client = cc.find_library('vitastor_client', has_headers: ['vitastor_c.h'],
++    required: get_option('vitastor'), kwargs: static_kwargs)
++  if libvitastor_client.found()
++    if cc.links('''
++      #include <vitastor_c.h>
++      int main(void) {
++        vitastor_c_create_qemu(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0);
++        return 0;
++      }''', dependencies: libvitastor_client)
++      vitastor = declare_dependency(dependencies: libvitastor_client)
++    elif get_option('vitastor').enabled()
++      error('could not link libvitastor_client')
++    else
++      warning('could not link libvitastor_client, disabling')
++    endif
++  endif
++endif
++
+ glusterfs = not_found
+ glusterfs_ftruncate_has_stat = false
+ glusterfs_iocb_has_stat = false
+@@ -1861,6 +1881,7 @@ config_host_data.set('CONFIG_NUMA', numa.found())
+ config_host_data.set('CONFIG_OPENGL', opengl.found())
+ config_host_data.set('CONFIG_PROFILER', get_option('profiler'))
+ config_host_data.set('CONFIG_RBD', rbd.found())
++config_host_data.set('CONFIG_VITASTOR', vitastor.found())
+ config_host_data.set('CONFIG_RDMA', rdma.found())
+ config_host_data.set('CONFIG_SDL', sdl.found())
+ config_host_data.set('CONFIG_SDL_IMAGE', sdl_image.found())
+@@ -3945,6 +3966,7 @@ if spice_protocol.found()
+   summary_info += {'  spice server support': spice}
+ endif
+ summary_info += {'rbd support':       rbd}
++summary_info += {'vitastor support':  vitastor}
+ summary_info += {'smartcard support': cacard}
+ summary_info += {'U2F support':       u2f}
+ summary_info += {'libusb':            libusb}
+diff --git a/meson_options.txt b/meson_options.txt
+index 4b749ca549..6b37bd6b77 100644
+--- a/meson_options.txt
++++ b/meson_options.txt
+@@ -169,6 +169,8 @@ option('lzo', type : 'feature', value : 'auto',
+        description: 'lzo compression support')
+ option('rbd', type : 'feature', value : 'auto',
+        description: 'Ceph block device driver')
++option('vitastor', type : 'feature', value : 'auto',
++       description: 'Vitastor block device driver')
+ option('opengl', type : 'feature', value : 'auto',
+        description: 'OpenGL support')
+ option('rdma', type : 'feature', value : 'auto',
+diff --git a/qapi/block-core.json b/qapi/block-core.json
+index 95ac4fa634..7a240827e4 100644
+--- a/qapi/block-core.json
++++ b/qapi/block-core.json
+@@ -2959,7 +2959,7 @@
+             'parallels', 'preallocate', 'qcow', 'qcow2', 'qed', 'quorum',
+             'raw', 'rbd',
+             { 'name': 'replication', 'if': 'CONFIG_REPLICATION' },
+-            'ssh', 'throttle', 'vdi', 'vhdx',
++            'ssh', 'throttle', 'vdi', 'vhdx', 'vitastor',
+             { 'name': 'virtio-blk-vfio-pci', 'if': 'CONFIG_BLKIO' },
+             { 'name': 'virtio-blk-vhost-user', 'if': 'CONFIG_BLKIO' },
+             { 'name': 'virtio-blk-vhost-vdpa', 'if': 'CONFIG_BLKIO' },
+@@ -3957,6 +3957,28 @@
+             '*key-secret': 'str',
+             '*server': ['InetSocketAddressBase'] } }
+ 
++##
++# @BlockdevOptionsVitastor:
++#
++# Driver specific block device options for vitastor
++#
++# @image:       Image name
++# @inode:       Inode number
++# @pool:        Pool ID
++# @size:        Desired image size in bytes
++# @config-path: Path to Vitastor configuration
++# @etcd-host:   etcd connection address(es)
++# @etcd-prefix: etcd key/value prefix
++##
++{ 'struct': 'BlockdevOptionsVitastor',
++  'data': { '*inode': 'uint64',
++            '*pool': 'uint64',
++            '*size': 'uint64',
++            '*image': 'str',
++            '*config-path': 'str',
++            '*etcd-host': 'str',
++            '*etcd-prefix': 'str' } }
++
+ ##
+ # @ReplicationMode:
+ #
+@@ -4405,6 +4427,7 @@
+       'throttle':   'BlockdevOptionsThrottle',
+       'vdi':        'BlockdevOptionsGenericFormat',
+       'vhdx':       'BlockdevOptionsGenericFormat',
++      'vitastor':   'BlockdevOptionsVitastor',
+       'virtio-blk-vfio-pci':
+                     { 'type': 'BlockdevOptionsVirtioBlkVfioPci',
+                       'if': 'CONFIG_BLKIO' },
+@@ -4804,6 +4827,17 @@
+             '*cluster-size' :   'size',
+             '*encrypt' :        'RbdEncryptionCreateOptions' } }
+ 
++##
++# @BlockdevCreateOptionsVitastor:
++#
++# Driver specific image creation options for Vitastor.
++#
++# @size: Size of the virtual disk in bytes
++##
++{ 'struct': 'BlockdevCreateOptionsVitastor',
++  'data': { 'location':         'BlockdevOptionsVitastor',
++            'size':             'size' } }
++
+ ##
+ # @BlockdevVmdkSubformat:
+ #
+@@ -5002,6 +5036,7 @@
+       'ssh':            'BlockdevCreateOptionsSsh',
+       'vdi':            'BlockdevCreateOptionsVdi',
+       'vhdx':           'BlockdevCreateOptionsVhdx',
++      'vitastor':       'BlockdevCreateOptionsVitastor',
+       'vmdk':           'BlockdevCreateOptionsVmdk',
+       'vpc':            'BlockdevCreateOptionsVpc'
+   } }
+diff --git a/scripts/ci/org.centos/stream/8/x86_64/configure b/scripts/ci/org.centos/stream/8/x86_64/configure
+index a7f92aff90..53dc55be2e 100755
+--- a/scripts/ci/org.centos/stream/8/x86_64/configure
++++ b/scripts/ci/org.centos/stream/8/x86_64/configure
+@@ -31,7 +31,7 @@
+ --with-git=meson \
+ --with-git-submodules=update \
+ --target-list="x86_64-softmmu" \
+---block-drv-rw-whitelist="qcow2,raw,file,host_device,nbd,iscsi,rbd,blkdebug,luks,null-co,nvme,copy-on-read,throttle,gluster" \
++--block-drv-rw-whitelist="qcow2,raw,file,host_device,nbd,iscsi,rbd,vitastor,blkdebug,luks,null-co,nvme,copy-on-read,throttle,gluster" \
+ --audio-drv-list="" \
+ --block-drv-ro-whitelist="vmdk,vhdx,vpc,https,ssh" \
+ --with-coroutine=ucontext \
+@@ -179,6 +179,7 @@
+ --enable-opengl \
+ --enable-pie \
+ --enable-rbd \
++--enable-vitastor \
+ --enable-rdma \
+ --enable-seccomp \
+ --enable-snappy \
+diff --git a/scripts/meson-buildoptions.sh b/scripts/meson-buildoptions.sh
+index aa6e30ea91..c45d21c40f 100644
+--- a/scripts/meson-buildoptions.sh
++++ b/scripts/meson-buildoptions.sh
+@@ -135,6 +135,7 @@ meson_options_help() {
+   printf "%s\n" '  qed             qed image format support'
+   printf "%s\n" '  qga-vss         build QGA VSS support (broken with MinGW)'
+   printf "%s\n" '  rbd             Ceph block device driver'
++  printf "%s\n" '  vitastor        Vitastor block device driver'
+   printf "%s\n" '  rdma            Enable RDMA-based migration'
+   printf "%s\n" '  replication     replication support'
+   printf "%s\n" '  sdl             SDL user interface'
+@@ -376,6 +377,8 @@ _meson_option_parse() {
+     --disable-qom-cast-debug) printf "%s" -Dqom_cast_debug=false ;;
+     --enable-rbd) printf "%s" -Drbd=enabled ;;
+     --disable-rbd) printf "%s" -Drbd=disabled ;;
++    --enable-vitastor) printf "%s" -Dvitastor=enabled ;;
++    --disable-vitastor) printf "%s" -Dvitastor=disabled ;;
+     --enable-rdma) printf "%s" -Drdma=enabled ;;
+     --disable-rdma) printf "%s" -Drdma=disabled ;;
+     --enable-replication) printf "%s" -Dreplication=enabled ;;

patches/qemu-8.0-vitastor.patch

@@ -0,0 +1,190 @@
+diff --git a/block/meson.build b/block/meson.build
+index 382bec0e7d..af6207dbce 100644
+--- a/block/meson.build
++++ b/block/meson.build
+@@ -101,6 +101,7 @@ foreach m : [
+   [libnfs, 'nfs', files('nfs.c')],
+   [libssh, 'ssh', files('ssh.c')],
+   [rbd, 'rbd', files('rbd.c')],
++  [vitastor, 'vitastor', files('vitastor.c')],
+ ]
+   if m[0].found()
+     module_ss = ss.source_set()
+diff --git a/meson.build b/meson.build
+index c44d05a13f..ebedb42843 100644
+--- a/meson.build
++++ b/meson.build
+@@ -1028,6 +1028,26 @@ if not get_option('rbd').auto() or have_block
+   endif
+ endif
+ 
++vitastor = not_found
++if not get_option('vitastor').auto() or have_block
++  libvitastor_client = cc.find_library('vitastor_client', has_headers: ['vitastor_c.h'],
++    required: get_option('vitastor'), kwargs: static_kwargs)
++  if libvitastor_client.found()
++    if cc.links('''
++      #include <vitastor_c.h>
++      int main(void) {
++        vitastor_c_create_qemu(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0);
++        return 0;
++      }''', dependencies: libvitastor_client)
++      vitastor = declare_dependency(dependencies: libvitastor_client)
++    elif get_option('vitastor').enabled()
++      error('could not link libvitastor_client')
++    else
++      warning('could not link libvitastor_client, disabling')
++    endif
++  endif
++endif
++
+ glusterfs = not_found
+ glusterfs_ftruncate_has_stat = false
+ glusterfs_iocb_has_stat = false
+@@ -1878,6 +1898,7 @@ endif
+ config_host_data.set('CONFIG_OPENGL', opengl.found())
+ config_host_data.set('CONFIG_PROFILER', get_option('profiler'))
+ config_host_data.set('CONFIG_RBD', rbd.found())
++config_host_data.set('CONFIG_VITASTOR', vitastor.found())
+ config_host_data.set('CONFIG_RDMA', rdma.found())
+ config_host_data.set('CONFIG_SDL', sdl.found())
+ config_host_data.set('CONFIG_SDL_IMAGE', sdl_image.found())
+@@ -4002,6 +4023,7 @@ if spice_protocol.found()
+   summary_info += {'  spice server support': spice}
+ endif
+ summary_info += {'rbd support':       rbd}
++summary_info += {'vitastor support':  vitastor}
+ summary_info += {'smartcard support': cacard}
+ summary_info += {'U2F support':       u2f}
+ summary_info += {'libusb':            libusb}
+diff --git a/meson_options.txt b/meson_options.txt
+index fc9447d267..c4ac55c283 100644
+--- a/meson_options.txt
++++ b/meson_options.txt
+@@ -173,6 +173,8 @@ option('lzo', type : 'feature', value : 'auto',
+        description: 'lzo compression support')
+ option('rbd', type : 'feature', value : 'auto',
+        description: 'Ceph block device driver')
++option('vitastor', type : 'feature', value : 'auto',
++       description: 'Vitastor block device driver')
+ option('opengl', type : 'feature', value : 'auto',
+        description: 'OpenGL support')
+ option('rdma', type : 'feature', value : 'auto',
+diff --git a/qapi/block-core.json b/qapi/block-core.json
+index c05ad0c07e..f5eb701604 100644
+--- a/qapi/block-core.json
++++ b/qapi/block-core.json
+@@ -3054,7 +3054,7 @@
+             'parallels', 'preallocate', 'qcow', 'qcow2', 'qed', 'quorum',
+             'raw', 'rbd',
+             { 'name': 'replication', 'if': 'CONFIG_REPLICATION' },
+-            'ssh', 'throttle', 'vdi', 'vhdx',
++            'ssh', 'throttle', 'vdi', 'vhdx', 'vitastor',
+             { 'name': 'virtio-blk-vfio-pci', 'if': 'CONFIG_BLKIO' },
+             { 'name': 'virtio-blk-vhost-user', 'if': 'CONFIG_BLKIO' },
+             { 'name': 'virtio-blk-vhost-vdpa', 'if': 'CONFIG_BLKIO' },
+@@ -4073,6 +4073,28 @@
+             '*key-secret': 'str',
+             '*server': ['InetSocketAddressBase'] } }
+ 
++##
++# @BlockdevOptionsVitastor:
++#
++# Driver specific block device options for vitastor
++#
++# @image:       Image name
++# @inode:       Inode number
++# @pool:        Pool ID
++# @size:        Desired image size in bytes
++# @config-path: Path to Vitastor configuration
++# @etcd-host:   etcd connection address(es)
++# @etcd-prefix: etcd key/value prefix
++##
++{ 'struct': 'BlockdevOptionsVitastor',
++  'data': { '*inode': 'uint64',
++            '*pool': 'uint64',
++            '*size': 'uint64',
++            '*image': 'str',
++            '*config-path': 'str',
++            '*etcd-host': 'str',
++            '*etcd-prefix': 'str' } }
++
+ ##
+ # @ReplicationMode:
+ #
+@@ -4521,6 +4543,7 @@
+       'throttle':   'BlockdevOptionsThrottle',
+       'vdi':        'BlockdevOptionsGenericFormat',
+       'vhdx':       'BlockdevOptionsGenericFormat',
++      'vitastor':   'BlockdevOptionsVitastor',
+       'virtio-blk-vfio-pci':
+                     { 'type': 'BlockdevOptionsVirtioBlkVfioPci',
+                       'if': 'CONFIG_BLKIO' },
+@@ -4920,6 +4943,17 @@
+             '*cluster-size' :   'size',
+             '*encrypt' :        'RbdEncryptionCreateOptions' } }
+ 
++##
++# @BlockdevCreateOptionsVitastor:
++#
++# Driver specific image creation options for Vitastor.
++#
++# @size: Size of the virtual disk in bytes
++##
++{ 'struct': 'BlockdevCreateOptionsVitastor',
++  'data': { 'location':         'BlockdevOptionsVitastor',
++            'size':             'size' } }
++
+ ##
+ # @BlockdevVmdkSubformat:
+ #
+@@ -5118,6 +5152,7 @@
+       'ssh':            'BlockdevCreateOptionsSsh',
+       'vdi':            'BlockdevCreateOptionsVdi',
+       'vhdx':           'BlockdevCreateOptionsVhdx',
++      'vitastor':       'BlockdevCreateOptionsVitastor',
+       'vmdk':           'BlockdevCreateOptionsVmdk',
+       'vpc':            'BlockdevCreateOptionsVpc'
+   } }
+diff --git a/scripts/ci/org.centos/stream/8/x86_64/configure b/scripts/ci/org.centos/stream/8/x86_64/configure
+index 6e8983f39c..1b0b9fcf3e 100755
+--- a/scripts/ci/org.centos/stream/8/x86_64/configure
++++ b/scripts/ci/org.centos/stream/8/x86_64/configure
+@@ -32,7 +32,7 @@
+ --with-git=meson \
+ --with-git-submodules=update \
+ --target-list="x86_64-softmmu" \
+---block-drv-rw-whitelist="qcow2,raw,file,host_device,nbd,iscsi,rbd,blkdebug,luks,null-co,nvme,copy-on-read,throttle,gluster" \
++--block-drv-rw-whitelist="qcow2,raw,file,host_device,nbd,iscsi,rbd,vitastor,blkdebug,luks,null-co,nvme,copy-on-read,throttle,gluster" \
+ --audio-drv-list="" \
+ --block-drv-ro-whitelist="vmdk,vhdx,vpc,https,ssh" \
+ --with-coroutine=ucontext \
+@@ -179,6 +179,7 @@
+ --enable-opengl \
+ --enable-pie \
+ --enable-rbd \
++--enable-vitastor \
+ --enable-rdma \
+ --enable-seccomp \
+ --enable-snappy \
+diff --git a/scripts/meson-buildoptions.sh b/scripts/meson-buildoptions.sh
+index 009fab1515..95914e6ebc 100644
+--- a/scripts/meson-buildoptions.sh
++++ b/scripts/meson-buildoptions.sh
+@@ -142,6 +142,7 @@ meson_options_help() {
+   printf "%s\n" '  qed             qed image format support'
+   printf "%s\n" '  qga-vss         build QGA VSS support (broken with MinGW)'
+   printf "%s\n" '  rbd             Ceph block device driver'
++  printf "%s\n" '  vitastor        Vitastor block device driver'
+   printf "%s\n" '  rdma            Enable RDMA-based migration'
+   printf "%s\n" '  replication     replication support'
+   printf "%s\n" '  sdl             SDL user interface'
+@@ -388,6 +389,8 @@ _meson_option_parse() {
+     --disable-qom-cast-debug) printf "%s" -Dqom_cast_debug=false ;;
+     --enable-rbd) printf "%s" -Drbd=enabled ;;
+     --disable-rbd) printf "%s" -Drbd=disabled ;;
++    --enable-vitastor) printf "%s" -Dvitastor=enabled ;;
++    --disable-vitastor) printf "%s" -Dvitastor=disabled ;;
+     --enable-rdma) printf "%s" -Drdma=enabled ;;
+     --disable-rdma) printf "%s" -Drdma=disabled ;;
+     --enable-replication) printf "%s" -Dreplication=enabled ;;

rpm/build-tarball.sh

@@ -7,13 +7,12 @@ set -e
 VITASTOR=$(dirname $0)
 VITASTOR=$(realpath "$VITASTOR/..")
 
-if [ -d /opt/rh/gcc-toolset-9 ]; then
+EL=$(rpm --eval '%dist')
+if [ "$EL" = ".el8" ]; then
     # CentOS 8
-    EL=8
     . /opt/rh/gcc-toolset-9/enable
-else
+elif [ "$EL" = ".el7" ]; then
     # CentOS 7
-    EL=7
     . /opt/rh/devtoolset-9/enable
 fi
 cd ~/rpmbuild/SPECS
@@ -25,4 +24,4 @@ rm fio
 mv fio-copy fio
 FIO=`rpm -qi fio | perl -e 'while(<>) { /^Epoch[\s:]+(\S+)/ && print "$1:"; /^Version[\s:]+(\S+)/ && print $1; /^Release[\s:]+(\S+)/ && print "-$1"; }'`
 perl -i -pe 's/(Requires:\s*fio)([^\n]+)?/$1 = '$FIO'/' $VITASTOR/rpm/vitastor-el$EL.spec
-tar --transform 's#^#vitastor-0.8.3/#' --exclude 'rpm/*.rpm' -czf $VITASTOR/../vitastor-0.8.3$(rpm --eval '%dist').tar.gz *
+tar --transform 's#^#vitastor-0.8.8/#' --exclude 'rpm/*.rpm' -czf $VITASTOR/../vitastor-0.8.8$(rpm --eval '%dist').tar.gz *

rpm/qemu-kvm-7.0-el9.spec.patch

@@ -0,0 +1,93 @@
+--- qemu-kvm.spec.orig	2023-02-28 08:04:06.000000000 +0000
++++ qemu-kvm.spec	2023-04-27 22:29:18.094878829 +0000
+@@ -100,8 +100,6 @@
+ %endif
+ 
+ %global target_list %{kvm_target}-softmmu
+-%global block_drivers_rw_list qcow2,raw,file,host_device,nbd,iscsi,rbd,blkdebug,luks,null-co,nvme,copy-on-read,throttle,compress
+-%global block_drivers_ro_list vdi,vmdk,vhdx,vpc,https
+ %define qemudocdir %{_docdir}/%{name}
+ %global firmwaredirs "%{_datadir}/qemu-firmware:%{_datadir}/ipxe/qemu:%{_datadir}/seavgabios:%{_datadir}/seabios"
+ 
+@@ -129,6 +127,7 @@ Requires: %{name}-device-usb-host = %{ep
+ Requires: %{name}-device-usb-redirect = %{epoch}:%{version}-%{release}   \
+ %endif                                                           \
+ Requires: %{name}-block-rbd = %{epoch}:%{version}-%{release}     \
++Requires: %{name}-block-vitastor = %{epoch}:%{version}-%{release}\
+ Requires: %{name}-audio-pa = %{epoch}:%{version}-%{release}
+ 
+ # Since SPICE is removed from RHEL-9, the following Obsoletes:
+@@ -151,7 +150,7 @@ Obsoletes: %{name}-block-ssh <= %{epoch}
+ Summary: QEMU is a machine emulator and virtualizer
+ Name: qemu-kvm
+ Version: 7.0.0
+-Release: 13%{?rcrel}%{?dist}%{?cc_suffix}.2
++Release: 13.vitastor%{?rcrel}%{?dist}%{?cc_suffix}
+ # Epoch because we pushed a qemu-1.0 package. AIUI this can't ever be dropped
+ # Epoch 15 used for RHEL 8
+ # Epoch 17 used for RHEL 9 (due to release versioning offset in RHEL 8.5)
+@@ -174,6 +173,7 @@ Source28: 95-kvm-memlock.conf
+ Source30: kvm-s390x.conf
+ Source31: kvm-x86.conf
+ Source36: README.tests
++Source37: qemu-vitastor.c
+ 
+ 
+ Patch0004: 0004-Initial-redhat-build.patch
+@@ -498,6 +498,7 @@ Patch171: kvm-i386-do-kvm_put_msr_featur
+ Patch172: kvm-target-i386-kvm-fix-kvmclock_current_nsec-Assertion-.patch
+ # For bz#2168221 - while live-migrating many instances concurrently, libvirt sometimes return internal error: migration was active, but no RAM info was set [rhel-9.1.0.z]
+ Patch173: kvm-migration-Read-state-once.patch
++Patch174: qemu-7.0-vitastor.patch
+ 
+ # Source-git patches
+ 
+@@ -531,6 +532,7 @@ BuildRequires: libcurl-devel
+ %if %{have_block_rbd}
+ BuildRequires: librbd-devel
+ %endif
++BuildRequires: vitastor-client-devel
+ # We need both because the 'stap' binary is probed for by configure
+ BuildRequires: systemtap
+ BuildRequires: systemtap-sdt-devel
+@@ -718,6 +720,14 @@ using the rbd protocol.
+ %endif
+ 
+ 
++%package  block-vitastor
++Summary: QEMU Vitastor block driver
++Requires: %{name}-common%{?_isa} = %{epoch}:%{version}-%{release}
++
++%description block-vitastor
++This package provides the additional Vitastor block driver for QEMU.
++
++
+ %package  audio-pa
+ Summary: QEMU PulseAudio audio driver
+ Requires: %{name}-common%{?_isa} = %{epoch}:%{version}-%{release}
+@@ -811,6 +821,7 @@ This package provides usbredir support.
+ %prep
+ %setup -q -n qemu-%{version}%{?rcstr}
+ %autopatch -p1
++cp %{SOURCE37} ./block/vitastor.c
+ 
+ %global qemu_kvm_build qemu_kvm_build
+ mkdir -p %{qemu_kvm_build}
+@@ -1032,6 +1043,7 @@ run_configure \
+ %if %{have_block_rbd}
+   --enable-rbd \
+ %endif
++  --enable-vitastor \
+ %if %{have_librdma}
+   --enable-rdma \
+ %endif
+@@ -1511,6 +1523,9 @@ useradd -r -u 107 -g qemu -G kvm -d / -s
+ %files block-rbd
+ %{_libdir}/%{name}/block-rbd.so
+ %endif
++%files block-vitastor
++%{_libdir}/%{name}/block-vitastor.so
++
+ %files audio-pa
+ %{_libdir}/%{name}/audio-pa.so
+ 

rpm/vitastor-el7.Dockerfile

@@ -35,7 +35,7 @@ ADD . /root/vitastor
 RUN set -e; \
     cd /root/vitastor/rpm; \
     sh build-tarball.sh; \
-    cp /root/vitastor-0.8.3.el7.tar.gz ~/rpmbuild/SOURCES; \
+    cp /root/vitastor-0.8.8.el7.tar.gz ~/rpmbuild/SOURCES; \
     cp vitastor-el7.spec ~/rpmbuild/SPECS/vitastor.spec; \
     cd ~/rpmbuild/SPECS/; \
     rpmbuild -ba vitastor.spec; \

rpm/vitastor-el7.spec

@@ -1,11 +1,11 @@
 Name:           vitastor
-Version:        0.8.3
+Version:        0.8.8
 Release:        1%{?dist}
 Summary:        Vitastor, a fast software-defined clustered block storage
 
