feat: add a new setting to enable v2 data engine fast rebuilding

Longhorn 9488 Signed-off-by: Shuo Wu <[email protected]>
shuo-wu · Dec 3, 2024 · bf5ec5c · bf5ec5c
1 parent 2d5d202
commit bf5ec5c
Show file tree

Hide file tree

Showing 3 changed files with 29 additions and 4 deletions.
diff --git a/controller/volume_controller.go b/controller/volume_controller.go
@@ -4892,7 +4892,12 @@ func (c *VolumeController) shouldCleanUpFailedReplica(v *longhorn.Volume, r *lon
 	}
 	// TODO: Remove it once we can reuse failed replicas during v2 rebuilding
 	if types.IsDataEngineV2(v.Spec.DataEngine) {
-		return true
+		v2DataEngineFastRebuilding, err := c.ds.GetSettingAsBool(types.SettingNameV2DataEngineFastRebuilding)
+		if err != nil {
+			log.WithError(err).Warn("Failed to get the setting for v2 data engine fast rebuilding, will consider it as false")
+			v2DataEngineFastRebuilding = false
+		}
+		return !v2DataEngineFastRebuilding
 	}
 	// Failed too long ago to be useful during a rebuild.
 	if v.Spec.StaleReplicaTimeout > 0 &&

diff --git a/scheduler/replica_scheduler.go b/scheduler/replica_scheduler.go
@@ -568,7 +568,14 @@ func filterActiveReplicas(replicas map[string]*longhorn.Replica) map[string]*lon
 func (rcs *ReplicaScheduler) CheckAndReuseFailedReplica(replicas map[string]*longhorn.Replica, volume *longhorn.Volume, hardNodeAffinity string) (*longhorn.Replica, error) {
 	// TODO: Remove it once we can reuse failed replicas during v2 rebuilding
 	if types.IsDataEngineV2(volume.Spec.DataEngine) {
-		return nil, nil
+		v2DataEngineFastRebuilding, err := rcs.ds.GetSettingAsBool(types.SettingNameV2DataEngineFastRebuilding)
+		if err != nil {
+			logrus.WithError(err).Warn("Failed to get the setting for v2 data engine fast rebuilding, will consider it as false")
+			v2DataEngineFastRebuilding = false
+		}
+		if !v2DataEngineFastRebuilding {
+			return nil, nil
+		}
 	}
 
 	replicas = filterActiveReplicas(replicas)

diff --git a/types/setting.go b/types/setting.go
@@ -135,6 +135,7 @@ const (
 	SettingNameV2DataEngineCPUMask                                      = SettingName("v2-data-engine-cpu-mask")
 	SettingNameV2DataEngineLogLevel                                     = SettingName("v2-data-engine-log-level")
 	SettingNameV2DataEngineLogFlags                                     = SettingName("v2-data-engine-log-flags")
+	SettingNameV2DataEngineFastRebuilding                               = SettingName("v2-data-engine-fast-rebuilding")
 	SettingNameFreezeFilesystemForSnapshot                              = SettingName("freeze-filesystem-for-snapshot")
 	SettingNameAutoCleanupSnapshotWhenDeleteBackup                      = SettingName("auto-cleanup-when-delete-backup")
 	SettingNameDefaultMinNumberOfBackingImageCopies                     = SettingName("default-min-number-of-backing-image-copies")
@@ -225,6 +226,7 @@ var (
 		SettingNameV2DataEngineCPUMask,
 		SettingNameV2DataEngineLogLevel,
 		SettingNameV2DataEngineLogFlags,
+		SettingNameV2DataEngineFastRebuilding,
 		SettingNameReplicaDiskSoftAntiAffinity,
 		SettingNameAllowEmptyNodeSelectorVolume,
 		SettingNameAllowEmptyDiskSelectorVolume,
@@ -347,6 +349,7 @@ var (
 		SettingNameV2DataEngineCPUMask:                                      SettingDefinitionV2DataEngineCPUMask,
 		SettingNameV2DataEngineLogLevel:                                     SettingDefinitionV2DataEngineLogLevel,
 		SettingNameV2DataEngineLogFlags:                                     SettingDefinitionV2DataEngineLogFlags,
+		SettingNameV2DataEngineFastRebuilding:                               SettingDefinitionV2DataEngineFastRebuilding,
 		SettingNameReplicaDiskSoftAntiAffinity:                              SettingDefinitionReplicaDiskSoftAntiAffinity,
 		SettingNameAllowEmptyNodeSelectorVolume:                             SettingDefinitionAllowEmptyNodeSelectorVolume,
 		SettingNameAllowEmptyDiskSelectorVolume:                             SettingDefinitionAllowEmptyDiskSelectorVolume,
@@ -1260,12 +1263,12 @@ var (
 
 	SettingDefinitionFastReplicaRebuildEnabled = SettingDefinition{
 		DisplayName: "Fast Replica Rebuild Enabled",
-		Description: "This setting enables the fast replica rebuilding feature. It relies on the checksums of snapshot disk files, so setting the snapshot-data-integrity to **enable** or **fast-check** is a prerequisite.",
+		Description: "This setting enables the fast replica rebuilding feature, which will try to reuse the existing data on a failed replica rather than directly creating a new replica for a degraded volume.",
 		Category:    SettingCategoryGeneral,
 		Type:        SettingTypeBool,
 		Required:    true,
 		ReadOnly:    false,
-		Default:     "true",
+		Default:     "false",
 	}
 
 	SettingDefinitionReplicaFileSyncHTTPClientTimeout = SettingDefinition{
@@ -1477,6 +1480,16 @@ var (
 		Default:     "",
 	}
 
+	SettingDefinitionV2DataEngineFastRebuilding = SettingDefinition{
+		DisplayName: "V2 Data Engine Fast Rebuilding",
+		Description: "Enable the fast rebuilding feature for V2 Data Engine.",
+		Category:    SettingCategoryV2DataEngine,
+		Type:        SettingTypeBool,
+		Required:    true,
+		ReadOnly:    false,
+		Default:     "false",
+	}
+
 	SettingDefinitionAutoCleanupSnapshotWhenDeleteBackup = SettingDefinition{
 		DisplayName: "Automatically Cleanup Snapshot When Deleting Backup",
 		Description: "This setting enables Longhorn to automatically cleanup snapshots when removing backup.",