Merge pull request #46 from supermeng/master

Make state of pod clearer

Author: supermeng

engine/pod.go

@@ -87,7 +87,7 @@ func (pc *podController) Deploy(cluster cluster.Cluster) {
 		id, err := pc.createContainer(cluster, filters, i)
 		if err != nil {
 			log.Warnf("%s Cannot create container, error=%q, spec=%+v", pc, err, cSpec)
-			pc.pod.State = RunStateFail
+			pc.pod.State = RunStateError
 			pc.pod.LastError = fmt.Sprintf("Cannot create container, %s", err)
 			return
 		}
@@ -187,7 +187,7 @@ func (pc *podController) Stop(cluster cluster.Cluster) {
 	for i, container := range pc.pod.Containers {
 		if err := cluster.StopContainer(container.Id, pc.spec.GetKillTimeout()); err != nil {
 			log.Warnf("%s Cannot stop the container %s, %s", pc, container.Id, err)
-			pc.pod.State = RunStateFail
+			pc.pod.State = RunStateError
 			pc.pod.LastError = fmt.Sprintf("Cannot stop container, %s", err)
 		} else {
 			pc.refreshContainer(cluster, i)
@@ -212,7 +212,7 @@ func (pc *podController) Start(cluster cluster.Cluster) {
 			pc.refreshContainer(cluster, i)
 		} else {
 			log.Warnf("%s Cannot start the container %s, %s", pc, container.Id, err)
-			pc.pod.State = RunStateFail
+			pc.pod.State = RunStateError
 			pc.pod.LastError = fmt.Sprintf("Cannot start container, %s", err)
 		}
 	}
@@ -231,7 +231,7 @@ func (pc *podController) Restart(cluster cluster.Cluster) {
 	for i, container := range pc.pod.Containers {
 		if err := cluster.RestartContainer(container.Id, pc.spec.GetKillTimeout()); err != nil {
 			log.Warnf("%s Cannot restart the container %s, %s", pc, container.Id, err)
-			pc.pod.State = RunStateFail
+			pc.pod.State = RunStateError
 			pc.pod.LastError = fmt.Sprintf("Cannot restart container, %s", err)
 		} else {
 			pc.refreshContainer(cluster, i)
@@ -285,7 +285,7 @@ func (pc *podController) startContainer(cluster cluster.Cluster, id string) erro
 				log.Warnf("%s Cannot release IP %s, %s", pc, id, err)
 			}
 		}
-		pc.pod.State = RunStateFail
+		pc.pod.State = RunStateError
 		pc.pod.LastError = fmt.Sprintf("Cannot start container, %s", err)
 		return err
 	}
@@ -329,7 +329,7 @@ func (pc *podController) refreshContainer(kluster cluster.Cluster, index int) {
 			pc.pod.LastError = fmt.Sprintf("Missing container %q, %s", id, err)
 		} else {
 			log.Warnf("%s Failed to inspect container %s, %s", pc, id, err)
-			pc.pod.State = RunStateFail
+			pc.pod.State = RunStateError
 			pc.pod.LastError = fmt.Sprintf("Cannot inspect the container, %s", err)
 		}
 	} else {

engine/podgroup.go

@@ -297,6 +297,7 @@ func (pgCtrl *podGroupController) Refresh(force bool) {
 	if pgCtrl.IsRemoved() || pgCtrl.IsPending() {
 		return
 	}
+	pgCtrl.emitOperationEvent(OperationStart)
 	pgCtrl.DisableRefresh()
 	defer func() {
 		pgCtrl.opsChan <- pgOperOver{}

engine/podgroup_ops.go

@@ -253,7 +253,7 @@ func (op pgOperRefreshInstance) Do(pgCtrl *podGroupController, c cluster.Cluster
 			runtime = podCtrl.pod.ImRuntime
 			consistent = false
 		}
-	} else if runtime.State == RunStateExit {
+	} else if runtime.State == RunStateExit || runtime.State == RunStateFail {
 		if !pgCtrl.engine.config.Maintenance && podCtrl.pod.NeedRestart(op.spec.RestartPolicy) {
 			if podCtrl.pod.RestartEnoughTimes() {
 				ntfController.Send(NewNotifySpec(podCtrl.spec.Namespace, podCtrl.spec.Name,
@@ -583,18 +583,18 @@ func (op pgOperChangeState) Do(pgCtrl *podGroupController, c cluster.Cluster, st
 	switch op.op {
 	case "start":
 		podCtrl.Start(c)
-		if podCtrl.pod.State != RunStateFail {
+		if podCtrl.pod.State != RunStateError {
 			podCtrl.pod.ChangeTargetState(ExpectStateRun)
 		}
 	case "stop":
 		podCtrl.Stop(c)
-		if podCtrl.pod.State != RunStateFail {
+		if podCtrl.pod.State != RunStateError {
 			podCtrl.pod.ChangeTargetState(ExpectStateStop)
 		}
 	case "restart":
 		pgCtrl.waitLastPodHealth(op.instance)
 		podCtrl.Restart(c)
-		if podCtrl.pod.State != RunStateFail {
+		if podCtrl.pod.State != RunStateError {
 			podCtrl.pod.ChangeTargetState(ExpectStateRun)
 		}
 	}

engine/runtimes.go

@@ -14,15 +14,16 @@ type PGOpState int32
 var RestartMaxCount int
 
 const (
-	RunStatePending = iota
-	RunStateDrift
-	RunStateSuccess
-	RunStateExit
-	RunStateFail
-	RunStateInconsistent
-	RunStateMissing
-	RunStateRemoved
-	RunStatePaused
+	RunStatePending = iota // waiting for operation
+	RunStateDrift          // drifting from one node to another
+	RunStateSuccess        // ok
+	RunStateExit           // exited
+	RunStateFail           // start failed with error
+	RunStateInconsistent   // container's state is different between deployd and swarm
+	RunStateMissing        // container is missing and need create it. happened when node down .etc
+	RunStateRemoved        // removed
+	RunStatePaused         // paused
+	RunStateError          // call docker interface with error
 )
 
 const (
@@ -68,6 +69,8 @@ func (rs RunState) String() string {
 		return "RunStateRemoved"
 	case RunStatePaused:
 		return "RunStatePaused"
+	case RunStateError:
+		return "RunStateError"
 	default:
 		return "Unknown RunState"
 	}