 License:        Vitastor Network Public License 1.1
 URL:            https://vitastor.io/
-Source0:        vitastor-0.8.3.el7.tar.gz
+Source0:        vitastor-0.8.8.el7.tar.gz
 
 BuildRequires:  liburing-devel >= 0.6
 BuildRequires:  gperftools-devel
@@ -35,6 +35,7 @@ Summary:        Vitastor - OSD
 Requires:       libJerasure2
 Requires:       libisa-l
 Requires:       liburing >= 0.6
+Requires:       liburing < 2
 Requires:       vitastor-client = %{version}-%{release}
 Requires:       util-linux
 Requires:       parted
@@ -59,6 +60,7 @@ scheduling cluster-level operations.
 %package -n vitastor-client
 Summary:        Vitastor - client
 Requires:       liburing >= 0.6
+Requires:       liburing < 2
 
 
 %description -n vitastor-client

rpm/vitastor-el8.Dockerfile

@@ -35,7 +35,7 @@ ADD . /root/vitastor
 RUN set -e; \
     cd /root/vitastor/rpm; \
     sh build-tarball.sh; \
-    cp /root/vitastor-0.8.3.el8.tar.gz ~/rpmbuild/SOURCES; \
+    cp /root/vitastor-0.8.8.el8.tar.gz ~/rpmbuild/SOURCES; \
     cp vitastor-el8.spec ~/rpmbuild/SPECS/vitastor.spec; \
     cd ~/rpmbuild/SPECS/; \
     rpmbuild -ba vitastor.spec; \

rpm/vitastor-el8.spec

@@ -1,11 +1,11 @@
 Name:           vitastor
-Version:        0.8.3
+Version:        0.8.8
 Release:        1%{?dist}
 Summary:        Vitastor, a fast software-defined clustered block storage
 
 License:        Vitastor Network Public License 1.1
 URL:            https://vitastor.io/
-Source0:        vitastor-0.8.3.el8.tar.gz
+Source0:        vitastor-0.8.8.el8.tar.gz
 
 BuildRequires:  liburing-devel >= 0.6
 BuildRequires:  gperftools-devel
@@ -34,6 +34,7 @@ Summary:        Vitastor - OSD
 Requires:       libJerasure2
 Requires:       libisa-l
 Requires:       liburing >= 0.6
+Requires:       liburing < 2
 Requires:       vitastor-client = %{version}-%{release}
 Requires:       util-linux
 Requires:       parted
@@ -57,6 +58,7 @@ scheduling cluster-level operations.
 %package -n vitastor-client
 Summary:        Vitastor - client
 Requires:       liburing >= 0.6
+Requires:       liburing < 2
 
 
 %description -n vitastor-client

rpm/vitastor-el9.Dockerfile

@@ -0,0 +1,28 @@
+# Build packages for AlmaLinux 9 inside a container
+# cd ..; podman build -t vitastor-el9 -v `pwd`/packages:/root/packages -f rpm/vitastor-el9.Dockerfile .
+
+FROM almalinux:9
+
+WORKDIR /root
+
+RUN sed -i 's/enabled=0/enabled=1/' /etc/yum.repos.d/*.repo
+RUN dnf -y install epel-release dnf-plugins-core
+RUN dnf -y install https://vitastor.io/rpms/centos/9/vitastor-release-1.0-1.el9.noarch.rpm
+RUN dnf -y install gcc-c++ gperftools-devel fio nodejs rpm-build jerasure-devel libisa-l-devel gf-complete-devel rdma-core-devel libarchive liburing-devel cmake
+RUN dnf download --source fio
+RUN rpm --nomd5 -i fio*.src.rpm
+RUN cd ~/rpmbuild/SPECS && dnf builddep -y --spec fio.spec
+
+ADD . /root/vitastor
+
+RUN set -e; \
+    cd /root/vitastor/rpm; \
+    sh build-tarball.sh; \
+    cp /root/vitastor-0.8.8.el9.tar.gz ~/rpmbuild/SOURCES; \
+    cp vitastor-el9.spec ~/rpmbuild/SPECS/vitastor.spec; \
+    cd ~/rpmbuild/SPECS/; \
+    rpmbuild -ba vitastor.spec; \
+    mkdir -p /root/packages/vitastor-el9; \
+    rm -rf /root/packages/vitastor-el9/*; \
+    cp ~/rpmbuild/RPMS/*/vitastor* /root/packages/vitastor-el9/; \
+    cp ~/rpmbuild/SRPMS/vitastor* /root/packages/vitastor-el9/

rpm/vitastor-el9.spec

@@ -0,0 +1,158 @@
+Name:           vitastor
+Version:        0.8.8
+Release:        1%{?dist}
+Summary:        Vitastor, a fast software-defined clustered block storage
+
+License:        Vitastor Network Public License 1.1
+URL:            https://vitastor.io/
+Source0:        vitastor-0.8.8.el9.tar.gz
+
+BuildRequires:  liburing-devel >= 0.6
+BuildRequires:  gperftools-devel
+BuildRequires:  gcc-c++
+BuildRequires:  nodejs >= 10
+BuildRequires:  jerasure-devel
+BuildRequires:  libisa-l-devel
+BuildRequires:  gf-complete-devel
+BuildRequires:  rdma-core-devel
+BuildRequires:  cmake
+Requires:       vitastor-osd = %{version}-%{release}
+Requires:       vitastor-mon = %{version}-%{release}
+Requires:       vitastor-client = %{version}-%{release}
+Requires:       vitastor-client-devel = %{version}-%{release}
+Requires:       vitastor-fio = %{version}-%{release}
+
+%description
+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).
+
+
+%package -n vitastor-osd
+Summary:        Vitastor - OSD
+Requires:       vitastor-client = %{version}-%{release}
+Requires:       util-linux
+Requires:       parted
+
+
+%description -n vitastor-osd
+Vitastor object storage daemon, i.e. server program that stores data.
+
+
+%package -n vitastor-mon
+Summary:        Vitastor - monitor
+Requires:       nodejs >= 10
+Requires:       lpsolve
+
+
+%description -n vitastor-mon
+Vitastor monitor, i.e. server program responsible for watching cluster state and
+scheduling cluster-level operations.
+
+
+%package -n vitastor-client
+Summary:        Vitastor - client
+
+
+%description -n vitastor-client
+Vitastor client library and command-line interface.
+
+
+%package -n vitastor-client-devel
+Summary:        Vitastor - development files
+Group:          Development/Libraries
+Requires:       vitastor-client = %{version}-%{release}
+
+
+%description -n vitastor-client-devel
+Vitastor library headers for development.
+
+
+%package -n vitastor-fio
+Summary:        Vitastor - fio drivers
+Group:          Development/Libraries
+Requires:       vitastor-client = %{version}-%{release}
+Requires:       fio = 3.27-7.el9
+
+
+%description -n vitastor-fio
+Vitastor fio drivers for benchmarking.
+
+
+%prep
+%setup -q
+
+
+%build
+%cmake
+%cmake_build
+
+
+%install
+rm -rf $RPM_BUILD_ROOT
+%cmake_install
+cd mon
+npm install
+cd ..
+mkdir -p %buildroot/usr/lib/vitastor
+cp -r mon %buildroot/usr/lib/vitastor
+mkdir -p %buildroot/lib/systemd/system
+cp mon/vitastor.target mon/vitastor-mon.service mon/vitastor-osd@.service %buildroot/lib/systemd/system
+mkdir -p %buildroot/lib/udev/rules.d
+cp mon/90-vitastor.rules %buildroot/lib/udev/rules.d
+
+
+%files
+%doc GPL-2.0.txt VNPL-1.1.txt README.md README-ru.md
+
+
+%files -n vitastor-osd
+%_bindir/vitastor-osd
+%_bindir/vitastor-disk
+%_bindir/vitastor-dump-journal
+/lib/systemd/system/vitastor-osd@.service
+/lib/systemd/system/vitastor.target
+/lib/udev/rules.d/90-vitastor.rules
+
+
+%pre -n vitastor-osd
+groupadd -r -f vitastor 2>/dev/null ||:
+useradd -r -g vitastor -s /sbin/nologin -c "Vitastor daemons" -M -d /nonexistent vitastor 2>/dev/null ||:
+install -o vitastor -g vitastor -d /var/log/vitastor
+mkdir -p /etc/vitastor
+
+
+%files -n vitastor-mon
+/usr/lib/vitastor/mon
+/lib/systemd/system/vitastor-mon.service
+
+
+%pre -n vitastor-mon
+groupadd -r -f vitastor 2>/dev/null ||:
+useradd -r -g vitastor -s /sbin/nologin -c "Vitastor daemons" -M -d /nonexistent vitastor 2>/dev/null ||:
+mkdir -p /etc/vitastor
+
+
+%files -n vitastor-client
+%_bindir/vitastor-nbd
+%_bindir/vitastor-nfs
+%_bindir/vitastor-cli
+%_bindir/vitastor-rm
+%_bindir/vita
+%_libdir/libvitastor_blk.so*
+%_libdir/libvitastor_client.so*
+
+
+%files -n vitastor-client-devel
+%_includedir/vitastor_c.h
+%_libdir/pkgconfig
+
+
+%files -n vitastor-fio
+%_libdir/libfio_vitastor.so
+%_libdir/libfio_vitastor_blk.so
+%_libdir/libfio_vitastor_sec.so
+
+
+%changelog

src/CMakeLists.txt

@@ -1,8 +1,9 @@
-cmake_minimum_required(VERSION 2.8)
+cmake_minimum_required(VERSION 2.8.12)
 
 project(vitastor)
 
 include(GNUInstallDirs)
+include(CTest)
 
 set(WITH_QEMU false CACHE BOOL "Build QEMU driver inside Vitastor source tree")
 set(WITH_FIO true CACHE BOOL "Build FIO driver")
@@ -15,7 +16,7 @@ if("${CMAKE_INSTALL_PREFIX}" MATCHES "^/usr/local/?$")
 	set(CMAKE_INSTALL_RPATH "${CMAKE_INSTALL_PREFIX}/${CMAKE_INSTALL_LIBDIR}")
 endif()
 
-add_definitions(-DVERSION="0.8.3")
+add_definitions(-DVERSION="0.8.8")
 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)
@@ -55,6 +56,14 @@ if (ISAL_LIBRARIES)
 	add_definitions(-DWITH_ISAL)
 endif (ISAL_LIBRARIES)
 
+add_custom_target(build_tests)
+add_custom_target(test
+	COMMAND
+	echo leak:tcmalloc > ${CMAKE_CURRENT_BINARY_DIR}/lsan-suppress.txt &&
+	env LSAN_OPTIONS=suppressions=${CMAKE_CURRENT_BINARY_DIR}/lsan-suppress.txt ${CMAKE_CTEST_COMMAND}
+)
+add_dependencies(test build_tests)
+
 include_directories(
 	../
 	/usr/include/jerasure
@@ -145,7 +154,6 @@ add_library(vitastor_client SHARED
 set_target_properties(vitastor_client PROPERTIES PUBLIC_HEADER "vitastor_c.h")
 target_link_libraries(vitastor_client
 	vitastor_common
-	tcmalloc_minimal
 	${LIBURING_LIBRARIES}
 	${IBVERBS_LIBRARIES}
 )
@@ -235,8 +243,18 @@ add_executable(osd_test osd_test.cpp rw_blocking.cpp addr_util.cpp)
 target_link_libraries(osd_test tcmalloc_minimal)
 
 # osd_rmw_test
-add_executable(osd_rmw_test osd_rmw_test.cpp allocator.cpp)
+add_executable(osd_rmw_test EXCLUDE_FROM_ALL osd_rmw_test.cpp allocator.cpp)
 target_link_libraries(osd_rmw_test Jerasure ${ISAL_LIBRARIES} tcmalloc_minimal)
+add_dependencies(build_tests osd_rmw_test)
+add_test(NAME osd_rmw_test COMMAND osd_rmw_test)
+
+if (ISAL_LIBRARIES)
+	add_executable(osd_rmw_test_je EXCLUDE_FROM_ALL osd_rmw_test.cpp allocator.cpp)
+	target_compile_definitions(osd_rmw_test_je PUBLIC -DNO_ISAL)
+	target_link_libraries(osd_rmw_test_je Jerasure tcmalloc_minimal)
+	add_dependencies(build_tests osd_rmw_test_je)
+	add_test(NAME osd_rmw_test_jerasure COMMAND osd_rmw_test_je)
+endif (ISAL_LIBRARIES)
 
 # stub_uring_osd
 add_executable(stub_uring_osd
@@ -250,11 +268,15 @@ target_link_libraries(stub_uring_osd
 )
 
 # osd_peering_pg_test
-add_executable(osd_peering_pg_test osd_peering_pg_test.cpp osd_peering_pg.cpp)
+add_executable(osd_peering_pg_test EXCLUDE_FROM_ALL osd_peering_pg_test.cpp osd_peering_pg.cpp)
 target_link_libraries(osd_peering_pg_test tcmalloc_minimal)
+add_dependencies(build_tests osd_peering_pg_test)
+add_test(NAME osd_peering_pg_test COMMAND osd_peering_pg_test)
 
 # test_allocator
-add_executable(test_allocator test_allocator.cpp allocator.cpp)
+add_executable(test_allocator EXCLUDE_FROM_ALL test_allocator.cpp allocator.cpp)
+add_dependencies(build_tests test_allocator)
+add_test(NAME test_allocator COMMAND test_allocator)
 
 # test_cas
 add_executable(test_cas
@@ -274,12 +296,15 @@ target_link_libraries(test_crc32
 
 # test_cluster_client
 add_executable(test_cluster_client
+	EXCLUDE_FROM_ALL
 	test_cluster_client.cpp
 	pg_states.cpp osd_ops.cpp cluster_client.cpp cluster_client_list.cpp msgr_op.cpp mock/messenger.cpp msgr_stop.cpp
 	etcd_state_client.cpp timerfd_manager.cpp ../json11/json11.cpp
 )
 target_compile_definitions(test_cluster_client PUBLIC -D__MOCK__)
 target_include_directories(test_cluster_client PUBLIC ${CMAKE_SOURCE_DIR}/src/mock)
+add_dependencies(build_tests test_cluster_client)
+add_test(NAME test_cluster_client COMMAND test_cluster_client)
 
 ## test_blockstore, test_shit
 #add_executable(test_blockstore test_blockstore.cpp)

src/blockstore.cpp

@@ -13,6 +13,11 @@ blockstore_t::~blockstore_t()
     delete impl;
 }
 
+void blockstore_t::parse_config(blockstore_config_t & config)
+{
+    impl->parse_config(config, false);
+}
+
 void blockstore_t::loop()
 {
     impl->loop();

src/blockstore.h

@@ -107,7 +107,7 @@ Input:
 - buf = pre-allocated obj_ver_id array <len> units long
 
 Output:
-- retval = 0 or negative error number (-EINVAL, -ENOENT if no such version or -EBUSY if not synced)
+- retval = 0 or negative error number (-ENOENT if no such version for stabilize)
 
 ## BS_OP_SYNC_STAB_ALL
 
@@ -165,6 +165,9 @@ public:
     blockstore_t(blockstore_config_t & config, ring_loop_t *ringloop, timerfd_manager_t *tfd);
     ~blockstore_t();
 
+    // Update configuration
+    void parse_config(blockstore_config_t & config);
+
     // Event loop
     void loop();
 

src/blockstore_flush.cpp

@@ -162,7 +162,8 @@ void journal_flusher_t::mark_trim_possible()
     if (trim_wanted > 0)
     {
         dequeuing = true;
-        journal_trim_counter++;
+        if (!journal_trim_counter)
+            journal_trim_counter = journal_trim_interval;
         bs->ringloop->wakeup();
     }
 }
@@ -931,7 +932,7 @@ bool journal_flusher_co::fsync_batch(bool fsync_meta, int wait_base)
     resume_1:
         if (!cur_sync->state)
         {
-            if (flusher->syncing_flushers >= flusher->cur_flusher_count || !flusher->flush_queue.size())
+            if (flusher->syncing_flushers >= flusher->active_flushers || !flusher->flush_queue.size())
             {
                 // Sync batch is ready. Do it.
                 await_sqe(0);

src/blockstore_impl.cpp

@@ -11,7 +11,7 @@ blockstore_impl_t::blockstore_impl_t(blockstore_config_t & config, ring_loop_t *
     ring_consumer.loop = [this]() { loop(); };
     ringloop->register_consumer(&ring_consumer);
     initialized = 0;
-    parse_config(config);
+    parse_config(config, true);
     zero_object = (uint8_t*)memalign_or_die(MEM_ALIGNMENT, dsk.data_block_size);
     try
     {
@@ -171,7 +171,7 @@ void blockstore_impl_t::loop()
                     // Can't submit SYNC before previous writes
                     continue;
                 }
-                wr_st = continue_sync(op, false);
+                wr_st = continue_sync(op);
                 if (wr_st != 2)
                 {
                     has_writes = wr_st > 0 ? 1 : 2;
@@ -193,6 +193,7 @@ void blockstore_impl_t::loop()
             }
             if (wr_st == 2)
             {
+                submit_queue[op_idx] = NULL;
                 new_idx--;
             }
             if (wr_st == 0)
@@ -306,6 +307,18 @@ void blockstore_impl_t::check_wait(blockstore_op_t *op)
         }
         PRIV(op)->wait_for = 0;
     }
+    else if (PRIV(op)->wait_for == WAIT_FREE)
+    {
+        if (!data_alloc->get_free_count() && big_to_flush > 0)
+        {
+#ifdef BLOCKSTORE_DEBUG
+            printf("Still waiting for free space on the data device\n");
+#endif
+            return;
+        }
+        flusher->release_trim();
+        PRIV(op)->wait_for = 0;
+    }
     else
     {
         throw std::runtime_error("BUG: op->wait_for value is unexpected");
@@ -324,7 +337,7 @@ void blockstore_impl_t::enqueue_op(blockstore_op_t *op)
     {
         // Basic verification not passed
         op->retval = -EINVAL;
-        std::function<void (blockstore_op_t*)>(op->callback)(op);
+        ringloop->set_immediate([op]() { std::function<void (blockstore_op_t*)>(op->callback)(op); });
         return;
     }
     if (op->opcode == BS_OP_SYNC_STAB_ALL)
@@ -367,16 +380,21 @@ void blockstore_impl_t::enqueue_op(blockstore_op_t *op)
     }
     if ((op->opcode == BS_OP_WRITE || op->opcode == BS_OP_WRITE_STABLE || op->opcode == BS_OP_DELETE) && !enqueue_write(op))
     {
-        std::function<void (blockstore_op_t*)>(op->callback)(op);
+        ringloop->set_immediate([op]() { std::function<void (blockstore_op_t*)>(op->callback)(op); });
         return;
     }
+    init_op(op);
+    submit_queue.push_back(op);
+    ringloop->wakeup();
+}
+
+void blockstore_impl_t::init_op(blockstore_op_t *op)
+{
     // Call constructor without allocating memory. We'll call destructor before returning op back
     new ((void*)op->private_data) blockstore_op_private_t;
     PRIV(op)->wait_for = 0;
     PRIV(op)->op_state = 0;
     PRIV(op)->pending_ops = 0;
-    submit_queue.push_back(op);
-    ringloop->wakeup();
 }
 
 static bool replace_stable(object_id oid, uint64_t version, int search_start, int search_end, obj_ver_id* list)
@@ -582,7 +600,7 @@ void blockstore_impl_t::process_list(blockstore_op_t *op)
                         replace_stable(dirty_it->first.oid, 0, clean_stable_count, stable_count, stable);
                     }
                 }
-                else if (IS_STABLE(dirty_it->second.state))
+                else if (IS_STABLE(dirty_it->second.state) || (dirty_it->second.state & BS_ST_INSTANT))
                 {
                     // First try to replace a clean stable version in the first part of the list
                     if (!replace_stable(dirty_it->first.oid, dirty_it->first.version, 0, clean_stable_count, stable))

src/blockstore_impl.h

@@ -160,6 +160,8 @@ struct __attribute__((__packed__)) dirty_entry
 #define WAIT_JOURNAL 3
 // Suspend operation until the next journal sector buffer is free
 #define WAIT_JOURNAL_BUFFER 4
+// Suspend operation until there is some free space on the data device
+#define WAIT_FREE 5
 
 struct fulfill_read_t
 {
@@ -216,6 +218,11 @@ struct pool_shard_settings_t
     uint32_t pg_stripe_size;
 };
 
