diff --git a/apis/apps/v1alpha1/opsrequest_types.go b/apis/apps/v1alpha1/opsrequest_types.go
index 915e4169806..410641b08e3 100644
--- a/apis/apps/v1alpha1/opsrequest_types.go
+++ b/apis/apps/v1alpha1/opsrequest_types.go
@@ -175,9 +175,9 @@ type ComponentOps struct {
type RebuildInstance struct {
ComponentOps `json:",inline"`
- // Defines the names of the instances that need to be rebuilt. These are essentially the names of the pods.
+ // Defines the instances that need to be rebuilt.
// +kubebuilder:validation:Required
- InstanceNames []string `json:"instanceNames"`
+ Instances []Instance `json:"instances"`
// Indicates the name of the backup from which to recover. Currently, only a full physical backup is supported
// unless your component only has one replica. Such as 'xtrabackup' is full physical backup for mysql and 'mysqldump' is not.
@@ -195,6 +195,17 @@ type RebuildInstance struct {
EnvForRestore []corev1.EnvVar `json:"envForRestore,omitempty" patchStrategy:"merge" patchMergeKey:"name"`
}
+type Instance struct {
+ // Pod name of the instance.
+ // +kubebuilder:validation:Required
+ Name string `json:"name"`
+
+ // The instance will rebuild on the specified node when the instance uses local PersistentVolume as the storage disk.
+ // If not set, it will rebuild on a random node.
+ // +optional
+ TargetNodeName string `json:"targetNodeName,omitempty"`
+}
+
type Switchover struct {
ComponentOps `json:",inline"`
diff --git a/apis/apps/v1alpha1/zz_generated.deepcopy.go b/apis/apps/v1alpha1/zz_generated.deepcopy.go
index cfa4336bccf..dd42372f698 100644
--- a/apis/apps/v1alpha1/zz_generated.deepcopy.go
+++ b/apis/apps/v1alpha1/zz_generated.deepcopy.go
@@ -3604,6 +3604,21 @@ func (in *IniConfig) DeepCopy() *IniConfig {
return out
}
+// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
+func (in *Instance) DeepCopyInto(out *Instance) {
+ *out = *in
+}
+
+// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new Instance.
+func (in *Instance) DeepCopy() *Instance {
+ if in == nil {
+ return nil
+ }
+ out := new(Instance)
+ in.DeepCopyInto(out)
+ return out
+}
+
// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
func (in *InstanceTemplate) DeepCopyInto(out *InstanceTemplate) {
*out = *in
@@ -4998,9 +5013,9 @@ func (in *RSMSpec) DeepCopy() *RSMSpec {
func (in *RebuildInstance) DeepCopyInto(out *RebuildInstance) {
*out = *in
out.ComponentOps = in.ComponentOps
- if in.InstanceNames != nil {
- in, out := &in.InstanceNames, &out.InstanceNames
- *out = make([]string, len(*in))
+ if in.Instances != nil {
+ in, out := &in.Instances, &out.Instances
+ *out = make([]Instance, len(*in))
copy(*out, *in)
}
if in.EnvForRestore != nil {
diff --git a/config/crd/bases/apps.kubeblocks.io_opsrequests.yaml b/config/crd/bases/apps.kubeblocks.io_opsrequests.yaml
index 8a43f7c909a..225e4f58cf6 100644
--- a/config/crd/bases/apps.kubeblocks.io_opsrequests.yaml
+++ b/config/crd/bases/apps.kubeblocks.io_opsrequests.yaml
@@ -506,15 +506,26 @@ spec:
type: object
type: array
x-kubernetes-preserve-unknown-fields: true
- instanceNames:
- description: Defines the names of the instances that need to
- be rebuilt. These are essentially the names of the pods.
+ instances:
+ description: Defines the instances that need to be rebuilt.
items:
- type: string
+ properties:
+ name:
+ description: Pod name of the instance.
+ type: string
+ targetNodeName:
+ description: The instance will rebuild on the specified
+ node when the instance uses local PersistentVolume as
+ the storage disk. If not set, it will rebuild on a random
+ node.
