feat: re-setup master when master role restarted or recreated and rec…

…over dataset

Signed-off-by: SimonCqk <[email protected]>

Makefile

@@ -161,13 +161,13 @@ csi-build:
 	CGO_ENABLED=0 GOOS=linux GOARCH=${ARCH} GO111MODULE=${GO_MODULE}  go build ${GC_FLAGS} -a -o bin/fluid-csi -ldflags '${LDFLAGS}' cmd/csi/main.go
 
 dataset-controller-build:
-	CGO_ENABLED=0 GOOS=linux GOARCH=${ARCH} GO111MODULE=${GO_MODULE}  go build ${GC_FLAGS} -a -o bin/dataset-controller -ldflags '${LDFLAGS}' cmd/dataset/main.go
+	CGO_ENABLED=0 GOOS=linux GOARCH=amd64 GO111MODULE=${GO_MODULE}  go build ${GC_FLAGS} -a -o bin/dataset-controller -ldflags '${LDFLAGS}' cmd/dataset/main.go
 
 alluxioruntime-controller-build:
 	CGO_ENABLED=0 GOOS=linux GOARCH=${ARCH} GO111MODULE=${GO_MODULE}  go build ${GC_FLAGS} -a -o bin/alluxioruntime-controller -ldflags '${LDFLAGS}' cmd/alluxio/main.go
 
 jindoruntime-controller-build:
-	CGO_ENABLED=0 GOOS=linux GOARCH=${ARCH} GO111MODULE=${GO_MODULE}  go build ${GC_FLAGS} -a -o bin/jindoruntime-controller -ldflags '${LDFLAGS}' cmd/jindo/main.go
+	CGO_ENABLED=0 GOOS=linux GOARCH=amd64 GO111MODULE=${GO_MODULE}  go build ${GC_FLAGS} -a -o bin/jindoruntime-controller -ldflags '${LDFLAGS}' cmd/jindo/main.go
 
 goosefsruntime-controller-build:
 	CGO_ENABLED=0 GOOS=linux GOARCH=${ARCH} GO111MODULE=${GO_MODULE}  go build ${GC_FLAGS} -a -o bin/goosefsruntime-controller -ldflags '${LDFLAGS}' cmd/goosefs/main.go

docker/Dockerfile.dataset

@@ -19,7 +19,7 @@ RUN apk add --update curl tzdata iproute2 bash libc6-compat vim &&  \
  	echo "Asia/Shanghai" >  /etc/timezone
 
 ARG TARGETARCH
-RUN curl -o helm-v3.11.3-linux-${TARGETARCH}.tar.gz https://get.helm.sh/helm-v3.11.3-linux-${TARGETARCH}.tar.gz && \
+RUN export HTTPS_PROXY=socks5://host.docker.internal:13659 && curl -o helm-v3.11.3-linux-${TARGETARCH}.tar.gz https://get.helm.sh/helm-v3.11.3-linux-${TARGETARCH}.tar.gz && \
     tar -xvf helm-v3.11.3-linux-${TARGETARCH}.tar.gz && \
     mv linux-${TARGETARCH}/helm /usr/local/bin/ddc-helm && \
     chmod u+x /usr/local/bin/ddc-helm && \

docker/Dockerfile.jindoruntime

@@ -16,7 +16,7 @@ RUN apk add --update curl tzdata iproute2 bash libc6-compat vim &&  \
  	echo "Asia/Shanghai" >  /etc/timezone
 
 ARG TARGETARCH
-RUN curl -o helm-v3.11.3-linux-${TARGETARCH}.tar.gz https://get.helm.sh/helm-v3.11.3-linux-${TARGETARCH}.tar.gz && \
+RUN export HTTPS_PROXY=socks5://host.docker.internal:13659 && curl -o helm-v3.11.3-linux-${TARGETARCH}.tar.gz https://get.helm.sh/helm-v3.11.3-linux-${TARGETARCH}.tar.gz && \
     tar -xvf helm-v3.11.3-linux-${TARGETARCH}.tar.gz && \
     mv linux-${TARGETARCH}/helm /usr/local/bin/ddc-helm && \
     chmod u+x /usr/local/bin/ddc-helm && \

pkg/common/constants.go

@@ -205,6 +205,10 @@ const (
 	EnvServerlessPlatformVal        = "VALUE_SERVERLESS_PLATFORM"
 	EnvDisableApplicationController = "KEY_DISABLE_APP_CONTROLLER"
 	EnvImagePullSecretsKey          = "IMAGE_PULL_SECRETS"
+
+	// AnnotationLatestMasterStartedTime annotates latest master containers start-at timestamp
+	// to identify master restarted or recreated by accidentally.
+	AnnotationLatestMasterStartedTime = "fluid.io/master-latest-started-at"
 )
 
 const (

pkg/ctrl/master.go

@@ -20,18 +20,21 @@ import (
 	"context"
 	"fmt"
 	"reflect"
+	"time"
 