+#define STAB_SPLIT_DONE 1
+#define STAB_SPLIT_WAIT 2
+#define STAB_SPLIT_SYNC 3
+#define STAB_SPLIT_TODO 4
+
 class blockstore_impl_t
 {
     blockstore_disk_t dsk;
@@ -258,6 +265,7 @@ class blockstore_impl_t
 
     struct journal_t journal;
     journal_flusher_t *flusher;
+    int big_to_flush = 0;
     int write_iodepth = 0;
 
     bool live = false, queue_stall = false;
@@ -277,7 +285,6 @@ class blockstore_impl_t
     friend class journal_flusher_t;
     friend class journal_flusher_co;
 
-    void parse_config(blockstore_config_t & config);
     void calc_lengths();
     void open_data();
     void open_meta();
@@ -299,6 +306,7 @@ class blockstore_impl_t
     blockstore_init_journal* journal_init_reader;
 
     void check_wait(blockstore_op_t *op);
+    void init_op(blockstore_op_t *op);
 
     // Read
     int dequeue_read(blockstore_op_t *read_op);
@@ -318,7 +326,7 @@ class blockstore_impl_t
     void handle_write_event(ring_data_t *data, blockstore_op_t *op);
 
     // Sync
-    int continue_sync(blockstore_op_t *op, bool queue_has_in_progress_sync);
+    int continue_sync(blockstore_op_t *op);
     void ack_sync(blockstore_op_t *op);
 
     // Stabilize
@@ -326,6 +334,8 @@ class blockstore_impl_t
     int continue_stable(blockstore_op_t *op);
     void mark_stable(const obj_ver_id & ov, bool forget_dirty = false);
     void stabilize_object(object_id oid, uint64_t max_ver);
+    blockstore_op_t* selective_sync(blockstore_op_t *op);
+    int split_stab_op(blockstore_op_t *op, std::function<int(obj_ver_id v)> decider);
 
     // Rollback
     int dequeue_rollback(blockstore_op_t *op);
@@ -341,6 +351,8 @@ public:
     blockstore_impl_t(blockstore_config_t & config, ring_loop_t *ringloop, timerfd_manager_t *tfd);
     ~blockstore_impl_t();
 
+    void parse_config(blockstore_config_t & config, bool init);
+
     // Event loop
     void loop();
 

src/blockstore_journal.h

@@ -16,6 +16,7 @@
 // FIXME: This value should be dynamic i.e. Blockstore ideally shouldn't allow
 // writing more than can be stabilized afterwards
 #define JOURNAL_STABILIZE_RESERVATION 65536
+#define JOURNAL_INSTANT_RESERVATION 131072
 
 // Journal entries
 // Journal entries are linked to each other by their crc32 value

src/blockstore_open.cpp

@@ -4,8 +4,54 @@
 #include <sys/file.h>
 #include "blockstore_impl.h"
 
-void blockstore_impl_t::parse_config(blockstore_config_t & config)
+void blockstore_impl_t::parse_config(blockstore_config_t & config, bool init)
 {
+    // Online-configurable options:
+    max_flusher_count = strtoull(config["max_flusher_count"].c_str(), NULL, 10);
+    if (!max_flusher_count)
+    {
+        max_flusher_count = strtoull(config["flusher_count"].c_str(), NULL, 10);
+    }
+    min_flusher_count = strtoull(config["min_flusher_count"].c_str(), NULL, 10);
+    max_write_iodepth = strtoull(config["max_write_iodepth"].c_str(), NULL, 10);
+    throttle_small_writes = config["throttle_small_writes"] == "true" || config["throttle_small_writes"] == "1" || config["throttle_small_writes"] == "yes";
+    throttle_target_iops = strtoull(config["throttle_target_iops"].c_str(), NULL, 10);
+    throttle_target_mbs = strtoull(config["throttle_target_mbs"].c_str(), NULL, 10);
+    throttle_target_parallelism = strtoull(config["throttle_target_parallelism"].c_str(), NULL, 10);
+    throttle_threshold_us = strtoull(config["throttle_threshold_us"].c_str(), NULL, 10);
+    if (!max_flusher_count)
+    {
+        max_flusher_count = 256;
+    }
+    if (!min_flusher_count || journal.flush_journal)
+    {
+        min_flusher_count = 1;
+    }
+    if (!max_write_iodepth)
+    {
+        max_write_iodepth = 128;
+    }
+    if (!throttle_target_iops)
+    {
+        throttle_target_iops = 100;
+    }
+    if (!throttle_target_mbs)
+    {
+        throttle_target_mbs = 100;
+    }
+    if (!throttle_target_parallelism)
+    {
+        throttle_target_parallelism = 1;
+    }
+    if (!throttle_threshold_us)
+    {
+        throttle_threshold_us = 50;
+    }
+    if (!init)
+    {
+        return;
+    }
+    // Offline-configurable options:
     // Common disk options
     dsk.parse_config(config);
     // Parse
@@ -44,29 +90,7 @@ void blockstore_impl_t::parse_config(blockstore_config_t & config)
     journal.no_same_sector_overwrites = config["journal_no_same_sector_overwrites"] == "true" ||
         config["journal_no_same_sector_overwrites"] == "1" || config["journal_no_same_sector_overwrites"] == "yes";
     journal.inmemory = config["inmemory_journal"] != "false";
-    max_flusher_count = strtoull(config["max_flusher_count"].c_str(), NULL, 10);
-    if (!max_flusher_count)
-        max_flusher_count = strtoull(config["flusher_count"].c_str(), NULL, 10);
-    min_flusher_count = strtoull(config["min_flusher_count"].c_str(), NULL, 10);
-    max_write_iodepth = strtoull(config["max_write_iodepth"].c_str(), NULL, 10);
-    throttle_small_writes = config["throttle_small_writes"] == "true" || config["throttle_small_writes"] == "1" || config["throttle_small_writes"] == "yes";
-    throttle_target_iops = strtoull(config["throttle_target_iops"].c_str(), NULL, 10);
-    throttle_target_mbs = strtoull(config["throttle_target_mbs"].c_str(), NULL, 10);
-    throttle_target_parallelism = strtoull(config["throttle_target_parallelism"].c_str(), NULL, 10);
-    throttle_threshold_us = strtoull(config["throttle_threshold_us"].c_str(), NULL, 10);
     // Validate
-    if (!max_flusher_count)
-    {
-        max_flusher_count = 256;
-    }
-    if (!min_flusher_count || journal.flush_journal)
-    {
-        min_flusher_count = 1;
-    }
-    if (!max_write_iodepth)
-    {
-        max_write_iodepth = 128;
-    }
     if (journal.sector_count < 2)
     {
         journal.sector_count = 32;
@@ -91,22 +115,6 @@ void blockstore_impl_t::parse_config(blockstore_config_t & config)
     {
         throw std::runtime_error("immediate_commit=all requires disable_journal_fsync and disable_data_fsync");
     }
-    if (!throttle_target_iops)
-    {
-        throttle_target_iops = 100;
-    }
-    if (!throttle_target_mbs)
-    {
-        throttle_target_mbs = 100;
-    }
-    if (!throttle_target_parallelism)
-    {
-        throttle_target_parallelism = 1;
-    }
-    if (!throttle_threshold_us)
-    {
-        throttle_threshold_us = 50;
-    }
     // init some fields
     journal.block_size = dsk.journal_block_size;
     journal.next_free = dsk.journal_block_size;

src/blockstore_read.cpp

@@ -286,7 +286,10 @@ void blockstore_impl_t::handle_read_event(ring_data_t *data, blockstore_op_t *op
                 {
                     auto used = --journal.used_sectors[rv.journal_sector-1];
                     if (used == 0)
+                    {
                         journal.used_sectors.erase(rv.journal_sector-1);
+                        flusher->mark_trim_possible();
+                    }
                 }
             }
         }

src/blockstore_rollback.cpp

@@ -9,48 +9,39 @@ int blockstore_impl_t::dequeue_rollback(blockstore_op_t *op)
     {
         return continue_rollback(op);
     }
-    obj_ver_id *v, *nv;
-    int i, todo = op->len;
-    for (i = 0, v = (obj_ver_id*)op->buf, nv = (obj_ver_id*)op->buf; i < op->len; i++, v++, nv++)
+    int r = split_stab_op(op, [this](obj_ver_id ov)
     {
-        if (nv != v)
-        {
-            *nv = *v;
-        }
         // Check that there are some versions greater than v->version (which may be zero),
         // check that they're unstable, synced, and not currently written to
         auto dirty_it = dirty_db.lower_bound((obj_ver_id){
-            .oid = v->oid,
+            .oid = ov.oid,
             .version = UINT64_MAX,
         });
         if (dirty_it == dirty_db.begin())
         {
-skip_ov:
             // Already rolled back, skip this object version
-            todo--;
-            nv--;
-            continue;
+            return STAB_SPLIT_DONE;
         }
         else
         {
             dirty_it--;
-            if (dirty_it->first.oid != v->oid || dirty_it->first.version < v->version)
+            if (dirty_it->first.oid != ov.oid || dirty_it->first.version < ov.version)
             {
-                goto skip_ov;
+                // Already rolled back, skip this object version
+                return STAB_SPLIT_DONE;
             }
-            while (dirty_it->first.oid == v->oid && dirty_it->first.version > v->version)
+            while (dirty_it->first.oid == ov.oid && dirty_it->first.version > ov.version)
             {
                 if (IS_IN_FLIGHT(dirty_it->second.state))
                 {
                     // Object write is still in progress. Wait until the write request completes
-                    return 0;
+                    return STAB_SPLIT_WAIT;
                 }
                 else if (!IS_SYNCED(dirty_it->second.state) ||
                     IS_STABLE(dirty_it->second.state))
                 {
-                    op->retval = -EBUSY;
-                    FINISH_OP(op);
-                    return 2;
+                    // Sync the object
+                    return STAB_SPLIT_SYNC;
                 }
                 if (dirty_it == dirty_db.begin())
                 {
@@ -58,19 +49,16 @@ skip_ov:
                 }
                 dirty_it--;
             }
+            return STAB_SPLIT_TODO;
         }
-    }
-    op->len = todo;
-    if (!todo)
+    });
+    if (r != 1)
     {
-        // Already rolled back
-        op->retval = 0;
-        FINISH_OP(op);
-        return 2;
+        return r;
     }
     // Check journal space
     blockstore_journal_check_t space_check(this);
-    if (!space_check.check_available(op, todo, sizeof(journal_entry_rollback), 0))
+    if (!space_check.check_available(op, op->len, sizeof(journal_entry_rollback), 0))
     {
         return 0;
     }
@@ -78,7 +66,8 @@ skip_ov:
     BS_SUBMIT_CHECK_SQES(space_check.sectors_to_write);
     // Prepare and submit journal entries
     int s = 0;
-    for (i = 0, v = (obj_ver_id*)op->buf; i < op->len; i++, v++)
+    auto v = (obj_ver_id*)op->buf;
+    for (int i = 0; i < op->len; i++, v++)
     {
         if (!journal.entry_fits(sizeof(journal_entry_rollback)) &&
             journal.sector_info[journal.cur_sector].dirty)
@@ -127,7 +116,6 @@ resume_4:
     {
         mark_rolled_back(*v);
     }
-    flusher->mark_trim_possible();
     // Acknowledge op
     op->retval = 0;
     FINISH_OP(op);
@@ -213,6 +201,11 @@ void blockstore_impl_t::erase_dirty(blockstore_dirty_db_t::iterator dirty_start,
     }
     while (1)
     {
+        if ((IS_BIG_WRITE(dirty_it->second.state) || IS_DELETE(dirty_it->second.state)) &&
+            IS_STABLE(dirty_it->second.state))
+        {
+            big_to_flush--;
+        }
         if (IS_BIG_WRITE(dirty_it->second.state) && dirty_it->second.location != clean_loc &&
             dirty_it->second.location != UINT64_MAX)
         {
@@ -232,6 +225,7 @@ void blockstore_impl_t::erase_dirty(blockstore_dirty_db_t::iterator dirty_start,
         if (used == 0)
         {
             journal.used_sectors.erase(dirty_it->second.journal_sector);
+            flusher->mark_trim_possible();
         }
         if (dsk.clean_entry_bitmap_size > sizeof(void*))
         {

src/blockstore_stable.cpp

@@ -41,60 +41,309 @@
 // 4) after a while it takes his synced object list and sends stabilize requests
 //    to peers and to its own blockstore, thus freeing the old version
 
-int blockstore_impl_t::dequeue_stable(blockstore_op_t *op)
+struct ver_vector_t
 {
-    if (PRIV(op)->op_state)
+    obj_ver_id *items = NULL;
+    uint64_t alloc = 0, size = 0;
+};
+
+static void init_versions(ver_vector_t & vec, obj_ver_id *start, obj_ver_id *end, uint64_t len)
+{
+    if (!vec.items)
     {
-        return continue_stable(op);
+        vec.alloc = len;
+        vec.items = (obj_ver_id*)malloc_or_die(sizeof(obj_ver_id) * vec.alloc);
+        for (auto sv = start; sv < end; sv++)
+        {
+            vec.items[vec.size++] = *sv;
+        }
     }
+}
+
+static void append_version(ver_vector_t & vec, obj_ver_id ov)
+{
+    if (vec.size >= vec.alloc)
+    {
+        vec.alloc = !vec.alloc ? 4 : vec.alloc*2;
+        vec.items = (obj_ver_id*)realloc_or_die(vec.items, sizeof(obj_ver_id) * vec.alloc);
+    }
+    vec.items[vec.size++] = ov;
+}
+
+static bool check_unsynced(std::vector<obj_ver_id> & check, obj_ver_id ov, std::vector<obj_ver_id> & to, int *count)
+{
+    bool found = false;
+    int j = 0, k = 0;
+    while (j < check.size())
+    {
+        if (check[j] == ov)
+            found = true;
+        if (check[j].oid == ov.oid && check[j].version <= ov.version)
+        {
+            to.push_back(check[j++]);
+            if (count)
+                (*count)--;
+        }
+        else
+            check[k++] = check[j++];
+    }
+    check.resize(k);
+    return found;
+}
+
+blockstore_op_t* blockstore_impl_t::selective_sync(blockstore_op_t *op)
+{
+    unsynced_big_write_count -= unsynced_big_writes.size();
+    unsynced_big_writes.swap(PRIV(op)->sync_big_writes);
+    unsynced_big_write_count += unsynced_big_writes.size();
+    unsynced_small_writes.swap(PRIV(op)->sync_small_writes);
+    // Create a sync operation, insert into the end of the queue
+    // And move ourselves into the end too!
+    // Rather hacky but that's what we need...
+    blockstore_op_t *sync_op = new blockstore_op_t;
+    sync_op->opcode = BS_OP_SYNC;
+    sync_op->buf = NULL;
+    sync_op->callback = [this](blockstore_op_t *sync_op)
+    {
+        delete sync_op;
+    };
+    init_op(sync_op);
+    int sync_res = continue_sync(sync_op);
+    if (sync_res != 2)
+    {
+        // Put SYNC into the queue if it's not finished yet
+        submit_queue.push_back(sync_op);
+    }
+    // Restore unsynced_writes
+    unsynced_small_writes.swap(PRIV(op)->sync_small_writes);
+    unsynced_big_write_count -= unsynced_big_writes.size();
+    unsynced_big_writes.swap(PRIV(op)->sync_big_writes);
+    unsynced_big_write_count += unsynced_big_writes.size();
+    if (sync_res == 2)
+    {
+        // Sync is immediately completed
+        return NULL;
+    }
+    return sync_op;
+}
+
+// Returns: 2 = stop processing and dequeue, 0 = stop processing and do not dequeue, 1 = proceed with op itself
+int blockstore_impl_t::split_stab_op(blockstore_op_t *op, std::function<int(obj_ver_id v)> decider)
+{
+    bool add_sync = false;
+    ver_vector_t good_vers, bad_vers;
     obj_ver_id* v;
     int i, todo = 0;
     for (i = 0, v = (obj_ver_id*)op->buf; i < op->len; i++, v++)
     {
-        auto dirty_it = dirty_db.find(*v);
-        if (dirty_it == dirty_db.end())
+        int action = decider(*v);
+        if (action < 0)
         {
-            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)
+            // Rollback changes
+            for (auto & ov: PRIV(op)->sync_big_writes)
             {
-                // No such object version
-                op->retval = -ENOENT;
-                FINISH_OP(op);
-                return 2;
+                unsynced_big_writes.push_back(ov);
+                unsynced_big_write_count++;
             }
-            else
+            for (auto & ov: PRIV(op)->sync_small_writes)
             {
-                // Already stable
+                unsynced_small_writes.push_back(ov);
             }
-        }
-        else if (IS_IN_FLIGHT(dirty_it->second.state))
-        {
-            // Object write is still in progress. Wait until the write request completes
-            return 0;
-        }
-        else if (!IS_SYNCED(dirty_it->second.state))
-        {
-            // Object not synced yet. Caller must sync it first
-            op->retval = -EBUSY;
+            free(good_vers.items);
+            good_vers.items = NULL;
+            free(bad_vers.items);
+            bad_vers.items = NULL;
+            // Error
+            op->retval = action;
             FINISH_OP(op);
             return 2;
         }
-        else if (!IS_STABLE(dirty_it->second.state))
+        else if (action == STAB_SPLIT_DONE)
         {
+            // Already done
+            init_versions(good_vers, (obj_ver_id*)op->buf, v, op->len);
+        }
+        else if (action == STAB_SPLIT_WAIT)
+        {
+            // Already in progress, we just have to wait until it finishes
+            init_versions(good_vers, (obj_ver_id*)op->buf, v, op->len);
+            append_version(bad_vers, *v);
+        }
+        else if (action == STAB_SPLIT_SYNC)
+        {
+            // Needs a SYNC, we have to send a SYNC if not already in progress
+            //
+            // If the object is not present in unsynced_(big|small)_writes then
+            // it's currently being synced. If it's present then we can initiate
+            // its sync ourselves.
+            init_versions(good_vers, (obj_ver_id*)op->buf, v, op->len);
+            append_version(bad_vers, *v);
+            if (!add_sync)
+            {
+                PRIV(op)->sync_big_writes.clear();
+                PRIV(op)->sync_small_writes.clear();
+                add_sync = true;
+            }
+            check_unsynced(unsynced_small_writes, *v, PRIV(op)->sync_small_writes, NULL);
+            check_unsynced(unsynced_big_writes, *v, PRIV(op)->sync_big_writes, &unsynced_big_write_count);
+        }
+        else /* if (action == STAB_SPLIT_TODO) */
+        {
+            if (good_vers.items)
+            {
+                // If we're selecting versions then append it
+                // Main idea is that 99% of the time all versions passed to BS_OP_STABLE are synced
+                // And we don't want to select/allocate anything in that optimistic case
+                append_version(good_vers, *v);
+            }
             todo++;
         }
     }
-    if (!todo)
+    // In a pessimistic scenario, an operation may be split into 3:
+    // - Stabilize synced entries
+    // - Sync unsynced entries
+    // - Continue for unsynced entries after sync
+    add_sync = add_sync && (PRIV(op)->sync_big_writes.size() || PRIV(op)->sync_small_writes.size());
+    if (!todo && !bad_vers.size)
     {
         // Already stable
         op->retval = 0;
         FINISH_OP(op);
         return 2;
     }
+    op->retval = 0;
+    if (!todo && !add_sync)
+    {
+        // Only wait for inflight writes or current in-progress syncs
+        return 0;
+    }
+    blockstore_op_t *sync_op = NULL, *split_stab_op = NULL;
+    if (add_sync)
+    {
+        // Initiate a selective sync for PRIV(op)->sync_(big|small)_writes
+        sync_op = selective_sync(op);
+    }
+    if (bad_vers.size)
+    {
+        // Split part of the request into a separate operation
+        split_stab_op = new blockstore_op_t;
+        split_stab_op->opcode = op->opcode;
+        split_stab_op->buf = bad_vers.items;
+        split_stab_op->len = bad_vers.size;
+        init_op(split_stab_op);
+        submit_queue.push_back(split_stab_op);
+    }
+    if (sync_op || split_stab_op || good_vers.items)
+    {
+        void *orig_buf = op->buf;
+        if (good_vers.items)
+        {
+            op->buf = good_vers.items;
+            op->len = good_vers.size;
+        }
+        // Make a wrapped callback
+        int *split_op_counter = (int*)malloc_or_die(sizeof(int));
+        *split_op_counter = (sync_op ? 1 : 0) + (split_stab_op ? 1 : 0) + (todo ? 1 : 0);
+        auto cb = [this, op, good_items = good_vers.items,
+            bad_items = bad_vers.items, split_op_counter,
+            orig_buf, real_cb = op->callback](blockstore_op_t *split_op)
+        {
+            if (split_op->retval != 0)
+                op->retval = split_op->retval;
+            (*split_op_counter)--;
+            assert((*split_op_counter) >= 0);
+            if (op != split_op)
+                delete split_op;
+            if (!*split_op_counter)
+            {
+                free(good_items);
+                free(bad_items);
+                free(split_op_counter);
+                op->buf = orig_buf;
+                real_cb(op);
+            }
+        };
+        if (sync_op)
+        {
+            sync_op->callback = cb;
+        }
+        if (split_stab_op)
+        {
+            split_stab_op->callback = cb;
+        }
+        op->callback = cb;
+    }
+    if (!todo)
+    {
+        // All work is postponed
+        op->callback = NULL;
+        return 2;
+    }
+    return 1;
+}
+
+int blockstore_impl_t::dequeue_stable(blockstore_op_t *op)
+{
+    if (PRIV(op)->op_state)
+    {
+        return continue_stable(op);
+    }
+    int r = split_stab_op(op, [this](obj_ver_id ov)
+    {
+        auto dirty_it = dirty_db.find(ov);
+        if (dirty_it == dirty_db.end())
+        {
+            auto & clean_db = clean_db_shard(ov.oid);
+            auto clean_it = clean_db.find(ov.oid);
+            if (clean_it == clean_db.end() || clean_it->second.version < ov.version)
+            {
+                // No such object version
+                printf("Error: %lx:%lx v%lu not found while stabilizing\n", ov.oid.inode, ov.oid.stripe, ov.version);
+                return -ENOENT;
+            }
+            else
+            {
+                // Already stable
+                return STAB_SPLIT_DONE;
+            }
+        }
+        else if (IS_IN_FLIGHT(dirty_it->second.state))
+        {
+            // Object write is still in progress. Wait until the write request completes
+            return STAB_SPLIT_WAIT;
+        }
+        else if (!IS_SYNCED(dirty_it->second.state))
+        {
+            // Object not synced yet - sync it
+            // In previous versions we returned EBUSY here and required
+            // the caller (OSD) to issue a global sync first. But a global sync
+            // waits for all writes in the queue including inflight writes. And
+            // inflight writes may themselves be blocked by unstable writes being
+            // still present in the journal and not flushed away from it.
+            // So we must sync specific objects here.
+            //
+            // Even more, we have to process "stabilize" request in parts. That is,
+            // we must stabilize all objects which are already synced. Otherwise
+            // they may block objects which are NOT synced yet.
+            return STAB_SPLIT_SYNC;
+        }
+        else if (IS_STABLE(dirty_it->second.state))
+        {
+            // Already stable
+            return STAB_SPLIT_DONE;
+        }
+        else
+        {
+            return STAB_SPLIT_TODO;
+        }
+    });
+    if (r != 1)
+    {
+        return r;
+    }
     // Check journal space
     blockstore_journal_check_t space_check(this);
-    if (!space_check.check_available(op, todo, sizeof(journal_entry_stable), 0))
+    if (!space_check.check_available(op, op->len, sizeof(journal_entry_stable), 0))
     {
         return 0;
     }
@@ -102,9 +351,9 @@ int blockstore_impl_t::dequeue_stable(blockstore_op_t *op)
     BS_SUBMIT_CHECK_SQES(space_check.sectors_to_write);
     // Prepare and submit journal entries
     int s = 0;
-    for (i = 0, v = (obj_ver_id*)op->buf; i < op->len; i++, v++)
+    auto v = (obj_ver_id*)op->buf;
+    for (int i = 0; i < op->len; i++, v++)
     {
-        // FIXME: Only stabilize versions that aren't stable yet
         if (!journal.entry_fits(sizeof(journal_entry_stable)) &&
             journal.sector_info[journal.cur_sector].dirty)
         {
@@ -197,6 +446,7 @@ void blockstore_impl_t::mark_stable(const obj_ver_id & v, bool forget_dirty)
                     {
                         inode_space_stats[dirty_it->first.oid.inode] += dsk.data_block_size;
                     }
+                    big_to_flush++;
                 }
                 else if (IS_DELETE(dirty_it->second.state))
                 {
@@ -205,6 +455,7 @@ void blockstore_impl_t::mark_stable(const obj_ver_id & v, bool forget_dirty)
                         sp -= dsk.data_block_size;
                     else
                         inode_space_stats.erase(dirty_it->first.oid.inode);
+                    big_to_flush++;
                 }
             }
             if (forget_dirty && (IS_BIG_WRITE(dirty_it->second.state) ||