+ type: string
+ required:
+ - name
+ type: object
type: array
required:
- componentName
- - instanceNames
+ - instances
type: object
type: array
x-kubernetes-list-map-keys:
diff --git a/controllers/apps/operations/rebuild_instance.go b/controllers/apps/operations/rebuild_instance.go
index abd04a0e18f..74cd403aa3c 100644
--- a/controllers/apps/operations/rebuild_instance.go
+++ b/controllers/apps/operations/rebuild_instance.go
@@ -58,6 +58,7 @@ type instanceHelper struct {
comp *appsv1alpha1.ClusterComponentSpec
targetPod *corev1.Pod
backup *dpv1alpha1.Backup
+ instance appsv1alpha1.Instance
actionSet *dpv1alpha1.ActionSet
// key: source pvc name, value: the tmp pvc which using to rebuild
pvcMap map[string]*corev1.PersistentVolumeClaim
@@ -106,9 +107,9 @@ func (r rebuildInstanceOpsHandler) Action(reqCtx intctrlutil.RequestCtx, cli cli
if err != nil {
return err
}
- for _, podName := range v.InstanceNames {
+ for _, ins := range v.Instances {
targetPod := &corev1.Pod{}
- if err = cli.Get(reqCtx.Ctx, client.ObjectKey{Name: podName, Namespace: opsRes.Cluster.Namespace}, targetPod); err != nil {
+ if err = cli.Get(reqCtx.Ctx, client.ObjectKey{Name: ins.Name, Namespace: opsRes.Cluster.Namespace}, targetPod); err != nil {
return err
}
isAvailable, err := r.instanceIsAvailable(synthesizedComp, targetPod)
@@ -116,7 +117,7 @@ func (r rebuildInstanceOpsHandler) Action(reqCtx intctrlutil.RequestCtx, cli cli
return err
}
if isAvailable {
- return intctrlutil.NewFatalError(fmt.Sprintf(`instance "%s" is availabled, can not rebuild it`, podName))
+ return intctrlutil.NewFatalError(fmt.Sprintf(`instance "%s" is availabled, can not rebuild it`, ins.Name))
}
}
}
@@ -127,13 +128,13 @@ func (r rebuildInstanceOpsHandler) SaveLastConfiguration(reqCtx intctrlutil.Requ
return nil
}
-func (r rebuildInstanceOpsHandler) getInstanceProgressDetail(compStatus appsv1alpha1.OpsRequestComponentStatus, instance string) *appsv1alpha1.ProgressStatusDetail {
+func (r rebuildInstanceOpsHandler) getInstanceProgressDetail(compStatus appsv1alpha1.OpsRequestComponentStatus, instance string) appsv1alpha1.ProgressStatusDetail {
objectKey := getProgressObjectKey(constant.PodKind, instance)
progressDetail := findStatusProgressDetail(compStatus.ProgressDetails, objectKey)
if progressDetail != nil {
- return progressDetail
+ return *progressDetail
}
- return &appsv1alpha1.ProgressStatusDetail{
+ return appsv1alpha1.ProgressStatusDetail{
ObjectKey: objectKey,
Status: appsv1alpha1.ProcessingProgressStatus,
Message: fmt.Sprintf("Start to rebuild pod %s", instance),
@@ -156,9 +157,9 @@ func (r rebuildInstanceOpsHandler) ReconcileAction(reqCtx intctrlutil.RequestCtx
for _, v := range opsRes.OpsRequest.Spec.RebuildFrom {
compStatus := opsRes.OpsRequest.Status.Components[v.ComponentName]
comp := opsRes.Cluster.Spec.GetComponentByName(v.ComponentName)
- for i, instance := range v.InstanceNames {
+ for i, instance := range v.Instances {
expectCount += 1
- progressDetail := r.getInstanceProgressDetail(compStatus, instance)
+ progressDetail := r.getInstanceProgressDetail(compStatus, instance.Name)
if isCompletedProgressStatus(progressDetail.Status) {
completedCount += 1
if progressDetail.Status == appsv1alpha1.FailedProgressStatus {
@@ -167,11 +168,11 @@ func (r rebuildInstanceOpsHandler) ReconcileAction(reqCtx intctrlutil.RequestCtx
continue
}
// rebuild instance
- completed, err := r.rebuildInstance(reqCtx, cli, opsRes, comp, v.EnvForRestore, progressDetail, instance, v.BackupName, i)
+ completed, err := r.rebuildInstance(reqCtx, cli, opsRes, comp, v.EnvForRestore, &progressDetail, instance, v.BackupName, i)
if intctrlutil.IsTargetError(err, intctrlutil.ErrorTypeFatal) {
// If a fatal error occurs, this instance rebuilds failed.
progressDetail.SetStatusAndMessage(appsv1alpha1.FailedProgressStatus, err.Error())
- setComponentStatusProgressDetail(opsRes.Recorder, opsRes.OpsRequest, &compStatus.ProgressDetails, *progressDetail)
+ setComponentStatusProgressDetail(opsRes.Recorder, opsRes.OpsRequest, &compStatus.ProgressDetails, progressDetail)
continue
}
if err != nil {
@@ -180,9 +181,9 @@ func (r rebuildInstanceOpsHandler) ReconcileAction(reqCtx intctrlutil.RequestCtx
if completed {
// if the pod has been rebuilt, set progressDetail phase to Succeed.