-	datav1alpha1 "github.com/fluid-cloudnative/fluid/api/v1alpha1"
-	"github.com/fluid-cloudnative/fluid/pkg/ddc/base"
-	"github.com/fluid-cloudnative/fluid/pkg/utils"
-	"github.com/fluid-cloudnative/fluid/pkg/utils/kubeclient"
 	appsv1 "k8s.io/api/apps/v1"
 	corev1 "k8s.io/api/core/v1"
-	"k8s.io/client-go/tools/record"
-
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/labels"
 	"k8s.io/apimachinery/pkg/types"
+	"k8s.io/client-go/tools/record"
+
+	datav1alpha1 "github.com/fluid-cloudnative/fluid/api/v1alpha1"
+	"github.com/fluid-cloudnative/fluid/pkg/common"
+	"github.com/fluid-cloudnative/fluid/pkg/ddc/base"
+	"github.com/fluid-cloudnative/fluid/pkg/utils"
+	"github.com/fluid-cloudnative/fluid/pkg/utils/kubeclient"
+	"github.com/fluid-cloudnative/fluid/pkg/utils/patch"
 )
 
 // CheckMasterHealthy checks the sts healthy with role
@@ -53,6 +56,26 @@ func (e *Helper) CheckMasterHealthy(recorder record.EventRecorder, runtime base.
 	}
 
 	if healthy {
+		realHealthy, readyTime, err := e.recheckMasterHealthyByEachContainerStartedTime(sts)
+		if err != nil {
+			e.log.Error(err, "failed to recheck master healthy by each container started time")
+			return err
+		}
+
+		if !readyTime.IsZero() && sts.Annotations[common.AnnotationLatestMasterStartedTime] != readyTime.Format(time.RFC3339) {
+			toPatch := patch.NewStrategicPatch().InsertAnnotation(common.AnnotationLatestMasterStartedTime, readyTime.Format(time.RFC3339))
+			if err = e.client.Patch(context.Background(), sts, toPatch); err != nil {
+				return err
+			}
+		}
+
+		if !realHealthy {
+			return fmt.Errorf("the master %s in %s is not ready. Master pod must have restarted since last ready time",
+				sts.Name,
+				sts.Namespace,
+			)
+		}
+
 		cond := utils.NewRuntimeCondition(datav1alpha1.RuntimeMasterReady, "The master is ready.",
 			"The master is ready.", corev1.ConditionTrue)
 		_, oldCond := utils.GetRuntimeCondition(statusToUpdate.Conditions, cond.Type)
@@ -116,3 +139,60 @@ func (e *Helper) CheckMasterHealthy(recorder record.EventRecorder, runtime base.
 	return
 
 }
+
+func (e *Helper) StopBindDataSetInHealthCheck() bool {
+	ds, err := utils.GetDataset(e.client, e.runtimeInfo.GetName(), e.runtimeInfo.GetNamespace())
+	if err != nil {
+		return false
+	}
+	// When current dataset is Failed, we expect to re-execute `Setup` in
+	// runtime controller to recover the corrupted dataset, and runtime controller
+	// will update dataset as Bound after successfully setup, that is, health check
+	// does not do the bind operation to dataset.
+	return ds.Status.Phase == datav1alpha1.FailedDatasetPhase
+}
+
+func (e *Helper) recheckMasterHealthyByEachContainerStartedTime(masterSts *appsv1.StatefulSet) (ready bool, readyTime metav1.Time, err error) {
+	lastLatest := time.Time{}
+
+	if len(masterSts.Annotations[common.AnnotationLatestMasterStartedTime]) > 0 {
+		lastLatest, err = time.Parse(time.RFC3339, masterSts.Annotations[common.AnnotationLatestMasterStartedTime])
+		if err != nil {
+			e.log.Error(err, "failed to parse last latest master started time from annotation",
+				"value", masterSts.Annotations[common.AnnotationLatestMasterStartedTime])
+			return true, metav1.Time{}, err
+		}
+	}
+
+	selector, err := metav1.LabelSelectorAsSelector(masterSts.Spec.Selector)
+	if err != nil {
+		return false, metav1.Time{}, fmt.Errorf("error converting StatefulSet %s in namespace %s selector: %v", masterSts.Name, masterSts.Namespace, err)
+	}
+	pods, err := kubeclient.GetPodsForStatefulSet(e.client, masterSts, selector)
+	if err != nil {
+		return false, metav1.Time{}, err
+	}
+	if len(pods) < int(*masterSts.Spec.Replicas) {
+		return false, metav1.Time{}, nil
+	}
+
+	latest := metav1.Time{}
+	for pi := range pods {
+		curLatest := findLatestContainersStartedTime(&pods[pi])
+		if curLatest.After(latest.Time) {
+			latest = curLatest
+		}
+	}
+	return !lastLatest.IsZero() && lastLatest.Equal(latest.Time), latest, nil
+}
+
+func findLatestContainersStartedTime(pod *corev1.Pod) metav1.Time {
+	latest := metav1.Time{}
+	for csi := range pod.Status.ContainerStatuses {
+		cs := &pod.Status.ContainerStatuses[csi]
+		if cs.State.Running != nil && cs.State.Running.StartedAt.After(latest.Time) {
+			latest = cs.State.Running.StartedAt
+		}
+	}
+	return latest
+}