src/blockstore_sync.cpp

@@ -12,7 +12,7 @@
 #define SYNC_JOURNAL_SYNC_SENT 7
 #define SYNC_DONE 8
 
-int blockstore_impl_t::continue_sync(blockstore_op_t *op, bool queue_has_in_progress_sync)
+int blockstore_impl_t::continue_sync(blockstore_op_t *op)
 {
     if (immediate_commit == IMMEDIATE_ALL)
     {
@@ -145,7 +145,7 @@ int blockstore_impl_t::continue_sync(blockstore_op_t *op, bool queue_has_in_prog
             PRIV(op)->op_state = SYNC_DONE;
         }
     }
-    if (PRIV(op)->op_state == SYNC_DONE && !queue_has_in_progress_sync)
+    if (PRIV(op)->op_state == SYNC_DONE)
     {
         ack_sync(op);
         return 2;

src/blockstore_write.cpp

@@ -89,6 +89,9 @@ bool blockstore_impl_t::enqueue_write(blockstore_op_t *op)
         else
         {
             // Invalid version requested
+#ifdef BLOCKSTORE_DEBUG
+            printf("Write %lx:%lx v%lu requested, but we already have v%lu\n", op->oid.inode, op->oid.stripe, op->version, version);
+#endif
             op->retval = -EEXIST;
             if (!is_del && dsk.clean_entry_bitmap_size > sizeof(void*))
             {
@@ -115,8 +118,8 @@ bool blockstore_impl_t::enqueue_write(blockstore_op_t *op)
     else if (!wait_del)
         printf("Write %lx:%lx v%lu offset=%u len=%u\n", op->oid.inode, op->oid.stripe, op->version, op->offset, op->len);
 #endif
-    // FIXME No strict need to add it into dirty_db here, it's just left
-    // from the previous implementation where reads waited for writes
+    // No strict need to add it into dirty_db here except maybe for listings to return
+    // correct data when there are inflight operations in the queue
     uint32_t state;
     if (is_del)
         state = BS_ST_DELETE | BS_ST_IN_FLIGHT;
@@ -182,9 +185,15 @@ void blockstore_impl_t::cancel_all_writes(blockstore_op_t *op, blockstore_dirty_
     bool found = false;
     for (auto other_op: submit_queue)
     {
-        // <op> may be present in queue multiple times due to moving operations in submit_queue
-        if (other_op == op)
+        if (!other_op)
+        {
+            // freed operations during submitting are zeroed
+        }
+        else if (other_op == op)
+        {
+            // <op> may be present in queue multiple times due to moving operations in submit_queue
             found = true;
+        }
         else if (found && other_op->oid == op->oid &&
             (other_op->opcode == BS_OP_WRITE || other_op->opcode == BS_OP_WRITE_STABLE))
         {
@@ -252,7 +261,8 @@ int blockstore_impl_t::dequeue_write(blockstore_op_t *op)
     {
         blockstore_journal_check_t space_check(this);
         if (!space_check.check_available(op, unsynced_big_write_count + 1,
-            sizeof(journal_entry_big_write) + dsk.clean_entry_bitmap_size, JOURNAL_STABILIZE_RESERVATION))
+            sizeof(journal_entry_big_write) + dsk.clean_entry_bitmap_size,
+            (dirty_it->second.state & BS_ST_INSTANT) ? JOURNAL_INSTANT_RESERVATION : JOURNAL_STABILIZE_RESERVATION))
         {
             return 0;
         }
@@ -261,6 +271,13 @@ int blockstore_impl_t::dequeue_write(blockstore_op_t *op)
         if (loc == UINT64_MAX)
         {
             // no space
+            if (big_to_flush > 0)
+            {
+                // hope that some space will be available after flush
+                flusher->request_trim();
+                PRIV(op)->wait_for = WAIT_FREE;
+                return 0;
+            }
             cancel_all_writes(op, dirty_it, -ENOSPC);
             return 2;
         }
@@ -332,7 +349,8 @@ int blockstore_impl_t::dequeue_write(blockstore_op_t *op)
             !space_check.check_available(op, unsynced_big_write_count,
                 sizeof(journal_entry_big_write) + dsk.clean_entry_bitmap_size, 0)
             || !space_check.check_available(op, 1,
-                sizeof(journal_entry_small_write) + dsk.clean_entry_bitmap_size, op->len + JOURNAL_STABILIZE_RESERVATION))
+                sizeof(journal_entry_small_write) + dsk.clean_entry_bitmap_size,
+                op->len + ((dirty_it->second.state & BS_ST_INSTANT) ? JOURNAL_INSTANT_RESERVATION : JOURNAL_STABILIZE_RESERVATION)))
         {
             return 0;
         }
@@ -443,18 +461,19 @@ int blockstore_impl_t::continue_write(blockstore_op_t *op)
 resume_2:
     // Only for the immediate_commit mode: prepare and submit big_write journal entry
     {
-        blockstore_journal_check_t space_check(this);
-        if (!space_check.check_available(op, 1,
-            sizeof(journal_entry_big_write) + dsk.clean_entry_bitmap_size, JOURNAL_STABILIZE_RESERVATION))
-        {
-            return 0;
-        }
-        BS_SUBMIT_CHECK_SQES(1);
         auto dirty_it = dirty_db.find((obj_ver_id){
             .oid = op->oid,
             .version = op->version,
         });
         assert(dirty_it != dirty_db.end());
+        blockstore_journal_check_t space_check(this);
+        if (!space_check.check_available(op, 1,
+            sizeof(journal_entry_big_write) + dsk.clean_entry_bitmap_size,
+            ((dirty_it->second.state & BS_ST_INSTANT) ? JOURNAL_INSTANT_RESERVATION : JOURNAL_STABILIZE_RESERVATION)))
+        {
+            return 0;
+        }
+        BS_SUBMIT_CHECK_SQES(1);
         journal_entry_big_write *je = (journal_entry_big_write*)prefill_single_journal_entry(
             journal, op->opcode == BS_OP_WRITE_STABLE ? JE_BIG_WRITE_INSTANT : JE_BIG_WRITE,
             sizeof(journal_entry_big_write) + dsk.clean_entry_bitmap_size
@@ -641,7 +660,7 @@ int blockstore_impl_t::dequeue_del(blockstore_op_t *op)
     });
     assert(dirty_it != dirty_db.end());
     blockstore_journal_check_t space_check(this);
-    if (!space_check.check_available(op, 1, sizeof(journal_entry_del), JOURNAL_STABILIZE_RESERVATION))
+    if (!space_check.check_available(op, 1, sizeof(journal_entry_del), JOURNAL_INSTANT_RESERVATION))
     {
         return 0;
     }

src/cli_df.cpp

@@ -121,8 +121,7 @@ resume_1:
             }
             if (pool_cfg.scheme != POOL_SCHEME_REPLICATED)
             {
-                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_avail *= (pool_cfg.pg_size - pool_cfg.parity_chunks);
             }
             pool_stats[pool_cfg.id] = json11::Json::object {
                 { "name", pool_cfg.name },

src/cli_merge.cpp

@@ -403,7 +403,7 @@ struct snap_merger_t
         op->opcode = OSD_OP_READ_BITMAP;
         op->inode = target;
         op->offset = offset;
-        op->len = 0;
+        op->len = target_block_size;
         op->callback = [this](cluster_op_t *op)
         {
             if (op->retval < 0)

src/cli_rm_data.cpp

@@ -92,6 +92,7 @@ struct rm_inode_t
 
     void send_ops(rm_pg_t *cur_list)
     {
+        parent->cli->init_msgr();
         if (parent->cli->msgr.osd_peer_fds.find(cur_list->rm_osd_num) ==
             parent->cli->msgr.osd_peer_fds.end())
         {

src/cli_rm_osd.cpp

@@ -5,6 +5,7 @@
 #include "cli.h"
 #include "cluster_client.h"
 #include "str_util.h"
+#include "epoll_manager.h"
 
 #include <algorithm>
 
@@ -14,13 +15,21 @@ struct rm_osd_t
     cli_tool_t *parent;
 
     bool dry_run, force_warning, force_dataloss;
+    uint64_t etcd_tx_retry_ms = 500;
+    uint64_t etcd_tx_retries = 10000;
     std::vector<uint64_t> osd_ids;
 
     int state = 0;
     cli_result_t result;
 
     std::set<uint64_t> to_remove;
+    std::set<uint64_t> to_restart;
     json11::Json::array pool_effects;
+    json11::Json::array history_updates, history_checks;
+    json11::Json new_pgs, new_clean_pgs;
+    uint64_t new_pgs_mod_rev, new_clean_pgs_mod_rev;
+    uint64_t cur_retry = 0;
+    uint64_t retry_wait = 0;
     bool is_warning, is_dataloss;
 
     bool is_done()
@@ -32,6 +41,12 @@ struct rm_osd_t
     {
         if (state == 1)
             goto resume_1;
+        else if (state == 2)
+            goto resume_2;
+        else if (state == 3)
+            goto resume_3;
+        else if (state == 4)
+            goto resume_4;
         if (!osd_ids.size())
         {
             result = (cli_result_t){ .err = EINVAL, .text = "OSD numbers are not specified" };
@@ -152,14 +167,48 @@ struct rm_osd_t
             result.text = error;
             if (dry_run || is_dataloss && !force_dataloss || is_warning && !force_warning)
             {
-                result.err = is_dataloss || is_warning ? EBUSY : 0;
+                result.err = is_dataloss && !force_dataloss || is_warning && !force_warning ? EBUSY : 0;
                 state = 100;
                 return;
             }
         }
+        parent->etcd_txn(json11::Json::object { { "success", json11::Json::array {
+            json11::Json::object {
+                { "request_range", json11::Json::object {
+                    { "key", base64_encode(
+                        parent->cli->st_cli.etcd_prefix+"/config/pgs"
+                    ) },
+                } },
+            },
+            json11::Json::object {
+                { "request_range", json11::Json::object {
+                    { "key", base64_encode(
+                        parent->cli->st_cli.etcd_prefix+"/history/last_clean_pgs"
+                    ) },
+                } },
+            },
+        } } });
+    resume_4:
+        state = 4;
+        if (parent->waiting > 0)
+            return;
+        if (parent->etcd_err.err)
+        {
+            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]);
+            new_pgs = remove_osds_from_pgs(kv);
+            new_pgs_mod_rev = kv.mod_revision;
+            kv = parent->cli->st_cli.parse_etcd_kv(parent->etcd_result["responses"][1]["response_range"]["kvs"][0]);
+            new_clean_pgs = remove_osds_from_pgs(kv);
+            new_clean_pgs_mod_rev = kv.mod_revision;
+        }
         // Remove keys from etcd
         {
-            json11::Json::array rm_items;
+            json11::Json::array rm_items, rm_checks;
             for (auto osd_id: osd_ids)
             {
                 rm_items.push_back("/config/osd/"+std::to_string(osd_id));
@@ -178,7 +227,39 @@ struct rm_osd_t
                     } },
                 };
             }
-            parent->etcd_txn(json11::Json::object { { "success", rm_items } });
+            if (!new_pgs.is_null())
+            {
+                auto pgs_key = base64_encode(parent->cli->st_cli.etcd_prefix+"/config/pgs");
+                rm_items.push_back(json11::Json::object {
+                    { "request_put", json11::Json::object {
+                        { "key", pgs_key },
+                        { "value", base64_encode(new_pgs.dump()) },
+                    } },
+                });
+                rm_checks.push_back(json11::Json::object {
+                    { "target", "MOD" },
+                    { "key", pgs_key },
+                    { "result", "LESS" },
+                    { "mod_revision", new_pgs_mod_rev+1 },
+                });
+            }
+            if (!new_clean_pgs.is_null())
+            {
+                auto pgs_key = base64_encode(parent->cli->st_cli.etcd_prefix+"/history/last_clean_pgs");
+                rm_items.push_back(json11::Json::object {
+                    { "request_put", json11::Json::object {
+                        { "key", pgs_key },
+                        { "value", base64_encode(new_clean_pgs.dump()) },
+                    } },
+                });
+                rm_checks.push_back(json11::Json::object {
+                    { "target", "MOD" },
+                    { "key", pgs_key },
+                    { "result", "LESS" },
+                    { "mod_revision", new_clean_pgs_mod_rev+1 },
+                });
+            }
+            parent->etcd_txn(json11::Json::object { { "success", rm_items }, { "checks", rm_checks } });
         }
     resume_1:
         state = 1;
@@ -190,6 +271,46 @@ struct rm_osd_t
             state = 100;
             return;
         }
+        // Remove old OSD from PG all_peers to prevent left_on_dead and from
+        // target_history to prevent INCOMPLETE if --allow-data-loss is specified
+        for (auto & rsp: parent->etcd_result["responses"].array_items())
+        {
+            if (rsp["response_delete_range"]["deleted"].uint64_value() > 0)
+            {
+                // Wait for mon_change_timeout before updating PG history, or the monitor's change will likely interfere with ours
+                retry_wait = parent->cli->config["mon_change_timeout"].uint64_value();
+                if (!retry_wait)
+                    retry_wait = 1000;
+                retry_wait += etcd_tx_retry_ms;
+            }
+        }
+        while (1)
+        {
+    resume_2:
+            if (!remove_osds_from_history(2))
+                return;
+    resume_3:
+            state = 3;
+            if (parent->waiting > 0)
+                return;
+            if (parent->etcd_err.err)
+            {
+                result = parent->etcd_err;
+                state = 100;
+                return;
+            }
+            if (parent->etcd_result["succeeded"].bool_value())
+                break;
+            if ((++cur_retry) >= etcd_tx_retries)
+            {
+                result.err = EAGAIN;
+                result.text += "Failed to remove OSDs from PG history due to update conflicts."
+                    " Some PGs may remain left_on_dead or incomplete. Please retry later\n";
+                state = 100;
+                return;
+            }
+            retry_wait = etcd_tx_retry_ms;
+        }
         std::string ids = "";
         for (auto osd_id: osd_ids)
         {
@@ -200,6 +321,138 @@ struct rm_osd_t
         result.text = (result.text != "" ? ids+"\n"+result.text : ids);
         result.err = 0;
     }
+
+    json11::Json remove_osds_from_pgs(const etcd_kv_t & kv)
+    {
+        if (kv.value.is_null())
+        {
+            return kv.value;
+        }
+        json11::Json::object new_pgs;
+        for (auto & pp: kv.value["items"].object_items())
+        {
+            if (pp.second.is_object())
+            {
+                json11::Json::object new_pool;
+                for (auto & pgp: pp.second.object_items())
+                {
+                    json11::Json::array osd_set;
+                    for (auto & osd_json: pgp.second["osd_set"].array_items())
+                    {
+                        uint64_t osd_num = osd_json.uint64_value();
+                        osd_set.push_back(osd_num == 0 || to_remove.find(osd_num) != to_remove.end() ? 0 : osd_num);
+                    }
+                    json11::Json::object new_pg = pgp.second.object_items();
+                    new_pg["osd_set"] = osd_set;
+                    new_pool[pgp.first] = new_pg;
+                }
+                new_pgs[pp.first] = new_pool;
+            }
+            else
+                new_pgs[pp.first] = pp.second;
+        }
+        auto res = kv.value.object_items();
+        res["items"] = new_pgs;
+        return res;
+    }
+
+    bool remove_osds_from_history(int base_state)
+    {
+        if (state == base_state+0)
+            goto resume_0;
+        history_updates.clear();
+        history_checks.clear();
+        for (auto & pp: parent->cli->st_cli.pool_config)
+        {
+            bool update_pg_history = false;
+            auto & pool_cfg = pp.second;
+            for (auto & pgp: pool_cfg.pg_config)
+            {
+                auto pg_num = pgp.first;
+                auto & pg_cfg = pgp.second;
+                for (int i = 0; i < pg_cfg.all_peers.size(); i++)
+                {
+                    if (to_remove.find(pg_cfg.all_peers[i]) != to_remove.end())
+                    {
+                        update_pg_history = true;
+                        pg_cfg.all_peers.erase(pg_cfg.all_peers.begin()+i, pg_cfg.all_peers.begin()+i+1);
+                        i--;
+                    }
+                }
+                for (int i = 0; i < pg_cfg.target_history.size(); i++)
+                {
+                    int hist_size = 0, hist_rm = 0;
+                    for (auto & old_osd: pg_cfg.target_history[i])
+                    {
+                        if (old_osd != 0)
+                        {
+                            hist_size++;
+                            if (to_remove.find(old_osd) != to_remove.end())
+                            {
+                                hist_rm++;
+                                old_osd = 0;
+                            }
+                        }
+                    }
+                    if (hist_rm > 0)
+                    {
+                        if (hist_size-hist_rm == 0)
+                        {
+                            pg_cfg.target_history.erase(pg_cfg.target_history.begin()+i, pg_cfg.target_history.begin()+i+1);
+                            i--;
+                        }
+                        update_pg_history = true;
+                    }
+                }
+                if (update_pg_history)
+                {
+                    std::string history_key = base64_encode(
+                        parent->cli->st_cli.etcd_prefix+"/pg/history/"+
+                        std::to_string(pool_cfg.id)+"/"+std::to_string(pg_num)
+                    );
+                    history_updates.push_back(json11::Json::object {
+                        { "request_put", json11::Json::object {
+                            { "key", history_key },
+                            { "value", base64_encode(json11::Json(json11::Json::object {
+                                { "epoch", pg_cfg.epoch },
+                                { "all_peers", pg_cfg.all_peers },
+                                { "osd_sets", pg_cfg.target_history },
+                            }).dump()) },
+                        } },
+                    });
+                    history_checks.push_back(json11::Json::object {
+                        { "target", "MOD" },
+                        { "key", history_key },
+                        { "result", "LESS" },
+                        { "mod_revision", parent->cli->st_cli.etcd_watch_revision+1 },
+                    });
+                }
+            }
+        }
+        if (history_updates.size())
+        {
+            if (retry_wait)
+            {
+                parent->waiting++;
+                parent->epmgr->tfd->set_timer(retry_wait, false, [this](int timer_id)
+                {
+                    parent->waiting--;
+                    parent->ringloop->wakeup();
+                });
+    resume_0:
+                state = base_state+0;
+                if (parent->waiting > 0)
+                    return false;
+            }
+            parent->etcd_txn(json11::Json::object {
+                { "success", history_updates },
+                { "compare", history_checks },
+            });
+        }
+        else
+            parent->etcd_result = json11::Json::object{ { "succeeded", true } };
+        return true;
+    }
 };
 
 std::function<bool(cli_result_t &)> cli_tool_t::start_rm_osd(json11::Json cfg)
@@ -209,6 +462,14 @@ std::function<bool(cli_result_t &)> cli_tool_t::start_rm_osd(json11::Json cfg)
     rm_osd->dry_run = cfg["dry_run"].bool_value();
     rm_osd->force_dataloss = cfg["allow_data_loss"].bool_value();
     rm_osd->force_warning = rm_osd->force_dataloss || cfg["force"].bool_value();
+    if (!cfg["etcd_tx_retries"].is_null())
+        rm_osd->etcd_tx_retries = cfg["etcd_tx_retries"].uint64_value();
+    if (!cfg["etcd_tx_retry_ms"].is_null())
+    {
+        rm_osd->etcd_tx_retry_ms = cfg["etcd_tx_retry_ms"].uint64_value();
+        if (rm_osd->etcd_tx_retry_ms < 100)
+            rm_osd->etcd_tx_retry_ms = 100;
+    }
     if (cfg["osd_id"].is_number() || cfg["osd_id"].is_string())
         rm_osd->osd_ids.push_back(cfg["osd_id"].uint64_value());
     else

src/cli_status.cpp

@@ -198,9 +198,9 @@ resume_2:
             }
             pgs_by_state_str += std::to_string(kv.second)+" "+kv.first;
         }
-        bool readonly = json_is_true(parent->cli->merged_config["readonly"]);
-        bool no_recovery = json_is_true(parent->cli->merged_config["no_recovery"]);
-        bool no_rebalance = json_is_true(parent->cli->merged_config["no_rebalance"]);
+        bool readonly = json_is_true(parent->cli->config["readonly"]);
+        bool no_recovery = json_is_true(parent->cli->config["no_recovery"]);
+        bool no_rebalance = json_is_true(parent->cli->config["no_rebalance"]);
         if (parent->json_output)
         {
             // JSON output

src/cluster_client.cpp

@@ -18,11 +18,12 @@
 
 cluster_client_t::cluster_client_t(ring_loop_t *ringloop, timerfd_manager_t *tfd, json11::Json & config)
 {
-    config = osd_messenger_t::read_config(config);
+    cli_config = config.object_items();
+    file_config = osd_messenger_t::read_config(config);
+    config = osd_messenger_t::merge_configs(cli_config, file_config, etcd_global_config, {});
 
     this->ringloop = ringloop;
     this->tfd = tfd;
-    this->config = config;
 
     msgr.osd_num = 0;
     msgr.tfd = tfd;
@@ -58,8 +59,7 @@ cluster_client_t::cluster_client_t(ring_loop_t *ringloop, timerfd_manager_t *tfd
         msgr.stop_client(op->peer_fd);
         delete op;
     };
-    msgr.parse_config(this->config);
-    msgr.init();
+    msgr.parse_config(config);
 
     st_cli.tfd = tfd;
     st_cli.on_load_config_hook = [this](json11::Json::object & cfg) { on_load_config_hook(cfg); };
@@ -73,17 +73,6 @@ cluster_client_t::cluster_client_t(ring_loop_t *ringloop, timerfd_manager_t *tfd
 
     scrap_buffer_size = SCRAP_BUFFER_SIZE;
     scrap_buffer = malloc_or_die(scrap_buffer_size);
-
-    if (ringloop)
-    {
-        consumer.loop = [this]()
-        {
-            msgr.read_requests();
-            msgr.send_replies();
-            this->ringloop->submit();
-        };
-        ringloop->register_consumer(&consumer);
-    }
 }
 
 cluster_client_t::~cluster_client_t()
@@ -115,6 +104,24 @@ cluster_op_t::~cluster_op_t()
     }
 }
 