progressDetail.SetStatusAndMessage(appsv1alpha1.SucceedProgressStatus,
- fmt.Sprintf("Rebuild pod %s successfully", instance))
+ fmt.Sprintf("Rebuild pod %s successfully", instance.Name))
}
- setComponentStatusProgressDetail(opsRes.Recorder, opsRes.OpsRequest, &compStatus.ProgressDetails, *progressDetail)
+ setComponentStatusProgressDetail(opsRes.Recorder, opsRes.OpsRequest, &compStatus.ProgressDetails, progressDetail)
}
opsRes.OpsRequest.Status.Components[v.ComponentName] = compStatus
}
@@ -206,10 +207,10 @@ func (r rebuildInstanceOpsHandler) rebuildInstance(reqCtx intctrlutil.RequestCtx
comp *appsv1alpha1.ClusterComponentSpec,
envForRestore []corev1.EnvVar,
progressDetail *appsv1alpha1.ProgressStatusDetail,
- targetPodName,
+ instance appsv1alpha1.Instance,
backupName string,
index int) (bool, error) {
- insHelper, err := r.prepareInstanceHelper(reqCtx, cli, opsRes, comp, envForRestore, targetPodName, backupName, index)
+ insHelper, err := r.prepareInstanceHelper(reqCtx, cli, opsRes, comp, envForRestore, instance, backupName, index)
if err != nil {
return false, err
}
@@ -224,7 +225,7 @@ func (r rebuildInstanceOpsHandler) prepareInstanceHelper(reqCtx intctrlutil.Requ
opsRes *OpsResource,
comp *appsv1alpha1.ClusterComponentSpec,
envForRestore []corev1.EnvVar,
- targetPodName,
+ instance appsv1alpha1.Instance,
backupName string,
index int) (*instanceHelper, error) {
var (
@@ -253,7 +254,7 @@ func (r rebuildInstanceOpsHandler) prepareInstanceHelper(reqCtx intctrlutil.Requ
}
}
targetPod := &corev1.Pod{}
- if err = cli.Get(reqCtx.Ctx, client.ObjectKey{Name: targetPodName, Namespace: opsRes.Cluster.Namespace}, targetPod); err != nil {
+ if err = cli.Get(reqCtx.Ctx, client.ObjectKey{Name: instance.Name, Namespace: opsRes.Cluster.Namespace}, targetPod); err != nil {
return nil, err
}
synthesizedComp, err := component.BuildSynthesizedComponentWrapper(reqCtx, cli, opsRes.Cluster, comp)
@@ -269,6 +270,7 @@ func (r rebuildInstanceOpsHandler) prepareInstanceHelper(reqCtx intctrlutil.Requ
index: index,
comp: comp,
backup: backup,
+ instance: instance,
actionSet: actionSet,
synthesizedComp: synthesizedComp,
pvcMap: pvcMap,
@@ -330,7 +332,6 @@ func (r rebuildInstanceOpsHandler) rebuildInstanceWithBackup(reqCtx intctrlutil.
if err != nil || !available {
return false, err
}
- progressDetail.Message = fmt.Sprintf(`Waiting for Restore "%s" to be completed`, restoreName)
return false, r.createPostReadyRestore(reqCtx, cli, opsRes.OpsRequest, insHelper, restoreName)
}
return false, r.createPrepareDataRestore(reqCtx, cli, opsRes.OpsRequest, insHelper, restoreName)
@@ -338,7 +339,11 @@ func (r rebuildInstanceOpsHandler) rebuildInstanceWithBackup(reqCtx intctrlutil.