+void cluster_client_t::init_msgr()
+{
+    if (msgr_initialized)
+        return;
+    msgr.init();
+    msgr_initialized = true;
+    if (ringloop)
+    {
+        consumer.loop = [this]()
+        {
+            msgr.read_requests();
+            msgr.send_replies();
+            this->ringloop->submit();
+        };
+        ringloop->register_consumer(&consumer);
+    }
+}
+
 void cluster_client_t::calc_wait(cluster_op_t *op)
 {
     op->prev_wait = 0;
@@ -143,7 +150,7 @@ void cluster_client_t::calc_wait(cluster_op_t *op)
         if (!op->prev_wait)
             continue_sync(op);
     }
-    else /* if (op->opcode == OSD_OP_READ || op->opcode == OSD_OP_READ_BITMAP) */
+    else /* if (op->opcode == OSD_OP_READ || op->opcode == OSD_OP_READ_BITMAP || op->opcode == OSD_OP_READ_CHAIN_BITMAP) */
     {
         for (auto prev = op_queue_head; prev && prev != op; prev = prev->next)
         {
@@ -151,7 +158,8 @@ void cluster_client_t::calc_wait(cluster_op_t *op)
             {
                 op->prev_wait++;
             }
-            else if (prev->opcode == OSD_OP_WRITE || prev->opcode == OSD_OP_READ || prev->opcode == OSD_OP_READ_BITMAP)
+            else if (prev->opcode == OSD_OP_WRITE || prev->opcode == OSD_OP_READ ||
+                prev->opcode == OSD_OP_READ_BITMAP || prev->opcode == OSD_OP_READ_CHAIN_BITMAP)
             {
                 // Flushes are always in the beginning (we're scanning from the beginning of the queue)
                 break;
@@ -171,7 +179,8 @@ void cluster_client_t::inc_wait(uint64_t opcode, uint64_t flags, cluster_op_t *n
             auto n2 = next->next;
             if (next->opcode == OSD_OP_SYNC && !(flags & OP_IMMEDIATE_COMMIT) ||
                 next->opcode == OSD_OP_WRITE && (flags & OP_FLUSH_BUFFER) && !(next->flags & OP_FLUSH_BUFFER) ||
-                (next->opcode == OSD_OP_READ || next->opcode == OSD_OP_READ_BITMAP) && (flags & OP_FLUSH_BUFFER))
+                (next->opcode == OSD_OP_READ || next->opcode == OSD_OP_READ_BITMAP ||
+                    next->opcode == OSD_OP_READ_CHAIN_BITMAP) && (flags & OP_FLUSH_BUFFER))
             {
                 next->prev_wait += inc;
                 assert(next->prev_wait >= 0);
@@ -221,11 +230,14 @@ void cluster_client_t::erase_op(cluster_op_t *op)
     if (op_queue_tail == op)
         op_queue_tail = op->prev;
     op->next = op->prev = NULL;
+    if (flags & OP_FLUSH_BUFFER)
+        std::function<void(cluster_op_t*)>(op->callback)(op);
     if (!(flags & OP_IMMEDIATE_COMMIT))
         inc_wait(opcode, flags, next, -1);
     // Call callback at the end to avoid inconsistencies in prev_wait
     // if the callback adds more operations itself
-    std::function<void(cluster_op_t*)>(op->callback)(op);
+    if (!(flags & OP_FLUSH_BUFFER))
+        std::function<void(cluster_op_t*)>(op->callback)(op);
 }
 
 void cluster_client_t::continue_ops(bool up_retry)
@@ -265,13 +277,10 @@ restart:
     continuing_ops = 0;
 }
 
-void cluster_client_t::on_load_config_hook(json11::Json::object & config)
+void cluster_client_t::on_load_config_hook(json11::Json::object & etcd_global_config)
 {
-    this->merged_config = config;
-    for (auto & kv: this->config.object_items())
-    {
-        this->merged_config[kv.first] = kv.second;
-    }
+    this->etcd_global_config = etcd_global_config;
+    config = osd_messenger_t::merge_configs(cli_config, file_config, etcd_global_config, {});
     if (config.find("client_max_dirty_bytes") != config.end())
     {
         client_max_dirty_bytes = config["client_max_dirty_bytes"].uint64_value();
@@ -281,14 +290,13 @@ void cluster_client_t::on_load_config_hook(json11::Json::object & config)
         // Old name
         client_max_dirty_bytes = config["client_dirty_limit"].uint64_value();
     }
-    if (config.find("client_max_dirty_ops") != config.end())
-    {
-        client_max_dirty_ops = config["client_max_dirty_ops"].uint64_value();
-    }
+    else
+        client_max_dirty_bytes = 0;
     if (!client_max_dirty_bytes)
     {
         client_max_dirty_bytes = DEFAULT_CLIENT_MAX_DIRTY_BYTES;
     }
+    client_max_dirty_ops = config["client_max_dirty_ops"].uint64_value();
     if (!client_max_dirty_ops)
     {
         client_max_dirty_ops = DEFAULT_CLIENT_MAX_DIRTY_OPS;
@@ -303,7 +311,7 @@ void cluster_client_t::on_load_config_hook(json11::Json::object & config)
         up_wait_retry_interval = 50;
     }
     msgr.parse_config(config);
-    msgr.parse_config(this->config);
+    st_cli.parse_config(config);
     st_cli.load_pgs();
 }
 
@@ -337,7 +345,8 @@ void cluster_client_t::on_change_hook(std::map<std::string, etcd_kv_t> & changes
             // And now they have to be resliced!
             for (auto op = op_queue_head; op; op = op->next)
             {
-                if ((op->opcode == OSD_OP_WRITE || op->opcode == OSD_OP_READ || op->opcode == OSD_OP_READ_BITMAP) &&
+                if ((op->opcode == OSD_OP_WRITE || op->opcode == OSD_OP_READ ||
+                    op->opcode == OSD_OP_READ_BITMAP || op->opcode == OSD_OP_READ_CHAIN_BITMAP) &&
                     INODE_POOL(op->cur_inode) == pool_item.first)
                 {
                     op->needs_reslice = true;
@@ -409,7 +418,7 @@ void cluster_client_t::on_ready(std::function<void(void)> fn)
 void cluster_client_t::execute(cluster_op_t *op)
 {
     if (op->opcode != OSD_OP_SYNC && op->opcode != OSD_OP_READ &&
-        op->opcode != OSD_OP_READ_BITMAP && op->opcode != OSD_OP_WRITE)
+        op->opcode != OSD_OP_READ_BITMAP && op->opcode != OSD_OP_READ_CHAIN_BITMAP && op->opcode != OSD_OP_WRITE)
     {
         op->retval = -EINVAL;
         std::function<void(cluster_op_t*)>(op->callback)(op);
@@ -441,7 +450,7 @@ void cluster_client_t::execute(cluster_op_t *op)
             return;
         }
         // Check alignment
-        if ((op->opcode == OSD_OP_READ || op->opcode == OSD_OP_WRITE) && !op->len ||
+        if (!op->len && (op->opcode == OSD_OP_READ || op->opcode == OSD_OP_READ_BITMAP || op->opcode == OSD_OP_READ_CHAIN_BITMAP || op->opcode == OSD_OP_WRITE) ||
             op->offset % pool_it->second.bitmap_granularity || op->len % pool_it->second.bitmap_granularity)
         {
             op->retval = -EINVAL;
@@ -702,8 +711,7 @@ resume_3:
         // Finished successfully
         // Even if the PG count has changed in meanwhile we treat it as success
         // because if some operations were invalid for the new PG count we'd get errors
-        bool is_read = op->opcode == OSD_OP_READ;
-        if (is_read)
+        if (op->opcode == OSD_OP_READ || op->opcode == OSD_OP_READ_CHAIN_BITMAP)
         {
             // Check parent inode
             auto ino_it = st_cli.inode_config.find(op->cur_inode);
@@ -727,6 +735,11 @@ resume_3:
             }
         }
         op->retval = op->len;
+        if (op->opcode == OSD_OP_READ_BITMAP || op->opcode == OSD_OP_READ_CHAIN_BITMAP)
+        {
+            auto & pool_cfg = st_cli.pool_config.at(INODE_POOL(op->inode));
+            op->retval = op->len / pool_cfg.bitmap_granularity;
+        }
         erase_op(op);
         return 1;
     }
@@ -750,7 +763,10 @@ resume_3:
         {
             for (int i = 0; i < op->parts.size(); i++)
             {
-                op->parts[i].flags = PART_RETRY;
+                if (!(op->parts[i].flags & PART_DONE))
+                {
+                    op->parts[i].flags = PART_RETRY;
+                }
             }
             goto resume_2;
         }
@@ -809,23 +825,19 @@ void cluster_client_t::slice_rw(cluster_op_t *op)
     uint64_t last_stripe = op->len > 0 ? ((op->offset + op->len - 1) / pg_block_size) * pg_block_size : first_stripe;
     op->retval = 0;
     op->parts.resize((last_stripe - first_stripe) / pg_block_size + 1);
-    if (op->opcode == OSD_OP_READ || op->opcode == OSD_OP_READ_BITMAP)
+    if (op->opcode == OSD_OP_READ || op->opcode == OSD_OP_READ_BITMAP || op->opcode == OSD_OP_READ_CHAIN_BITMAP)
     {
         // Allocate memory for the bitmap
-        unsigned object_bitmap_size = (((op->opcode == OSD_OP_READ_BITMAP ? pg_block_size : op->len) / pool_cfg.bitmap_granularity + 7) / 8);
+        unsigned object_bitmap_size = ((op->len / pool_cfg.bitmap_granularity + 7) / 8);
         object_bitmap_size = (object_bitmap_size < 8 ? 8 : object_bitmap_size);
         unsigned bitmap_mem = object_bitmap_size + (pool_cfg.data_block_size / pool_cfg.bitmap_granularity / 8 * pg_data_size) * op->parts.size();
-        if (op->bitmap_buf_size < bitmap_mem)
+        if (!op->bitmap_buf || op->bitmap_buf_size < bitmap_mem)
         {
             op->bitmap_buf = realloc_or_die(op->bitmap_buf, bitmap_mem);
-            if (!op->bitmap_buf_size)
-            {
-                // First allocation
-                memset(op->bitmap_buf, 0, object_bitmap_size);
-            }
             op->part_bitmaps = (uint8_t*)op->bitmap_buf + object_bitmap_size;
             op->bitmap_buf_size = bitmap_mem;
         }
+        memset(op->bitmap_buf, 0, bitmap_mem);
     }
     int iov_idx = 0;
     size_t iov_pos = 0;
@@ -876,13 +888,14 @@ void cluster_client_t::slice_rw(cluster_op_t *op)
             if (end == begin)
                 op->done_count++;
         }
-        else if (op->opcode != OSD_OP_READ_BITMAP && op->opcode != OSD_OP_DELETE)
+        else if (op->opcode != OSD_OP_READ_BITMAP && op->opcode != OSD_OP_READ_CHAIN_BITMAP && op->opcode != OSD_OP_DELETE)
         {
             add_iov(end-begin, false, op, iov_idx, iov_pos, op->parts[i].iov, NULL, 0);
         }
         op->parts[i].parent = op;
         op->parts[i].offset = begin;
-        op->parts[i].len = op->opcode == OSD_OP_READ_BITMAP || op->opcode == OSD_OP_DELETE ? 0 : (uint32_t)(end - begin);
+        op->parts[i].len = op->opcode == OSD_OP_READ_BITMAP || op->opcode == OSD_OP_READ_CHAIN_BITMAP ||
+            op->opcode == OSD_OP_DELETE ? 0 : (uint32_t)(end - begin);
         op->parts[i].pg_num = pg_num;
         op->parts[i].osd_num = 0;
         op->parts[i].flags = 0;
@@ -911,6 +924,10 @@ bool cluster_client_t::affects_osd(uint64_t inode, uint64_t offset, uint64_t len
 
 bool cluster_client_t::try_send(cluster_op_t *op, int i)
 {
+    if (!msgr_initialized)
+    {
+        init_msgr();
+    }
     auto part = &op->parts[i];
     auto & pool_cfg = st_cli.pool_config.at(INODE_POOL(op->cur_inode));
     auto pg_it = pool_cfg.pg_config.find(part->pg_num);
@@ -929,7 +946,7 @@ bool cluster_client_t::try_send(cluster_op_t *op, int i)
                 pool_cfg.scheme == POOL_SCHEME_REPLICATED ? 1 : pool_cfg.pg_size-pool_cfg.parity_chunks
             );
             uint64_t meta_rev = 0;
-            if (op->opcode != OSD_OP_READ_BITMAP && op->opcode != OSD_OP_DELETE)
+            if (op->opcode != OSD_OP_READ_BITMAP && op->opcode != OSD_OP_READ_CHAIN_BITMAP && op->opcode != OSD_OP_DELETE)
             {
                 auto ino_it = st_cli.inode_config.find(op->inode);
                 if (ino_it != st_cli.inode_config.end())
@@ -942,7 +959,7 @@ bool cluster_client_t::try_send(cluster_op_t *op, int i)
                     .header = {
                         .magic = SECONDARY_OSD_OP_MAGIC,
                         .id = next_op_id(),
-                        .opcode = op->opcode == OSD_OP_READ_BITMAP ? OSD_OP_READ : op->opcode,
+                        .opcode = op->opcode == OSD_OP_READ_BITMAP || op->opcode == OSD_OP_READ_CHAIN_BITMAP ? OSD_OP_READ : op->opcode,
                     },
                     .inode = op->cur_inode,
                     .offset = part->offset,
@@ -950,8 +967,10 @@ bool cluster_client_t::try_send(cluster_op_t *op, int i)
                     .meta_revision = meta_rev,
                     .version = op->opcode == OSD_OP_WRITE || op->opcode == OSD_OP_DELETE ? op->version : 0,
                 } },
-                .bitmap = (op->opcode == OSD_OP_READ || op->opcode == OSD_OP_READ_BITMAP ? (uint8_t*)op->part_bitmaps + pg_bitmap_size*i : NULL),
-                .bitmap_len = (unsigned)(op->opcode == OSD_OP_READ || op->opcode == OSD_OP_READ_BITMAP ? pg_bitmap_size : 0),
+                .bitmap = (op->opcode == OSD_OP_READ || op->opcode == OSD_OP_READ_BITMAP || op->opcode == OSD_OP_READ_CHAIN_BITMAP
+                    ? (uint8_t*)op->part_bitmaps + pg_bitmap_size*i : NULL),
+                .bitmap_len = (unsigned)(op->opcode == OSD_OP_READ || op->opcode == OSD_OP_READ_BITMAP || op->opcode == OSD_OP_READ_CHAIN_BITMAP
+                    ? pg_bitmap_size : 0),
                 .callback = [this, part](osd_op_t *op_part)
                 {
                     handle_op_part(part);
@@ -1099,6 +1118,24 @@ void cluster_client_t::handle_op_part(cluster_op_part_t *part)
     if (part->op.reply.hdr.retval != expected)
     {
         // Operation failed, retry
+        part->flags |= PART_ERROR;
+        if (!op->retval || op->retval == -EPIPE)
+        {
+            // Don't overwrite other errors with -EPIPE
+            op->retval = part->op.reply.hdr.retval;
+        }
+        int stop_fd = -1;
+        if (op->retval != -EINTR && op->retval != -EIO)
+        {
+            stop_fd = part->op.peer_fd;
+            fprintf(
+                stderr, "%s operation failed on OSD %lu: retval=%ld (expected %d), dropping connection\n",
+                osd_op_names[part->op.req.hdr.opcode], part->osd_num, part->op.reply.hdr.retval, expected
+            );
+        }
+        // All next things like timer, continue_sync/rw and stop_client may affect the operation again
+        // So do all these things after modifying operation state, otherwise we may hit reenterability bugs
+        // FIXME postpone such things to set_immediate here to avoid bugs
         if (part->op.reply.hdr.retval == -EPIPE)
         {
             // Mark op->up_wait = true before stopping the client
@@ -1112,40 +1149,37 @@ void cluster_client_t::handle_op_part(cluster_op_part_t *part)
                 });
             }
         }
-        if (!op->retval || op->retval == -EPIPE)
+        if (op->inflight_count == 0)
         {
-            // Don't overwrite other errors with -EPIPE
-            op->retval = part->op.reply.hdr.retval;
+            if (op->opcode == OSD_OP_SYNC)
+                continue_sync(op);
+            else
+                continue_rw(op);
         }
-        if (op->retval != -EINTR && op->retval != -EIO)
+        if (stop_fd >= 0)
         {
-            fprintf(
-                stderr, "%s operation failed on OSD %lu: retval=%ld (expected %d), dropping connection\n",
-                osd_op_names[part->op.req.hdr.opcode], part->osd_num, part->op.reply.hdr.retval, expected
-            );
-            msgr.stop_client(part->op.peer_fd);
+            msgr.stop_client(stop_fd);
         }
-        part->flags |= PART_ERROR;
     }
     else
     {
         // OK
-        if (!(op->flags & OP_IMMEDIATE_COMMIT))
+        if ((op->opcode == OSD_OP_WRITE || op->opcode == OSD_OP_DELETE) && !(op->flags & OP_IMMEDIATE_COMMIT))
             dirty_osds.insert(part->osd_num);
         part->flags |= PART_DONE;
         op->done_count++;
-        if (op->opcode == OSD_OP_READ || op->opcode == OSD_OP_READ_BITMAP)
+        if (op->opcode == OSD_OP_READ || op->opcode == OSD_OP_READ_BITMAP || op->opcode == OSD_OP_READ_CHAIN_BITMAP)
         {
             copy_part_bitmap(op, part);
             op->version = op->parts.size() == 1 ? part->op.reply.rw.version : 0;
         }
-    }
-    if (op->inflight_count == 0)
-    {
-        if (op->opcode == OSD_OP_SYNC)
-            continue_sync(op);
-        else
-            continue_rw(op);
+        if (op->inflight_count == 0)
+        {
+            if (op->opcode == OSD_OP_SYNC)
+                continue_sync(op);
+            else
+                continue_rw(op);
+        }
     }
 }
 
@@ -1158,7 +1192,12 @@ void cluster_client_t::copy_part_bitmap(cluster_op_t *op, cluster_op_part_t *par
     );
     uint32_t object_offset = (part->op.req.rw.offset - op->offset) / pool_cfg.bitmap_granularity;
     uint32_t part_offset = (part->op.req.rw.offset % pg_block_size) / pool_cfg.bitmap_granularity;
-    uint32_t part_len = (op->opcode == OSD_OP_READ_BITMAP ? pg_block_size : part->op.req.rw.len) / pool_cfg.bitmap_granularity;
+    uint32_t op_len = op->len / pool_cfg.bitmap_granularity;
+    uint32_t part_len = pg_block_size/pool_cfg.bitmap_granularity - part_offset;
+    if (part_len > op_len-object_offset)
+    {
+        part_len = op_len-object_offset;
+    }
     if (!(object_offset & 0x7) && !(part_offset & 0x7) && (part_len >= 8))
     {
         // Copy bytes

src/cluster_client.h

@@ -11,6 +11,7 @@
 #define INODE_LIST_DONE 1
 #define INODE_LIST_HAS_UNSTABLE 2
 #define OSD_OP_READ_BITMAP OSD_OP_SEC_READ_BMP
+#define OSD_OP_READ_CHAIN_BITMAP 0x102
 
 #define OSD_OP_IGNORE_READONLY 0x08
 
@@ -30,7 +31,7 @@ struct cluster_op_part_t
 
 struct cluster_op_t
 {
-    uint64_t opcode; // OSD_OP_READ, OSD_OP_WRITE, OSD_OP_SYNC, OSD_OP_DELETE, OSD_OP_READ_BITMAP
+    uint64_t opcode; // OSD_OP_READ, OSD_OP_WRITE, OSD_OP_SYNC, OSD_OP_DELETE, OSD_OP_READ_BITMAP, OSD_OP_READ_CHAIN_BITMAP
     uint64_t inode;
     uint64_t offset;
     uint64_t len;
@@ -39,9 +40,13 @@ struct cluster_op_t
     uint64_t version = 0;
     // now only OSD_OP_IGNORE_READONLY is supported
     uint64_t flags = 0;
+    // negative retval is an error number
+    // write and read return len on success
+    // sync and delete return 0 on success
+    // read_bitmap and read_chain_bitmap return the length of bitmap in bits(!)
     int retval;
     osd_op_buf_list_t iov;
-    // READ and READ_BITMAP return the bitmap here
+    // READ, READ_BITMAP, READ_CHAIN_BITMAP return the bitmap here
     void *bitmap_buf = NULL;
     std::function<void(cluster_op_t*)> callback;
     ~cluster_op_t();
@@ -99,12 +104,16 @@ class cluster_client_t
     std::vector<std::function<void(void)>> on_ready_hooks;
     std::vector<inode_list_t*> lists;
     int continuing_ops = 0;
+    bool msgr_initialized = false;
 
 public:
     etcd_state_client_t st_cli;
+
     osd_messenger_t msgr;
-    json11::Json config;
-    json11::Json::object merged_config;
+    void init_msgr();
+
+    json11::Json::object cli_config, file_config, etcd_global_config;
+    json11::Json::object config;
 
     cluster_client_t(ring_loop_t *ringloop, timerfd_manager_t *tfd, json11::Json & config);
     ~cluster_client_t();

src/cluster_client_list.cpp

@@ -43,6 +43,7 @@ struct inode_list_t
 inode_list_t* cluster_client_t::list_inode_start(inode_t inode,
     std::function<void(inode_list_t* lst, std::set<object_id>&& objects, pg_num_t pg_num, osd_num_t primary_osd, int status)> callback)
 {
+    init_msgr();
     int skipped_pgs = 0;
     pool_id_t pool_id = INODE_POOL(inode);
     if (!pool_id || st_cli.pool_config.find(pool_id) == st_cli.pool_config.end())

src/disk_tool_journal.cpp

@@ -281,7 +281,7 @@ void disk_tool_t::dump_journal_entry(int num, journal_entry *je, bool json)
         if (je->big_write.size > sizeof(journal_entry_big_write))
         {
             printf(json ? ",\"bitmap\":\"" : " (bitmap: ");
-            for (int i = sizeof(journal_entry_big_write); i < je->small_write.size; i++)
+            for (int i = sizeof(journal_entry_big_write); i < je->big_write.size; i++)
             {
                 printf("%02x", ((uint8_t*)je)[i]);
             }

src/disk_tool_meta.cpp

@@ -26,7 +26,7 @@ int disk_tool_t::process_meta(std::function<void(blockstore_meta_header_v1_t *)>
         buf_size = dsk.meta_len;
     void *data = memalign_or_die(MEM_ALIGNMENT, buf_size);
     lseek64(dsk.meta_fd, dsk.meta_offset, 0);
-    read_blocking(dsk.meta_fd, data, buf_size);
+    read_blocking(dsk.meta_fd, data, dsk.meta_block_size);
     // Check superblock
     blockstore_meta_header_v1_t *hdr = (blockstore_meta_header_v1_t *)data;
     if (hdr->zero == 0 &&
@@ -41,8 +41,11 @@ int disk_tool_t::process_meta(std::function<void(blockstore_meta_header_v1_t *)>
             if (buf_size % dsk.meta_block_size)
             {
                 buf_size = 8*dsk.meta_block_size;
+                void *new_data = memalign_or_die(MEM_ALIGNMENT, buf_size);
+                memcpy(new_data, data, dsk.meta_block_size);
                 free(data);
-                data = memalign_or_die(MEM_ALIGNMENT, buf_size);
+                data = new_data;
+                hdr = (blockstore_meta_header_v1_t *)data;
             }
         }
         dsk.bitmap_granularity = hdr->bitmap_granularity;

src/disk_tool_udev.cpp

@@ -387,6 +387,14 @@ int disk_tool_t::purge_devices(const std::vector<std::string> & devices)
         rm_osd_cli.push_back(std::to_string(osd_num));
     }
     // Check for data loss
+    if (options["force"] != "")
+    {
+        rm_osd_cli.push_back("--force");
+    }
+    else if (options["allow_data_loss"] != "")
+    {
+        rm_osd_cli.push_back("--allow-data-loss");
+    }
     rm_osd_cli.push_back("--dry-run");
     std::string dry_run_ignore_stdout;
     if (shell_exec(rm_osd_cli, "", &dry_run_ignore_stdout, NULL) != 0)
@@ -405,14 +413,6 @@ int disk_tool_t::purge_devices(const std::vector<std::string> & devices)
     }
     // Remove OSD metadata
     rm_osd_cli.pop_back();
-    if (options["force"] != "")
-    {
-        rm_osd_cli.push_back("--force");
-    }
-    else if (options["allow_data_loss"] != "")
-    {
-        rm_osd_cli.push_back("--allow-data-loss");
-    }
     if (shell_exec(rm_osd_cli, "", NULL, NULL) != 0)
     {
         return 1;

src/disk_tool_utils.cpp

@@ -305,10 +305,10 @@ int write_zero(int fd, uint64_t offset, uint64_t size)
 json11::Json read_parttable(std::string dev)
 {
     std::string part_dump;
-    int r = shell_exec({ "sfdisk", "--dump", dev, "--json" }, "", &part_dump, NULL);
+    int r = shell_exec({ "sfdisk", "--json", dev }, "", &part_dump, NULL);
     if (r == 255)
     {
-        fprintf(stderr, "Error running sfdisk --dump %s --json\n", dev.c_str());
+        fprintf(stderr, "Error running sfdisk --json %s\n", dev.c_str());
         return json11::Json(false);
     }
     // Decode partition table
@@ -319,7 +319,7 @@ json11::Json read_parttable(std::string dev)
         pt = json11::Json::parse(part_dump, err);
         if (err != "")
         {
-            fprintf(stderr, "sfdisk --dump %s --json returned bad JSON: %s\n", dev.c_str(), part_dump.c_str());
+            fprintf(stderr, "sfdisk --json %s returned bad JSON: %s\n", dev.c_str(), part_dump.c_str());
             return json11::Json(false);
         }
         pt = pt["partitiontable"];

src/epoll_manager.cpp

@@ -54,6 +54,13 @@ void epoll_manager_t::set_fd_handler(int fd, bool wr, std::function<void(int, in
         ev.events = (wr ? EPOLLOUT : 0) | EPOLLIN | EPOLLRDHUP | EPOLLET;
         if (epoll_ctl(epoll_fd, exists ? EPOLL_CTL_MOD : EPOLL_CTL_ADD, fd, &ev) < 0)
         {
+            if (errno == ENOENT)
+            {
+                // The FD is probably already closed
+                epoll_ctl(epoll_fd, EPOLL_CTL_DEL, fd, NULL);
+                epoll_handlers.erase(fd);
+                return;
+            }
             throw std::runtime_error(std::string("epoll_ctl: ") + strerror(errno));
         }
         epoll_handlers[fd] = handler;

src/etcd_state_client.cpp

@@ -18,12 +18,8 @@ etcd_state_client_t::~etcd_state_client_t()
     }
     watches.clear();
     etcd_watches_initialised = -1;
-    if (ws_keepalive_timer >= 0)
-    {
-        tfd->clear_timer(ws_keepalive_timer);
-        ws_keepalive_timer = -1;
-    }
 #ifndef __MOCK__
+    stop_ws_keepalive();
     if (etcd_watch_ws)
     {
         http_close(etcd_watch_ws);
@@ -245,6 +241,7 @@ void etcd_state_client_t::parse_config(const json11::Json & config)
         if (this->etcd_keepalive_timeout < 30)
             this->etcd_keepalive_timeout = 30;
     }
+    auto old_etcd_ws_keepalive_interval = this->etcd_ws_keepalive_interval;
     this->etcd_ws_keepalive_interval = config["etcd_ws_keepalive_interval"].uint64_value();
     if (this->etcd_ws_keepalive_interval <= 0)
     {
@@ -265,6 +262,13 @@ void etcd_state_client_t::parse_config(const json11::Json & config)
     {
         this->etcd_quick_timeout = 1000;
     }
+    if (this->etcd_ws_keepalive_interval != old_etcd_ws_keepalive_interval && ws_keepalive_timer >= 0)
+    {
+#ifndef __MOCK__
+        stop_ws_keepalive();
+        start_ws_keepalive();
+#endif
+    }
 }
 
 void etcd_state_client_t::pick_next_etcd()
@@ -478,6 +482,20 @@ void etcd_state_client_t::start_etcd_watcher()
     {
         on_start_watcher_hook(etcd_watch_ws);
     }
+    start_ws_keepalive();
+}
+
+void etcd_state_client_t::stop_ws_keepalive()
+{
+    if (ws_keepalive_timer >= 0)
+    {
+        tfd->clear_timer(ws_keepalive_timer);
+        ws_keepalive_timer = -1;
+    }
+}
+
+void etcd_state_client_t::start_ws_keepalive()
+{
     if (ws_keepalive_timer < 0)
     {
         ws_keepalive_timer = tfd->set_timer(etcd_ws_keepalive_interval*1000, true, [this](int)
@@ -871,19 +889,33 @@ void etcd_state_client_t::parse_state(const etcd_kv_t & kv)
             pg_cfg.target_history.clear();
             pg_cfg.all_peers.clear();
             // Refuse to start PG if any set of the <osd_sets> has no live OSDs
-            for (auto hist_item: value["osd_sets"].array_items())
+            for (auto & hist_item: value["osd_sets"].array_items())
             {
                 std::vector<osd_num_t> history_set;
-                for (auto pg_osd: hist_item.array_items())
+                for (auto & pg_osd: hist_item.array_items())
                 {
-                    history_set.push_back(pg_osd.uint64_value());
+                    osd_num_t pg_osd_num = pg_osd.uint64_value();
+                    if (pg_osd_num != 0)
+                    {
+                        auto it = std::lower_bound(history_set.begin(), history_set.end(), pg_osd_num);
+                        if (it == history_set.end() || *it != pg_osd_num)
+                            history_set.insert(it, pg_osd_num);
+                    }
                 }
-                pg_cfg.target_history.push_back(history_set);
+                auto it = std::lower_bound(pg_cfg.target_history.begin(), pg_cfg.target_history.end(), history_set);
+                if (it == pg_cfg.target_history.end() || *it != history_set)
+                    pg_cfg.target_history.insert(it, history_set);
             }
             // Include these additional OSDs when peering the PG
             for (auto pg_osd: value["all_peers"].array_items())
             {
-                pg_cfg.all_peers.push_back(pg_osd.uint64_value());
+                osd_num_t pg_osd_num = pg_osd.uint64_value();
+                if (pg_osd_num != 0)
+                {
+                    auto it = std::lower_bound(pg_cfg.all_peers.begin(), pg_cfg.all_peers.end(), pg_osd_num);
+                    if (it == pg_cfg.all_peers.end() || *it != pg_osd_num)
+                        pg_cfg.all_peers.insert(it, pg_osd_num);
+                }
             }
             // Read epoch
             pg_cfg.epoch = value["epoch"].uint64_value();

src/etcd_state_client.h

@@ -132,6 +132,8 @@ public:
     void etcd_txn(json11::Json txn, int timeout, int retries, int interval, std::function<void(std::string, json11::Json)> callback);
     void etcd_txn_slow(json11::Json txn, std::function<void(std::string, json11::Json)> callback);
     void start_etcd_watcher();
+    void stop_ws_keepalive();
+    void start_ws_keepalive();
     void load_global_config();
     void load_pgs();
     void parse_state(const etcd_kv_t & kv);

src/messenger.cpp

@@ -157,10 +157,10 @@ void osd_messenger_t::parse_config(const json11::Json & config)
         this->rdma_max_sge = 128;
     this->rdma_max_send = config["rdma_max_send"].uint64_value();
     if (!this->rdma_max_send)
-        this->rdma_max_send = 1;
+        this->rdma_max_send = 8;
     this->rdma_max_recv = config["rdma_max_recv"].uint64_value();
     if (!this->rdma_max_recv)
-        this->rdma_max_recv = 128;
+        this->rdma_max_recv = 16;
     this->rdma_max_msg = config["rdma_max_msg"].uint64_value();
     if (!this->rdma_max_msg || this->rdma_max_msg > 128*1024*1024)
         this->rdma_max_msg = 129*1024;
@@ -534,8 +534,9 @@ bool osd_messenger_t::is_rdma_enabled()
 }
 #endif
 
-json11::Json osd_messenger_t::read_config(const json11::Json & config)
+json11::Json::object osd_messenger_t::read_config(const json11::Json & config)
 {
+    json11::Json::object file_config;
     const char *config_path = config["config_path"].string_value() != ""
         ? config["config_path"].string_value().c_str() : VITASTOR_CONFIG_PATH;
     int fd = open(config_path, O_RDONLY);
@@ -543,14 +544,14 @@ json11::Json osd_messenger_t::read_config(const json11::Json & config)
     {
         if (errno != ENOENT)
             fprintf(stderr, "Error reading %s: %s\n", config_path, strerror(errno));
-        return config;
+        return file_config;
     }
     struct stat st;
     if (fstat(fd, &st) != 0)
     {
         fprintf(stderr, "Error reading %s: %s\n", config_path, strerror(errno));
         close(fd);
-        return config;
+        return file_config;
     }
     std::string buf;
     buf.resize(st.st_size);
@@ -562,23 +563,125 @@ json11::Json osd_messenger_t::read_config(const json11::Json & config)
         {
             fprintf(stderr, "Error reading %s: %s\n", config_path, strerror(errno));
             close(fd);
-            return config;
+            return file_config;
         }
         done += r;
     }
     close(fd);
     std::string json_err;
-    json11::Json::object file_config = json11::Json::parse(buf, json_err).object_items();
+    file_config = json11::Json::parse(buf, json_err).object_items();
     if (json_err != "")
     {
         fprintf(stderr, "Invalid JSON in %s: %s\n", config_path, json_err.c_str());
-        return config;
-    }
-    file_config.erase("config_path");
-    file_config.erase("osd_num");
-    for (auto kv: config.object_items())
-    {
-        file_config[kv.first] = kv.second;
     }
     return file_config;
 }
+
+static const char* cli_only_params[] = {
+    // The list has to be sorted
+    "bitmap_granularity",
+    "block_size",
+    "data_device",
+    "data_offset",
+    "data_size",
+    "disable_data_fsync",
+    "disable_device_lock",
+    "disable_journal_fsync",
+    "disable_meta_fsync",
+    "disk_alignment",
+    "flush_journal",
+    "immediate_commit",
+    "inmemory_journal",
+    "inmemory_metadata",
+    "journal_block_size",
+    "journal_device",
+    "journal_no_same_sector_overwrites",
+    "journal_offset",
+    "journal_sector_buffer_count",
+    "journal_size",
+    "meta_block_size",
+    "meta_buf_size",
+    "meta_device",
+    "meta_offset",
+    "osd_num",
+    "readonly",
+};
+
+static const char **cli_only_end = cli_only_params + (sizeof(cli_only_params)/sizeof(cli_only_params[0]));
+
+static const char* local_only_params[] = {
+    // The list has to be sorted
+    "config_path",
+    "rdma_device",
+    "rdma_gid_index",
+    "rdma_max_msg",
+    "rdma_max_recv",
+    "rdma_max_send",
+    "rdma_max_sge",
+    "rdma_mtu",
+    "rdma_port_num",
+    "tcp_header_buffer_size",
+    "use_rdma",
+    "use_sync_send_recv",
+};
+
+static const char **local_only_end = local_only_params + (sizeof(local_only_params)/sizeof(local_only_params[0]));
+
+// Basically could be replaced by std::lower_bound()...
+static int find_str_array(const char **start, const char **end, const std::string & s)
+{
+    int min = 0, max = end-start;
+    while (max-min >= 2)
+    {
+        int mid = (min+max)/2;
+        int r = strcmp(s.c_str(), start[mid]);
+        if (r < 0)
+            max = mid;
+        else if (r > 0)
+            min = mid;
+        else
+            return mid;
+    }
+    if (min < end-start && !strcmp(s.c_str(), start[min]))
+        return min;
+    return -1;
+}
+
+json11::Json::object osd_messenger_t::merge_configs(const json11::Json::object & cli_config,
+    const json11::Json::object & file_config,
+    const json11::Json::object & etcd_global_config,
+    const json11::Json::object & etcd_osd_config)
+{
+    // Priority: most important -> less important:
+    // etcd_osd_config -> cli_config -> etcd_global_config -> file_config
+    json11::Json::object res = file_config;
+    for (auto & kv: file_config)
+    {
+        int cli_only = find_str_array(cli_only_params, cli_only_end, kv.first);
+        if (cli_only < 0)
+        {
+            res[kv.first] = kv.second;
+        }
+    }
+    for (auto & kv: etcd_global_config)
+    {
+        int local_only = find_str_array(local_only_params, local_only_end, kv.first);
+        if (local_only < 0)
+        {
+            res[kv.first] = kv.second;
+        }
+    }
+    for (auto & kv: cli_config)
+    {
+        res[kv.first] = kv.second;
+    }
+    for (auto & kv: etcd_osd_config)
+    {
+        int local_only = find_str_array(local_only_params, local_only_end, kv.first);
+        if (local_only < 0)
+        {
+            res[kv.first] = kv.second;
+        }
+    }
+    return res;
+}

src/messenger.h

@@ -138,6 +138,7 @@ protected:
 
     std::vector<int> read_ready_clients;
     std::vector<int> write_ready_clients;
+    // We don't use ringloop->set_immediate here because we may have no ringloop in client :)
     std::vector<std::function<void()>> set_immediate;
 
 public:
@@ -165,7 +166,11 @@ public:
     void accept_connections(int listen_fd);
     ~osd_messenger_t();
 
-    static json11::Json read_config(const json11::Json & config);
+    static json11::Json::object read_config(const json11::Json & config);
+    static json11::Json::object merge_configs(const json11::Json::object & cli_config,
+        const json11::Json::object & file_config,
+        const json11::Json::object & etcd_global_config,
+        const json11::Json::object & etcd_osd_config);
 
 #ifdef WITH_RDMA
     bool is_rdma_enabled();

src/mock/messenger.cpp

@@ -43,7 +43,15 @@ void osd_messenger_t::send_replies()
 {
 }
 
-json11::Json osd_messenger_t::read_config(const json11::Json & config)
+json11::Json::object osd_messenger_t::read_config(const json11::Json & config)
 {
-    return config;
+    return json11::Json::object();
+}
+
+json11::Json::object osd_messenger_t::merge_configs(const json11::Json::object & cli_config,
+    const json11::Json::object & file_config,
+    const json11::Json::object & etcd_global_config,
+    const json11::Json::object & etcd_osd_config)
+{
+    return cli_config;
 }

src/msgr_rdma.cpp

@@ -368,9 +368,8 @@ static void try_send_rdma_wr(osd_client_t *cl, ibv_sge *sge, int op_sge)
 bool osd_messenger_t::try_send_rdma(osd_client_t *cl)
 {
     auto rc = cl->rdma_conn;
-    if (!cl->send_list.size() || rc->cur_send > 0)
+    if (!cl->send_list.size() || rc->cur_send >= rc->max_send)
     {
-        // Only send one batch at a time
         return true;
     }
     uint64_t op_size = 0, op_sge = 0;
@@ -380,6 +379,7 @@ bool osd_messenger_t::try_send_rdma(osd_client_t *cl)
         iovec & iov = cl->send_list[rc->send_pos];
         if (op_size >= rc->max_msg || op_sge >= rc->max_sge)
         {
+            rc->send_sizes.push_back(op_size);
             try_send_rdma_wr(cl, sge, op_sge);
             op_sge = 0;
             op_size = 0;
@@ -405,18 +405,24 @@ bool osd_messenger_t::try_send_rdma(osd_client_t *cl)
     }
     if (op_sge > 0)
     {
+        rc->send_sizes.push_back(op_size);
         try_send_rdma_wr(cl, sge, op_sge);
     }
     return true;
 }
 
-static void try_recv_rdma_wr(osd_client_t *cl, ibv_sge *sge, int op_sge)
+static void try_recv_rdma_wr(osd_client_t *cl, void *buf)
 {
+    ibv_sge sge = {
+        .addr = (uintptr_t)buf,
+        .length = (uint32_t)cl->rdma_conn->max_msg,
+        .lkey = cl->rdma_conn->ctx->mr->lkey,
+    };
     ibv_recv_wr *bad_wr = NULL;
     ibv_recv_wr wr = {
         .wr_id = (uint64_t)(cl->peer_fd*2),
-        .sg_list = sge,
-        .num_sge = op_sge,
+        .sg_list = &sge,
+        .num_sge = 1,
     };
     int err = ibv_post_recv(cl->rdma_conn->qp, &wr, &bad_wr);
     if (err || bad_wr)
@@ -434,12 +440,7 @@ bool osd_messenger_t::try_recv_rdma(osd_client_t *cl)
     {
         void *buf = malloc_or_die(rc->max_msg);
         rc->recv_buffers.push_back(buf);
-        ibv_sge sge = {
-            .addr = (uintptr_t)buf,
-            .length = (uint32_t)rc->max_msg,
-            .lkey = rc->ctx->mr->lkey,
-        };
-        try_recv_rdma_wr(cl, &sge, 1);
+        try_recv_rdma_wr(cl, buf);
     }
     return true;
 }
@@ -476,6 +477,7 @@ void osd_messenger_t::handle_rdma_events()
                 continue;
             }
             osd_client_t *cl = cl_it->second;
+            auto rc = cl->rdma_conn;
             if (wc[i].status != IBV_WC_SUCCESS)
             {
                 fprintf(stderr, "RDMA work request failed for client %d", client_id);
@@ -489,44 +491,59 @@ void osd_messenger_t::handle_rdma_events()
             }
             if (!is_send)
             {
-                cl->rdma_conn->cur_recv--;
-                if (!handle_read_buffer(cl, cl->rdma_conn->recv_buffers[0], wc[i].byte_len))
+                rc->cur_recv--;
+                if (!handle_read_buffer(cl, rc->recv_buffers[rc->next_recv_buf], wc[i].byte_len))
                 {
                     // handle_read_buffer may stop the client
                     continue;
                 }
-                free(cl->rdma_conn->recv_buffers[0]);
-                cl->rdma_conn->recv_buffers.erase(cl->rdma_conn->recv_buffers.begin(), cl->rdma_conn->recv_buffers.begin()+1);
-                try_recv_rdma(cl);
+                try_recv_rdma_wr(cl, rc->recv_buffers[rc->next_recv_buf]);
+                rc->next_recv_buf = (rc->next_recv_buf+1) % rc->recv_buffers.size();
             }
             else
             {
-                cl->rdma_conn->cur_send--;
-                if (!cl->rdma_conn->cur_send)
+                rc->cur_send--;
+                uint64_t sent_size = rc->send_sizes.at(0);
+                rc->send_sizes.erase(rc->send_sizes.begin(), rc->send_sizes.begin()+1);
+                int send_pos = 0, send_buf_pos = 0;
+                while (sent_size > 0)
                 {
-                    // Wait for the whole batch
-                    for (int i = 0; i < cl->rdma_conn->send_pos; i++)
+                    if (sent_size >= cl->send_list.at(send_pos).iov_len)
                     {
-                        if (cl->outbox[i].flags & MSGR_SENDP_FREE)
-                        {
-                            // Reply fully sent
-                            delete cl->outbox[i].op;
-                        }
+                        sent_size -= cl->send_list[send_pos].iov_len;
+                        send_pos++;
                     }
-                    if (cl->rdma_conn->send_pos > 0)
+                    else
                     {
-                        cl->send_list.erase(cl->send_list.begin(), cl->send_list.begin()+cl->rdma_conn->send_pos);
-                        cl->outbox.erase(cl->outbox.begin(), cl->outbox.begin()+cl->rdma_conn->send_pos);
-                        cl->rdma_conn->send_pos = 0;
+                        send_buf_pos = sent_size;
+                        sent_size = 0;
                     }
-                    if (cl->rdma_conn->send_buf_pos > 0)
-                    {
-                        cl->send_list[0].iov_base = (uint8_t*)cl->send_list[0].iov_base + cl->rdma_conn->send_buf_pos;
-                        cl->send_list[0].iov_len -= cl->rdma_conn->send_buf_pos;
-                        cl->rdma_conn->send_buf_pos = 0;
-                    }
-                    try_send_rdma(cl);
                 }
+                assert(rc->send_pos >= send_pos);
+                if (rc->send_pos == send_pos)
+                {
+                    rc->send_buf_pos -= send_buf_pos;
+                }
+                rc->send_pos -= send_pos;
+                for (int i = 0; i < send_pos; i++)
+                {
+                    if (cl->outbox[i].flags & MSGR_SENDP_FREE)
+                    {
+                        // Reply fully sent
+                        delete cl->outbox[i].op;
+                    }
+                }
+                if (send_pos > 0)
+                {
+                    cl->send_list.erase(cl->send_list.begin(), cl->send_list.begin()+send_pos);
+                    cl->outbox.erase(cl->outbox.begin(), cl->outbox.begin()+send_pos);
+                }
+                if (send_buf_pos > 0)
+                {
+                    cl->send_list[0].iov_base = (uint8_t*)cl->send_list[0].iov_base + send_buf_pos;
+                    cl->send_list[0].iov_len -= send_buf_pos;
+                }
+                try_send_rdma(cl);
             }
         }
     } while (event_count > 0);

src/msgr_rdma.h

@@ -49,8 +49,9 @@ struct msgr_rdma_connection_t
     uint64_t max_msg = 0;
 
     int send_pos = 0, send_buf_pos = 0;
-    int recv_pos = 0, recv_buf_pos = 0;
+    int next_recv_buf = 0;
     std::vector<void*> recv_buffers;
+    std::vector<uint64_t> send_sizes;
 