if restore.Status.Phase == dpv1alpha1.RestorePhaseFailed {
return false, intctrlutil.NewFatalError(fmt.Sprintf(`pod "%s" rebuild failed, due to the Restore "%s" is Failed`, insHelper.targetPod.Name, restoreName))
}
- return restore.Status.Phase == dpv1alpha1.RestorePhaseCompleted, nil
+ if restore.Status.Phase != dpv1alpha1.RestorePhaseCompleted {
+ progressDetail.Message = fmt.Sprintf(`Waiting for Restore "%s" to be completed`, restoreName)
+ return false, nil
+ }
+ return true, nil
}
var (
@@ -488,6 +493,16 @@ func (r rebuildInstanceOpsHandler) createPrepareDataRestore(reqCtx intctrlutil.R
}
volumeClaims = append(volumeClaims, volumeClaim)
}
+ schedulePolicy := dpv1alpha1.SchedulingSpec{
+ Tolerations: insHelper.targetPod.Spec.Tolerations,
+ Affinity: insHelper.targetPod.Spec.Affinity,
+ TopologySpreadConstraints: insHelper.targetPod.Spec.TopologySpreadConstraints,
+ }
+ if insHelper.instance.TargetNodeName != "" {
+ schedulePolicy.NodeSelector = map[string]string{
+ corev1.LabelHostname: insHelper.instance.TargetNodeName,
+ }
+ }
restore := &dpv1alpha1.Restore{
ObjectMeta: r.buildRestoreMetaObject(opsRequest, restoreName),
Spec: dpv1alpha1.RestoreSpec{
@@ -497,11 +512,7 @@ func (r rebuildInstanceOpsHandler) createPrepareDataRestore(reqCtx intctrlutil.R
},
Env: insHelper.envForRestore,
PrepareDataConfig: &dpv1alpha1.PrepareDataConfig{
- SchedulingSpec: dpv1alpha1.SchedulingSpec{
- Tolerations: insHelper.targetPod.Spec.Tolerations,
- Affinity: insHelper.targetPod.Spec.Affinity,
- TopologySpreadConstraints: insHelper.targetPod.Spec.TopologySpreadConstraints,
- },
+ SchedulingSpec: schedulePolicy,
VolumeClaimRestorePolicy: dpv1alpha1.VolumeClaimRestorePolicySerial,
RestoreVolumeClaims: volumeClaims,
},
@@ -571,14 +582,20 @@ func (r rebuildInstanceOpsHandler) createTmpPVCsAndPod(reqCtx intctrlutil.Reques
VolumeMounts: insHelper.volumeMounts,
}
intctrlutil.InjectZeroResourcesLimitsIfEmpty(container)
- rebuildPod := builder.NewPodBuilder(insHelper.targetPod.Namespace, tmpPodName).AddTolerations(insHelper.targetPod.Spec.Tolerations...).
+ rebuildPodBuilder := builder.NewPodBuilder(insHelper.targetPod.Namespace, tmpPodName).AddTolerations(insHelper.targetPod.Spec.Tolerations...).
AddContainer(*container).
AddVolumes(insHelper.volumes...).
SetRestartPolicy(corev1.RestartPolicyNever).
AddLabels(constant.OpsRequestNameLabelKey, opsRequest.Name).
AddLabels(constant.OpsRequestNamespaceLabelKey, opsRequest.Namespace).
SetTopologySpreadConstraints(insHelper.targetPod.Spec.TopologySpreadConstraints).
- SetAffinity(insHelper.targetPod.Spec.Affinity).GetObject()
+ SetAffinity(insHelper.targetPod.Spec.Affinity)
+ if insHelper.instance.TargetNodeName != "" {
+ rebuildPodBuilder.SetNodeSelector(map[string]string{
+ corev1.LabelHostname: insHelper.instance.TargetNodeName,
+ })
+ }
+ rebuildPod := rebuildPodBuilder.GetObject()
_ = intctrlutil.SetControllerReference(opsRequest, rebuildPod)
return client.IgnoreAlreadyExists(cli.Create(reqCtx.Ctx, rebuildPod))
}
@@ -752,7 +769,7 @@ func (r rebuildInstanceOpsHandler) instanceIsAvailable(
}
isFailed, isTimeout, _ := intctrlutil.IsPodFailedAndTimedOut(targetPod)
if isFailed && isTimeout {
- return false, intctrlutil.NewFatalError(fmt.Sprintf(`create pod "%s" failed`, targetPod.Name))
+ return false, intctrlutil.NewFatalError(fmt.Sprintf(`the new instance "%s" is failed, please check it`, targetPod.Name))
}
if !podutils.IsPodAvailable(targetPod, synthesizedComp.MinReadySeconds, metav1.Now()) {
return false, nil
diff --git a/controllers/apps/operations/rebuild_instance_test.go b/controllers/apps/operations/rebuild_instance_test.go
index 1a4552ea24f..798ccce1836 100644
--- a/controllers/apps/operations/rebuild_instance_test.go
+++ b/controllers/apps/operations/rebuild_instance_test.go
@@ -78,11 +78,17 @@ var _ = Describe("OpsUtil functions", func() {