     ~msgr_rdma_connection_t();
     static msgr_rdma_connection_t *create(msgr_rdma_context_t *ctx, uint32_t max_send, uint32_t max_recv, uint32_t max_sge, uint32_t max_msg);

src/msgr_receive.cpp

@@ -313,17 +313,18 @@ bool osd_messenger_t::handle_reply_hdr(osd_client_t *cl)
             stop_client(cl->peer_fd);
             return false;
         }
-        if (op->reply.hdr.retval >= 0 && bmp_len > 0)
+        if (bmp_len > 0)
         {
             assert(op->bitmap);
             cl->recv_list.push_back(op->bitmap, bmp_len);
+            cl->read_remaining += bmp_len;
         }
         if (op->reply.hdr.retval > 0)
         {
             assert(op->iov.count > 0);
             cl->recv_list.append(op->iov);
+            cl->read_remaining += op->reply.hdr.retval;
         }
-        cl->read_remaining = op->reply.hdr.retval + bmp_len;
         if (cl->read_remaining == 0)
         {
             goto reuse;

src/object_id.h

@@ -39,6 +39,11 @@ struct __attribute__((__packed__)) obj_ver_id
     uint64_t version;
 };
 
+inline bool operator == (const obj_ver_id & a, const obj_ver_id & b)
+{
+    return a.oid == b.oid && a.version == b.version;
+}
+
 inline bool operator < (const obj_ver_id & a, const obj_ver_id & b)
 {
     return a.oid < b.oid || a.oid == b.oid && a.version < b.version;

src/osd.cpp

@@ -35,18 +35,18 @@ osd_t::osd_t(const json11::Json & config, ring_loop_t *ringloop)
 
     this->ringloop = ringloop;
 
-    this->config = msgr.read_config(config).object_items();
-    if (this->config.find("log_level") == this->config.end())
-        this->config["log_level"] = 1;
-    parse_config(this->config, true);
+    this->cli_config = config.object_items();
+    this->file_config = msgr.read_config(this->cli_config);
+    parse_config(true);
 
     epmgr = new epoll_manager_t(ringloop);
     // FIXME: Use timerfd_interval based directly on io_uring
     this->tfd = epmgr->tfd;
 
-    auto bs_cfg = json_to_bs(this->config);
-    this->bs = new blockstore_t(bs_cfg, ringloop, tfd);
+    if (!json_is_true(this->config["disable_blockstore"]))
     {
+        auto bs_cfg = json_to_bs(this->config);
+        this->bs = new blockstore_t(bs_cfg, ringloop, tfd);
         // Autosync based on the number of unstable writes to prevent stalls due to insufficient journal space
         uint64_t max_autosync = bs->get_journal_size() / bs->get_block_size() / 2;
         if (autosync_writes > max_autosync)
@@ -67,11 +67,11 @@ osd_t::osd_t(const json11::Json & config, ring_loop_t *ringloop)
         }
     }
 
-    this->tfd->set_timer(print_stats_interval*1000, true, [this](int timer_id)
+    print_stats_timer_id = this->tfd->set_timer(print_stats_interval*1000, true, [this](int timer_id)
     {
         print_stats();
     });
-    this->tfd->set_timer(slow_log_interval*1000, true, [this](int timer_id)
+    slow_log_timer_id = this->tfd->set_timer(slow_log_interval*1000, true, [this](int timer_id)
     {
         print_slow();
     });
@@ -91,18 +91,42 @@ osd_t::osd_t(const json11::Json & config, ring_loop_t *ringloop)
 
 osd_t::~osd_t()
 {
+    if (slow_log_timer_id >= 0)
+    {
+        tfd->clear_timer(slow_log_timer_id);
+        slow_log_timer_id = -1;
+    }
+    if (print_stats_timer_id >= 0)
+    {
+        tfd->clear_timer(print_stats_timer_id);
+        print_stats_timer_id = -1;
+    }
+    if (autosync_timer_id >= 0)
+    {
+        tfd->clear_timer(autosync_timer_id);
+        autosync_timer_id = -1;
+    }
     ringloop->unregister_consumer(&consumer);
     delete epmgr;
-    delete bs;
+    if (bs)
+        delete bs;
     close(listen_fd);
     free(zero_buffer);
 }
 
-void osd_t::parse_config(const json11::Json & config, bool allow_disk_params)
+void osd_t::parse_config(bool init)
 {
+    config = msgr.merge_configs(cli_config, file_config, etcd_global_config, etcd_osd_config);
+    if (config.find("log_level") == this->config.end())
+        config["log_level"] = 1;
+    if (bs)
+    {
+        auto bs_cfg = json_to_bs(config);
+        bs->parse_config(bs_cfg);
+    }
     st_cli.parse_config(config);
     msgr.parse_config(config);
-    if (allow_disk_params)
+    if (init)
     {
         // OSD number
         osd_num = config["osd_num"].uint64_value();
@@ -124,24 +148,27 @@ void osd_t::parse_config(const json11::Json & config, bool allow_disk_params)
             immediate_commit = IMMEDIATE_SMALL;
         else
             immediate_commit = IMMEDIATE_NONE;
+        // Bind address
+        bind_address = config["bind_address"].string_value();
+        if (bind_address == "")
+            bind_address = "0.0.0.0";
+        bind_port = config["bind_port"].uint64_value();
+        if (bind_port <= 0 || bind_port > 65535)
+            bind_port = 0;
+        // OSD configuration
+        etcd_report_interval = config["etcd_report_interval"].uint64_value();
+        if (etcd_report_interval <= 0)
+            etcd_report_interval = 5;
+        readonly = json_is_true(config["readonly"]);
+        run_primary = !json_is_false(config["run_primary"]);
+        allow_test_ops = json_is_true(config["allow_test_ops"]);
     }
-    // Bind address
-    bind_address = config["bind_address"].string_value();
-    if (bind_address == "")
-        bind_address = "0.0.0.0";
-    bind_port = config["bind_port"].uint64_value();
-    if (bind_port <= 0 || bind_port > 65535)
-        bind_port = 0;
-    // OSD configuration
     log_level = config["log_level"].uint64_value();
-    etcd_report_interval = config["etcd_report_interval"].uint64_value();
-    if (etcd_report_interval <= 0)
-        etcd_report_interval = 5;
-    readonly = json_is_true(config["readonly"]);
-    run_primary = !json_is_false(config["run_primary"]);
+    auto old_no_rebalance = no_rebalance;
     no_rebalance = json_is_true(config["no_rebalance"]);
+    auto old_no_recovery = no_recovery;
     no_recovery = json_is_true(config["no_recovery"]);
-    allow_test_ops = json_is_true(config["allow_test_ops"]);
+    auto old_autosync_interval = autosync_interval;
     if (!config["autosync_interval"].is_null())
     {
         // Allow to set it to 0
@@ -163,18 +190,52 @@ void osd_t::parse_config(const json11::Json & config, bool allow_disk_params)
     recovery_queue_depth = config["recovery_queue_depth"].uint64_value();
     if (recovery_queue_depth < 1 || recovery_queue_depth > MAX_RECOVERY_QUEUE)
         recovery_queue_depth = DEFAULT_RECOVERY_QUEUE;
+    recovery_pg_switch = config["recovery_pg_switch"].uint64_value();
+    if (recovery_pg_switch < 1)
+        recovery_pg_switch = DEFAULT_RECOVERY_PG_SWITCH;
     recovery_sync_batch = config["recovery_sync_batch"].uint64_value();
     if (recovery_sync_batch < 1 || recovery_sync_batch > MAX_RECOVERY_QUEUE)
         recovery_sync_batch = DEFAULT_RECOVERY_BATCH;
+    auto old_print_stats_interval = print_stats_interval;
     print_stats_interval = config["print_stats_interval"].uint64_value();
     if (!print_stats_interval)
         print_stats_interval = 3;
+    auto old_slow_log_interval = slow_log_interval;
     slow_log_interval = config["slow_log_interval"].uint64_value();
     if (!slow_log_interval)
         slow_log_interval = 10;
     inode_vanish_time = config["inode_vanish_time"].uint64_value();
     if (!inode_vanish_time)
         inode_vanish_time = 60;
+    if ((old_no_rebalance && !no_rebalance || old_no_recovery && !no_recovery) &&
+        !(peering_state & (OSD_RECOVERING | OSD_FLUSHING_PGS)))
+    {
+        peering_state = peering_state | OSD_RECOVERING;
+    }
+    if (old_autosync_interval != autosync_interval && autosync_timer_id >= 0)
+    {
+        this->tfd->clear_timer(autosync_timer_id);
+        autosync_timer_id = this->tfd->set_timer(autosync_interval*1000, true, [this](int timer_id)
+        {
+            autosync();
+        });
+    }
+    if (old_print_stats_interval != print_stats_interval && print_stats_timer_id >= 0)
+    {
+        tfd->clear_timer(print_stats_timer_id);
+        print_stats_timer_id = this->tfd->set_timer(print_stats_interval*1000, true, [this](int timer_id)
+        {
+            print_stats();
+        });
+    }
+    if (old_slow_log_interval != slow_log_interval && slow_log_timer_id >= 0)
+    {
+        tfd->clear_timer(slow_log_timer_id);
+        slow_log_timer_id = this->tfd->set_timer(slow_log_interval*1000, true, [this](int timer_id)
+        {
+            print_slow();
+        });
+    }
 }
 
 void osd_t::bind_socket()
@@ -399,7 +460,7 @@ void osd_t::print_slow()
                 int l = sizeof(alloc), n;
                 char *buf = alloc;
 #define bufprintf(s, ...) { n = snprintf(buf, l, s, __VA_ARGS__); n = n < 0 ? 0 : n; buf += n; l -= n; }
-                bufprintf("[OSD %lu] Slow op", osd_num);
+                bufprintf("[OSD %lu] Slow op %lx", osd_num, (unsigned long)op);
                 if (kv.second->osd_num)
                 {
                     bufprintf(" from peer OSD %lu (client %d)", kv.second->osd_num, kv.second->peer_fd);
@@ -472,7 +533,7 @@ void osd_t::print_slow()
             }
         }
     }
-    if (has_slow)
+    if (has_slow && bs)
     {
         bs->dump_diagnostics();
     }

src/osd.h

@@ -34,6 +34,7 @@
 #define DEFAULT_AUTOSYNC_WRITES 128
 #define MAX_RECOVERY_QUEUE 2048
 #define DEFAULT_RECOVERY_QUEUE 4
+#define DEFAULT_RECOVERY_PG_SWITCH 128
 #define DEFAULT_RECOVERY_BATCH 16
 
 //#define OSD_STUB
@@ -89,7 +90,7 @@ class osd_t
 {
     // config
 
-    json11::Json::object config;
+    json11::Json::object cli_config, file_config, etcd_global_config, etcd_osd_config, config;
     int etcd_report_interval = 5;
 
     bool readonly = false;
@@ -108,6 +109,7 @@ class osd_t
     int autosync_interval = DEFAULT_AUTOSYNC_INTERVAL; // "emergency" sync every 5 seconds
     int autosync_writes = DEFAULT_AUTOSYNC_WRITES;
     int recovery_queue_depth = DEFAULT_RECOVERY_QUEUE;
+    int recovery_pg_switch = DEFAULT_RECOVERY_PG_SWITCH;
     int recovery_sync_batch = DEFAULT_RECOVERY_BATCH;
     int inode_vanish_time = 60;
     int log_level = 0;
@@ -124,6 +126,7 @@ class osd_t
     bool pg_config_applied = false;
     bool etcd_reporting_pg_state = false;
     bool etcd_reporting_stats = false;
+    int autosync_timer_id = -1, print_stats_timer_id = -1, slow_log_timer_id = -1;
 
     // peers and PGs
 
@@ -135,7 +138,10 @@ class osd_t
     uint64_t misplaced_objects = 0, degraded_objects = 0, incomplete_objects = 0;
     int peering_state = 0;
     std::map<object_id, osd_recovery_op_t> recovery_ops;
-    int recovery_done = 0;
+    bool recovery_last_degraded = true;
+    pool_pg_num_t recovery_last_pg;
+    object_id recovery_last_oid;
+    int recovery_pg_done = 0, recovery_done = 0;
     osd_op_t *autosync_op = NULL;
 
     // Unstable writes
@@ -147,7 +153,7 @@ class osd_t
 
     bool stopping = false;
     int inflight_ops = 0;
-    blockstore_t *bs;
+    blockstore_t *bs = NULL;
     void *zero_buffer = NULL;
     uint64_t zero_buffer_size = 0;
     uint32_t bs_block_size, bs_bitmap_granularity, clean_entry_bitmap_size;
@@ -168,7 +174,7 @@ class osd_t
     uint64_t recovery_stat_bytes[2][2] = {};
 
     // cluster connection
-    void parse_config(const json11::Json & config, bool allow_disk_params);
+    void parse_config(bool init);
     void init_cluster();
     void on_change_osd_state_hook(osd_num_t peer_osd);
     void on_change_pg_history_hook(pool_id_t pool_id, pg_num_t pg_num);
@@ -200,7 +206,6 @@ class osd_t
     bool check_peer_config(osd_client_t *cl, json11::Json conf);
     void repeer_pgs(osd_num_t osd_num);
     void start_pg_peering(pg_t & pg);
-    void submit_sync_and_list_subop(osd_num_t role_osd, pg_peering_state_t *ps);
     void submit_list_subop(osd_num_t role_osd, pg_peering_state_t *ps);
     void discard_list_subop(osd_op_t *list_op);
     bool stop_pg(pg_t & pg);

src/osd_cluster.cpp

@@ -75,7 +75,7 @@ void osd_t::init_cluster()
     }
     if (run_primary && autosync_interval > 0)
     {
-        this->tfd->set_timer(autosync_interval*1000, true, [this](int timer_id)
+        autosync_timer_id = this->tfd->set_timer(autosync_interval*1000, true, [this](int timer_id)
         {
             autosync();
         });
@@ -132,7 +132,7 @@ bool osd_t::check_peer_config(osd_client_t *cl, json11::Json conf)
                 this->osd_num, immediate_commit == IMMEDIATE_ALL ? "all" : "small",
                 cl->osd_num, conf["immediate_commit"].string_value().c_str()
             );
-            return true;
+            return false;
         }
         else if (conf["block_size"].uint64_value() != (uint64_t)this->bs_block_size)
         {
@@ -140,7 +140,7 @@ bool osd_t::check_peer_config(osd_client_t *cl, json11::Json conf)
                 "[OSD %lu] My block_size is %u, but peer OSD %lu has %lu. We can't work together\n",
                 this->osd_num, this->bs_block_size, cl->osd_num, conf["block_size"].uint64_value()
             );
-            return true;
+            return false;
         }
         else if (conf["bitmap_granularity"].uint64_value() != (uint64_t)this->bs_bitmap_granularity)
         {
@@ -148,7 +148,7 @@ bool osd_t::check_peer_config(osd_client_t *cl, json11::Json conf)
                 "[OSD %lu] My bitmap_granularity is %u, but peer OSD %lu has %lu. We can't work together\n",
                 this->osd_num, this->bs_bitmap_granularity, cl->osd_num, conf["bitmap_granularity"].uint64_value()
             );
-            return true;
+            return false;
         }
     }
     return true;
@@ -182,10 +182,10 @@ json11::Json osd_t::get_statistics()
     char time_str[50] = { 0 };
     sprintf(time_str, "%ld.%03ld", ts.tv_sec, ts.tv_nsec/1000000);
     st["time"] = time_str;
-    st["blockstore_ready"] = bs->is_started();
-    st["data_block_size"] = (uint64_t)bs->get_block_size();
     if (bs)
     {
+        st["blockstore_ready"] = bs->is_started();
+        st["data_block_size"] = (uint64_t)bs->get_block_size();
         st["size"] = bs->get_block_count() * bs->get_block_size();
         st["free"] = bs->get_free_block_count() * bs->get_block_size();
     }
@@ -233,7 +233,8 @@ void osd_t::report_statistics()
     json11::Json::object inode_space;
     json11::Json::object last_stat;
     pool_id_t last_pool = 0;
-    auto & bs_inode_space = bs->get_inode_space_stats();
+    std::map<uint64_t, uint64_t> bs_empty_space;
+    auto & bs_inode_space = bs ? bs->get_inode_space_stats() : bs_empty_space;
     for (auto kv: bs_inode_space)
     {
         pool_id_t pool_id = INODE_POOL(kv.first);
@@ -374,7 +375,11 @@ void osd_t::on_change_osd_state_hook(osd_num_t peer_osd)
 
 void osd_t::on_change_etcd_state_hook(std::map<std::string, etcd_kv_t> & changes)
 {
-    // FIXME apply config changes in runtime (maybe, some)
+    if (changes.find(st_cli.etcd_prefix+"/config/global") != changes.end())
+    {
+        etcd_global_config = changes[st_cli.etcd_prefix+"/config/global"].value.object_items();
+        parse_config(false);
+    }
     if (run_primary)
     {
         apply_pg_count();
@@ -382,37 +387,10 @@ void osd_t::on_change_etcd_state_hook(std::map<std::string, etcd_kv_t> & changes
     }
 }
 
-void osd_t::on_change_pg_history_hook(pool_id_t pool_id, pg_num_t pg_num)
-{
-    auto pg_it = pgs.find({
-        .pool_id = pool_id,
-        .pg_num = pg_num,
-    });
-    if (pg_it != pgs.end() && pg_it->second.epoch > pg_it->second.reported_epoch &&
-        st_cli.pool_config[pool_id].pg_config[pg_num].epoch >= pg_it->second.epoch)
-    {
-        pg_it->second.reported_epoch = st_cli.pool_config[pool_id].pg_config[pg_num].epoch;
-        object_id oid = { 0 };
-        bool first = true;
-        for (auto op: pg_it->second.write_queue)
-        {
-            if (first || oid != op.first)
-            {
-                oid = op.first;
-                first = false;
-                continue_primary_write(op.second);
-            }
-        }
-    }
-}
-
 void osd_t::on_load_config_hook(json11::Json::object & global_config)
 {
-    json11::Json::object osd_config = this->config;
-    for (auto & kv: global_config)
-        if (osd_config.find(kv.first) == osd_config.end())
-            osd_config[kv.first] = kv.second;
-    parse_config(osd_config, false);
+    etcd_global_config = global_config;
+    parse_config(true);
     bind_socket();
     acquire_lease();
 }
@@ -704,13 +682,16 @@ void osd_t::apply_pg_config()
                         }
                     }
                 }
+                auto vec_all_peers = std::vector<osd_num_t>(all_peers.begin(), all_peers.end());
                 if (currently_taken)
                 {
-                    if (pg_it->second.state & (PG_ACTIVE | PG_INCOMPLETE | PG_PEERING | PG_REPEERING | PG_PEERED))
+                    if (pg_it->second.state & (PG_ACTIVE | PG_INCOMPLETE | PG_PEERING | PG_REPEERING))
                     {
-                        if (pg_it->second.target_set == pg_cfg.target_set)
+                        if (pg_it->second.target_set == pg_cfg.target_set &&
+                            pg_it->second.target_history == pg_cfg.target_history &&
+                            pg_it->second.all_peers == vec_all_peers)
                         {
-                            // No change in osd_set; history changes are ignored
+                            // No change in osd_set and history
                             continue;
                         }
                         else
@@ -755,13 +736,13 @@ void osd_t::apply_pg_config()
                     .pg_cursize = 0,
                     .pg_size = pool_item.second.pg_size,
                     .pg_minsize = pool_item.second.pg_minsize,
-                    .pg_data_size = pg.scheme == POOL_SCHEME_REPLICATED
+                    .pg_data_size = pool_item.second.scheme == POOL_SCHEME_REPLICATED
                          ? 1 : pool_item.second.pg_size - pool_item.second.parity_chunks,
                     .pool_id = pool_id,
                     .pg_num = pg_num,
                     .reported_epoch = pg_cfg.epoch,
                     .target_history = pg_cfg.target_history,
-                    .all_peers = std::vector<osd_num_t>(all_peers.begin(), all_peers.end()),
+                    .all_peers = vec_all_peers,
                     .target_set = pg_cfg.target_set,
                 };
                 if (pg.scheme == POOL_SCHEME_EC)
@@ -984,13 +965,6 @@ void osd_t::report_pg_states()
                         }
                         this->pgs.erase(pg_it);
                     }
-                    else if (pg_it->second.state & PG_PEERED)
-                    {
-                        // Activate PG after PG PEERED state is reported along with history
-                        // (if the state wasn't changed again)
-                        pg_it->second.state = pg_it->second.state & ~PG_PEERED | PG_ACTIVE;
-                        report_pg_state(pg_it->second);
-                    }
                 }
             }
             // Push other PG state updates, if any

src/osd_flush.cpp

@@ -64,6 +64,11 @@ void osd_t::submit_pg_flush_ops(pg_t & pg)
 
 void osd_t::handle_flush_op(bool rollback, pool_id_t pool_id, pg_num_t pg_num, pg_flush_batch_t *fb, osd_num_t peer_osd, int retval)
 {
+    if (log_level > 2)
+    {
+        printf("[PG %u/%u] flush batch %lx completed on OSD %lu with result %d\n",
+            pool_id, pg_num, (uint64_t)fb, peer_osd, retval);
+    }
     pool_pg_num_t pg_id = { .pool_id = pool_id, .pg_num = pg_num };
     if (pgs.find(pg_id) == pgs.end() || pgs[pg_id].flush_batch != fb)
     {
@@ -99,10 +104,9 @@ void osd_t::handle_flush_op(bool rollback, pool_id_t pool_id, pg_num_t pg_num, p
         std::vector<osd_op_t*> continue_ops;
         auto & pg = pgs.at(pg_id);
         auto it = pg.flush_actions.begin(), prev_it = it;
-        auto erase_start = it;
         while (1)
         {
-            if (it == pg.flush_actions.end() ||
+            if (it == pg.flush_actions.end() || !it->second.submitted ||
                 it->first.oid.inode != prev_it->first.oid.inode ||
                 (it->first.oid.stripe & ~STRIPE_MASK) != (prev_it->first.oid.stripe & ~STRIPE_MASK))
             {
@@ -116,29 +120,23 @@ void osd_t::handle_flush_op(bool rollback, pool_id_t pool_id, pg_num_t pg_num, p
                 });
                 if (wr_it != pg.write_queue.end())
                 {
+                    if (log_level > 2)
+                    {
+                        printf("[PG %u/%u] continuing write %lx to object %lx:%lx after flush\n",
+                            pool_id, pg_num, (uint64_t)wr_it->second, wr_it->first.inode, wr_it->first.stripe);
+                    }
                     continue_ops.push_back(wr_it->second);
-                    pg.write_queue.erase(wr_it);
                 }
             }
-            if ((it == pg.flush_actions.end() || !it->second.submitted) &&
-                erase_start != it)
-            {
-                pg.flush_actions.erase(erase_start, it);
-            }
-            if (it == pg.flush_actions.end())
+            if (it == pg.flush_actions.end() || !it->second.submitted)
             {
+                if (it != pg.flush_actions.begin())
+                {
+                    pg.flush_actions.erase(pg.flush_actions.begin(), it);
+                }
                 break;
             }
-            prev_it = it;
-            if (!it->second.submitted)
-            {
-                it++;
-                erase_start = it;
-            }
-            else
-            {
-                it++;
-            }
+            prev_it = it++;
         }
         delete fb;
         pg.flush_batch = NULL;
@@ -168,6 +166,18 @@ bool osd_t::submit_flush_op(pool_id_t pool_id, pg_num_t pg_num, pg_flush_batch_t
     // Copy buffer so it gets freed along with the operation
     op->buf = malloc_or_die(sizeof(obj_ver_id) * count);
     memcpy(op->buf, data, sizeof(obj_ver_id) * count);
+    if (log_level > 2)
+    {
+        printf(
+            "[PG %u/%u] flush batch %lx on OSD %lu: %s objects: ",
+            pool_id, pg_num, (uint64_t)fb, peer_osd, rollback ? "rollback" : "stabilize"
+        );
+        for (int i = 0; i < count; i++)
+        {
+            printf(i > 0 ? ", %lx:%lx v%lu" : "%lx:%lx v%lu", data[i].oid.inode, data[i].oid.stripe, data[i].version);
+        }
+        printf("\n");
+    }
     if (peer_osd == this->osd_num)
     {
         // local
@@ -226,42 +236,51 @@ bool osd_t::submit_flush_op(pool_id_t pool_id, pg_num_t pg_num, pg_flush_batch_t
 
 bool osd_t::pick_next_recovery(osd_recovery_op_t &op)
 {
-    if (!no_recovery)
+    if (!pgs.size())
     {
-        for (auto pg_it = pgs.begin(); pg_it != pgs.end(); pg_it++)
-        {
-            if ((pg_it->second.state & (PG_ACTIVE | PG_HAS_DEGRADED)) == (PG_ACTIVE | PG_HAS_DEGRADED))
-            {
-                for (auto obj_it = pg_it->second.degraded_objects.begin(); obj_it != pg_it->second.degraded_objects.end(); obj_it++)
-                {
-                    if (recovery_ops.find(obj_it->first) == recovery_ops.end())
-                    {
-                        op.degraded = true;
-                        op.oid = obj_it->first;
-                        return true;
-                    }
-                }
-            }
-        }
+        return false;
     }
-    if (!no_rebalance)
+    // Restart scanning from the same degraded/misplaced status as the last time
+    for (int tried_degraded = 0; tried_degraded < 2; tried_degraded++)
     {
-        for (auto pg_it = pgs.begin(); pg_it != pgs.end(); pg_it++)
+        if (recovery_last_degraded ? !no_recovery : !no_rebalance)
         {
             // Don't try to "recover" misplaced objects if "recovery" would make them degraded
-            if ((pg_it->second.state & (PG_ACTIVE | PG_DEGRADED | PG_HAS_MISPLACED)) == (PG_ACTIVE | PG_HAS_MISPLACED))
+            auto mask = recovery_last_degraded ? (PG_ACTIVE | PG_HAS_DEGRADED) : (PG_ACTIVE | PG_DEGRADED | PG_HAS_MISPLACED);
+            auto check = recovery_last_degraded ? (PG_ACTIVE | PG_HAS_DEGRADED) : (PG_ACTIVE | PG_HAS_MISPLACED);
+            // Restart scanning from the same PG as the last time
+            for (auto pg_it = pgs.lower_bound(recovery_last_pg); pg_it != pgs.end(); pg_it++)
             {
-                for (auto obj_it = pg_it->second.misplaced_objects.begin(); obj_it != pg_it->second.misplaced_objects.end(); obj_it++)
+                if ((pg_it->second.state & mask) == check)
                 {
-                    if (recovery_ops.find(obj_it->first) == recovery_ops.end())
+                    auto & src = recovery_last_degraded ? pg_it->second.degraded_objects : pg_it->second.misplaced_objects;
+                    assert(src.size() > 0);
+                    // Restart scanning from the next object
+                    for (auto obj_it = src.upper_bound(recovery_last_oid); obj_it != src.end(); obj_it++)
                     {
-                        op.degraded = false;
-                        op.oid = obj_it->first;
-                        return true;
+                        if (recovery_ops.find(obj_it->first) == recovery_ops.end())
+                        {
+                            op.degraded = recovery_last_degraded;
+                            recovery_last_oid = op.oid = obj_it->first;
+                            recovery_pg_done++;
+                            // Switch to another PG after recovery_pg_switch operations
+                            // to always mix all PGs during recovery but still benefit
+                            // from recovery queue depth greater than 1
+                            if (recovery_pg_done >= recovery_pg_switch)
+                            {
+                                recovery_pg_done = 0;
+                                recovery_last_pg.pg_num++;
+                                recovery_last_oid = {};
+                            }
+                            return true;
+                        }
                     }
                 }
             }
         }
+        recovery_last_degraded = !recovery_last_degraded;
+        recovery_last_pg = {};
+        recovery_last_oid = {};
     }
     return false;
 }
@@ -295,9 +314,10 @@ void osd_t::submit_recovery_op(osd_recovery_op_t *op)
             {
                 // PG is stopped or one of the OSDs is gone, error is harmless
                 printf(
-                    "Recovery operation failed with object %lx:%lx (PG %u/%u)\n",
-                    op->oid.inode, op->oid.stripe, INODE_POOL(op->oid.inode),
-                    map_to_pg(op->oid, st_cli.pool_config.at(INODE_POOL(op->oid.inode)).pg_stripe_size)
+                    "[PG %u/%u] Recovery operation failed with object %lx:%lx\n",
+                    INODE_POOL(op->oid.inode),
+                    map_to_pg(op->oid, st_cli.pool_config.at(INODE_POOL(op->oid.inode)).pg_stripe_size),
+                    op->oid.inode, op->oid.stripe
                 );
             }
             else

src/osd_id.h

@@ -28,3 +28,13 @@ inline bool operator < (const pool_pg_num_t & a, const pool_pg_num_t & b)
 {
     return a.pool_id < b.pool_id || a.pool_id == b.pool_id && a.pg_num < b.pg_num;
 }
+
+inline bool operator == (const pool_pg_num_t & a, const pool_pg_num_t & b)
+{
+    return a.pool_id == b.pool_id && a.pg_num == b.pg_num;
+}
+
+inline bool operator != (const pool_pg_num_t & a, const pool_pg_num_t & b)
+{
+    return a.pool_id != b.pool_id || a.pg_num != b.pg_num;
+}

src/osd_ops.h

@@ -191,7 +191,7 @@ struct __attribute__((__packed__)) osd_op_rw_t
     uint64_t inode;
     // offset
     uint64_t offset;
-    // length
+    // length. 0 means to read all bitmaps of the specified range, but no data.
     uint32_t len;
     // flags (for future)
     uint32_t flags;

src/osd_peering.cpp

@@ -32,7 +32,16 @@ void osd_t::handle_peers()
                     if (p.second.state & PG_HAS_UNCLEAN)
                         peering_state = peering_state | OSD_FLUSHING_PGS;
                     else if (p.second.state & (PG_HAS_DEGRADED | PG_HAS_MISPLACED))
+                    {
                         peering_state = peering_state | OSD_RECOVERING;
+                        if (p.second.state & PG_HAS_DEGRADED)
+                        {
+                            // Restart recovery from degraded objects
+                            recovery_last_degraded = true;
+                            recovery_last_pg = {};
+                            recovery_last_oid = {};
+                        }
+                    }
                     ringloop->wakeup();
                     return;
                 }
@@ -41,10 +50,6 @@ void osd_t::handle_peers()
                     still = true;
                 }
             }
-            else if (p.second.state & PG_PEERED)
-            {
-                still = true;
-            }
         }
         if (!still)
         {
@@ -65,17 +70,13 @@ void osd_t::handle_peers()
                 }
                 still = true;
             }
-            else if (p.second.state & PG_PEERED)
-            {
-                still = true;
-            }
         }
         if (!still)
         {
             peering_state = peering_state & ~OSD_FLUSHING_PGS | OSD_RECOVERING;
         }
     }
-    if ((peering_state & OSD_RECOVERING) && !readonly)
+    if (!(peering_state & OSD_FLUSHING_PGS) && (peering_state & OSD_RECOVERING) && !readonly)
     {
         if (!continue_recovery())
         {
@@ -91,7 +92,7 @@ void osd_t::repeer_pgs(osd_num_t peer_osd)
     {
         auto & pg = p.second;
         bool repeer = false;
-        if (pg.state & (PG_PEERING | PG_PEERED | PG_ACTIVE | PG_INCOMPLETE))
+        if (pg.state & (PG_PEERING | PG_ACTIVE | PG_INCOMPLETE))
         {
             for (osd_num_t pg_osd: pg.all_peers)
             {
@@ -302,82 +303,11 @@ void osd_t::start_pg_peering(pg_t & pg)
         {
             continue;
         }
-        submit_sync_and_list_subop(peer_osd, pg.peering_state);
+        submit_list_subop(peer_osd, pg.peering_state);
     }
     ringloop->wakeup();
 }
 
-void osd_t::submit_sync_and_list_subop(osd_num_t role_osd, pg_peering_state_t *ps)
-{
-    // Sync before listing, if not readonly
-    if (readonly)
-    {
-        submit_list_subop(role_osd, ps);
-    }
-    else if (role_osd == this->osd_num)
-    {
-        // Self
-        osd_op_t *op = new osd_op_t();
-        op->op_type = 0;
-        op->peer_fd = SELF_FD;
-        clock_gettime(CLOCK_REALTIME, &op->tv_begin);
-        op->bs_op = new blockstore_op_t();
-        op->bs_op->opcode = BS_OP_SYNC;
-        op->bs_op->callback = [this, ps, op, role_osd](blockstore_op_t *bs_op)
-        {
-            if (bs_op->retval < 0)
-            {
-                printf("Local OP_SYNC failed: %d (%s)\n", bs_op->retval, strerror(-bs_op->retval));
-                force_stop(1);
-                return;
-            }
-            add_bs_subop_stats(op);
-            delete op->bs_op;
-            op->bs_op = NULL;
-            delete op;
-            ps->list_ops.erase(role_osd);
-            submit_list_subop(role_osd, ps);
-        };
-        ps->list_ops[role_osd] = op;
-        bs->enqueue_op(op->bs_op);
-    }
-    else
-    {
-        // Peer
-        auto & cl = msgr.clients.at(msgr.osd_peer_fds.at(role_osd));
-        osd_op_t *op = new osd_op_t();
-        op->op_type = OSD_OP_OUT;
-        op->peer_fd = cl->peer_fd;
-        op->req = (osd_any_op_t){
-            .sec_sync = {
-                .header = {
-                    .magic = SECONDARY_OSD_OP_MAGIC,
-                    .id = msgr.next_subop_id++,
-                    .opcode = OSD_OP_SEC_SYNC,
-                },
-            },
-        };
-        op->callback = [this, ps, role_osd](osd_op_t *op)
-        {
-            if (op->reply.hdr.retval < 0)
-            {
-                // FIXME: Mark peer as failed and don't reconnect immediately after dropping the connection
-                printf("Failed to sync OSD %lu: %ld (%s), disconnecting peer\n", role_osd, op->reply.hdr.retval, strerror(-op->reply.hdr.retval));
-                int fail_fd = op->peer_fd;
-                ps->list_ops.erase(role_osd);
-                delete op;
-                msgr.stop_client(fail_fd);
-                return;
-            }
-            delete op;
-            ps->list_ops.erase(role_osd);
-            submit_list_subop(role_osd, ps);
-        };
-        ps->list_ops[role_osd] = op;
-        msgr.outbox_push(op);
-    }
-}
-
 void osd_t::submit_list_subop(osd_num_t role_osd, pg_peering_state_t *ps)
 {
     if (role_osd == this->osd_num)
@@ -551,13 +481,21 @@ void osd_t::report_pg_state(pg_t & pg)
         pg.history_changed = true;
         pg.target_history.clear();
         pg.all_peers = pg.target_set;
+        std::sort(pg.all_peers.begin(), pg.all_peers.end());
         pg.cur_peers = pg.target_set;
+        // Change pg_config at the same time, otherwise our PG reconciling loop may try to apply the old metadata
+        auto & pg_cfg = st_cli.pool_config[pg.pool_id].pg_config[pg.pg_num];
+        pg_cfg.target_history = pg.target_history;
+        pg_cfg.all_peers = pg.all_peers;
     }
     else if (pg.state == (PG_ACTIVE|PG_LEFT_ON_DEAD))
     {
         // Clear history of active+left_on_dead PGs, but leave dead OSDs in all_peers
-        pg.history_changed = true;
-        pg.target_history.clear();
+        if (pg.target_history.size())
+        {
+            pg.history_changed = true;
+            pg.target_history.clear();
+        }
         std::set<osd_num_t> dead_peers;
         for (auto pg_osd: pg.all_peers)
         {
@@ -574,8 +512,12 @@ void osd_t::report_pg_state(pg_t & pg)
                 dead_peers.insert(pg_osd);
             }
         }
-        pg.all_peers.clear();
-        pg.all_peers.insert(pg.all_peers.begin(), dead_peers.begin(), dead_peers.end());
+        auto new_all_peers = std::vector<osd_num_t>(dead_peers.begin(), dead_peers.end());
+        if (pg.all_peers != new_all_peers)
+        {
+            pg.history_changed = true;
+            pg.all_peers = new_all_peers;
+        }
         pg.cur_peers.clear();
         for (auto pg_osd: pg.target_set)
         {
@@ -584,6 +526,9 @@ void osd_t::report_pg_state(pg_t & pg)
                 pg.cur_peers.push_back(pg_osd);
             }
         }
+        auto & pg_cfg = st_cli.pool_config[pg.pool_id].pg_config[pg.pg_num];
+        pg_cfg.target_history = pg.target_history;
+        pg_cfg.all_peers = pg.all_peers;
     }
     if (pg.state == PG_OFFLINE && !this->pg_config_applied)
     {

src/osd_peering_pg.cpp

@@ -86,25 +86,12 @@ void pg_obj_state_check_t::walk()
     }
     if (pg->pg_cursize < pg->pg_size)
     {
-        // Report PG history and activate
-        pg->state |= PG_DEGRADED | PG_PEERED;
-        std::vector<osd_num_t> history_set;
-        for (auto peer_osd: pg->cur_set)
-        {
-            if (peer_osd != 0)
-            {
-                history_set.push_back(peer_osd);
-            }
-        }
-        pg->target_history.push_back(history_set);
-        pg->history_changed = true;
+        // Activate as degraded
+        // Current OSD set will be added into target_history on first write
+        pg->state |= PG_DEGRADED;
     }
-    else
-    {
-        // Just activate
-        pg->state |= PG_ACTIVE;
-    }
-    if (pg->state == PG_ACTIVE && pg->cur_peers.size() < pg->all_peers.size())
+    pg->state |= PG_ACTIVE;
+    if (pg->cur_peers.size() < pg->all_peers.size())
     {
         pg->state |= PG_LEFT_ON_DEAD;
     }
@@ -435,21 +422,44 @@ void pg_t::calc_object_states(int log_level)
     std::sort(st.list.begin(), st.list.end());
     // Walk over it and check object states
     st.walk();
-    if (this->state & (PG_DEGRADED|PG_LEFT_ON_DEAD))
+    if (this->state != PG_ACTIVE)
     {
         assert(epoch != (((uint64_t)1 << PG_EPOCH_BITS)-1));
         epoch++;
     }
+    if (log_level > 0)
+    {
+        std::string osd_set_desc;
+        for (auto & osd_num: target_set)
+        {
+            osd_set_desc += (osd_set_desc == "" ? "" : ", ")+std::to_string(osd_num);
+        }
+        printf(
+            "[PG %u/%u] %lu clean objects on target OSD set %s\n",
+            pool_id, pg_num, clean_count, osd_set_desc.c_str()
+        );
+        for (auto & stp: state_dict)
+        {
+            osd_set_desc = "";
+            for (auto & loc: stp.first)
+            {
+                osd_set_desc += (osd_set_desc == "" ? "" : ", ")+
+                    std::to_string(loc.osd_num)+
+                    (st.replicated ? "" : "("+std::to_string(loc.role)+")")+
+                    (loc.outdated ? "(old)" : "");
+            }
+            printf("[PG %u/%u] %lu objects on OSD set %s\n", pool_id, pg_num, stp.second.object_count, osd_set_desc.c_str());
+        }
+    }
 }
 
 void pg_t::print_state()
 {
     printf(
-        "[PG %u/%u] is %s%s%s%s%s%s%s%s%s%s%s%s%s%s%s (%lu objects)\n", pool_id, pg_num,
+        "[PG %u/%u] is %s%s%s%s%s%s%s%s%s%s%s%s%s%s (%lu objects)\n", pool_id, pg_num,
         (state & PG_STARTING) ? "starting" : "",
         (state & PG_OFFLINE) ? "offline" : "",
         (state & PG_PEERING) ? "peering" : "",
-        (state & PG_PEERED) ? "peered" : "",
         (state & PG_INCOMPLETE) ? "incomplete" : "",
         (state & PG_ACTIVE) ? "active" : "",
         (state & PG_REPEERING) ? "repeering" : "",

src/osd_peering_pg_test.cpp

@@ -54,5 +54,6 @@ int main(int argc, char *argv[])
     {
         printf("dev: state=%lx\n", it.second.state);
     }
+    delete pg.peering_state;
     return 0;
 }

src/osd_primary.cpp

@@ -186,10 +186,22 @@ void osd_t::continue_primary_read(osd_op_t *cur_op)
     cur_op->reply.rw.bitmap_len = 0;
     {
         auto & pg = pgs.at({ .pool_id = INODE_POOL(op_data->oid.inode), .pg_num = op_data->pg_num });
-        for (int role = 0; role < op_data->pg_data_size; role++)
+        if (cur_op->req.rw.len == 0)
         {
-            op_data->stripes[role].read_start = op_data->stripes[role].req_start;
-            op_data->stripes[role].read_end = op_data->stripes[role].req_end;
+            // len=0 => bitmap read
+            for (int role = 0; role < op_data->pg_data_size; role++)
+            {
+                op_data->stripes[role].read_start = 0;
+                op_data->stripes[role].read_end = UINT32_MAX;
+            }
+        }
+        else
+        {
+            for (int role = 0; role < op_data->pg_data_size; role++)
+            {
+                op_data->stripes[role].read_start = op_data->stripes[role].req_start;
+                op_data->stripes[role].read_end = op_data->stripes[role].req_end;
+            }
         }
         // Determine version
         auto vo_it = pg.ver_override.find(op_data->oid);
@@ -228,7 +240,7 @@ resume_1:
 resume_2:
     if (op_data->errors > 0)
     {
-        finish_op(cur_op, op_data->epipe > 0 ? -EPIPE : -EIO);
+        finish_op(cur_op, op_data->errcode);
         return;
     }
     cur_op->reply.rw.version = op_data->fact_ver;
@@ -350,7 +362,7 @@ resume_2:
 resume_3:
     if (op_data->errors > 0)
     {
-        pg_cancel_write_queue(pg, cur_op, op_data->oid, op_data->epipe > 0 ? -EPIPE : -EIO);
+        pg_cancel_write_queue(pg, cur_op, op_data->oid, op_data->errcode);
         return;
     }
     // Check CAS version
@@ -371,7 +383,7 @@ resume_4:
 resume_5:
     if (op_data->errors > 0)
     {
-        pg_cancel_write_queue(pg, cur_op, op_data->oid, op_data->epipe > 0 ? -EPIPE : -EIO);
+        pg_cancel_write_queue(pg, cur_op, op_data->oid, op_data->errcode);
         return;
     }
     // Remove version override

src/osd_primary.h

@@ -24,7 +24,7 @@ struct osd_primary_op_data_t
     uint64_t target_ver;
     uint64_t orig_ver = 0, fact_ver = 0;
     uint64_t scheme = 0;
-    int n_subops = 0, done = 0, errors = 0, epipe = 0;
+    int n_subops = 0, done = 0, errors = 0, errcode = 0;
     int degraded = 0, pg_size, pg_data_size;
     osd_rmw_stripe_t *stripes;
     osd_op_t *subops = NULL;