opsName := "rebuild-instance-" + testCtx.GetRandomStr()
ops := testapps.NewOpsRequestObj(opsName, testCtx.DefaultNamespace,
clusterName, appsv1alpha1.RebuildInstanceType)
+ var instances []appsv1alpha1.Instance
+ for _, insName := range instanceNames {
+ instances = append(instances, appsv1alpha1.Instance{
+ Name: insName,
+ })
+ }
ops.Spec.RebuildFrom = []appsv1alpha1.RebuildInstance{
{
- ComponentOps: appsv1alpha1.ComponentOps{ComponentName: consensusComp},
- InstanceNames: instanceNames,
- BackupName: backupName,
+ ComponentOps: appsv1alpha1.ComponentOps{ComponentName: consensusComp},
+ Instances: instances,
+ BackupName: backupName,
},
}
opsRequest := testapps.CreateOpsRequest(ctx, testCtx, ops)
@@ -170,8 +176,8 @@ var _ = Describe("OpsUtil functions", func() {
By("fake pod is unavailable")
opsRes.OpsRequest.Status.Phase = appsv1alpha1.OpsCreatingPhase
- for _, podName := range opsRes.OpsRequest.Spec.RebuildFrom[0].InstanceNames {
- Expect(testapps.GetAndChangeObjStatus(&testCtx, client.ObjectKey{Name: podName, Namespace: opsRes.OpsRequest.Namespace}, func(pod *corev1.Pod) {
+ for _, ins := range opsRes.OpsRequest.Spec.RebuildFrom[0].Instances {
+ Expect(testapps.GetAndChangeObjStatus(&testCtx, client.ObjectKey{Name: ins.Name, Namespace: opsRes.OpsRequest.Namespace}, func(pod *corev1.Pod) {
pod.Status.Conditions = nil
})()).Should(Succeed())
}
diff --git a/deploy/helm/crds/apps.kubeblocks.io_opsrequests.yaml b/deploy/helm/crds/apps.kubeblocks.io_opsrequests.yaml
index 8a43f7c909a..225e4f58cf6 100644
--- a/deploy/helm/crds/apps.kubeblocks.io_opsrequests.yaml
+++ b/deploy/helm/crds/apps.kubeblocks.io_opsrequests.yaml
@@ -506,15 +506,26 @@ spec:
type: object
type: array
x-kubernetes-preserve-unknown-fields: true
- instanceNames:
- description: Defines the names of the instances that need to
- be rebuilt. These are essentially the names of the pods.
+ instances:
+ description: Defines the instances that need to be rebuilt.
items:
- type: string
+ properties:
+ name:
+ description: Pod name of the instance.
+ type: string
+ targetNodeName:
+ description: The instance will rebuild on the specified
+ node when the instance uses local PersistentVolume as
+ the storage disk. If not set, it will rebuild on a random
+ node.
+ type: string
+ required:
+ - name
+ type: object
type: array
required:
- componentName
- - instanceNames
+ - instances
type: object
type: array
x-kubernetes-list-map-keys:
diff --git a/docker/Dockerfile-tools b/docker/Dockerfile-tools
index 23821f69230..d55238cac3a 100644
--- a/docker/Dockerfile-tools
+++ b/docker/Dockerfile-tools
@@ -80,7 +80,7 @@ RUN GRPC_HEALTH_PROBE_VERSION=v0.4.13 GOOS=${TARGETOS} GOARCH=${TARGETARCH} &&
# Use alpine with tag 20230329 is corresponding to "edge" tag (latest release to date is 3.18) as of 20230625
-FROM docker.io/alpine:edge as dist
+FROM docker.io/alpine:3.19.1 as dist
ARG APK_MIRROR
# install tools via apk
diff --git a/docs/developer_docs/api-reference/cluster.md b/docs/developer_docs/api-reference/cluster.md
index 83df35d6bc5..62b15edfa30 100644
--- a/docs/developer_docs/api-reference/cluster.md
+++ b/docs/developer_docs/api-reference/cluster.md
@@ -11412,6 +11412,47 @@ string
+Instance
+
+
+(Appears on:RebuildInstance)
+
+
+
+
+
+
+| Field |
+Description |
+
+
+
+
+
+name
+
+string
+
+ |
+
+ Pod name of the instance.
+ |
+
+
+
+targetNodeName
+
+string
+
+ |
+
+(Optional)
+ The instance will rebuild on the specified node when the instance uses local PersistentVolume as the storage disk.
+If not set, it will rebuild on a random node.
+ |
+
+
+
InstanceTemplate
@@ -15143,13 +15184,15 @@ ComponentOps
-instanceNames
+instances
-[]string
+
+[]Instance
+
|
- Defines the names of the instances that need to be rebuilt. These are essentially the names of the pods.
+Defines the instances that need to be rebuilt.
|