Pause zero scale Framework instead of completing it (#59)

pkg/apis/frameworkcontroller/v1/types.go

@@ -45,10 +45,11 @@ type FrameworkList struct {
 // 2. Partitioned to different heterogeneous TaskRoles which share the same lifecycle
 // 3. Ordered in the same homogeneous TaskRole by TaskIndex
 // 4. With consistent identity {FrameworkName}-{TaskRoleName}-{TaskIndex} as PodName
-// 5. With fine grained RetryPolicy for each Task and the whole Framework
-// 6. With fine grained FrameworkAttemptCompletionPolicy for each TaskRole
-// 7. With PodGracefulDeletionTimeoutSec for each Task to tune Consistency vs Availability
-// 8. With fine grained Status for each TaskAttempt/Task, each TaskRole and the whole
+// 5. With fine grained ExecutionType to Start/Stop the whole Framework
+// 6. With fine grained RetryPolicy for each Task and the whole Framework
+// 7. With fine grained FrameworkAttemptCompletionPolicy for each TaskRole
+// 8. With PodGracefulDeletionTimeoutSec for each Task to tune Consistency vs Availability
+// 9. With fine grained Status for each TaskAttempt/Task, each TaskRole and the whole
 //    FrameworkAttempt/Framework
 //
 // Notes:
@@ -204,14 +205,14 @@ type RetryPolicySpec struct {
 //    from the Task which triggers the completion.
 // 3. If MinSucceededTaskCount >= 1 and MinSucceededTaskCount <= succeeded Task
 //    count of current TaskRole, immediately complete the FrameworkAttempt, regardless
-//    of any uncompleted Task, and the CompletionStatus is succeeded which is
-//    inherited from the Task which triggers the completion.
+//    of any uncompleted Task, and the CompletionStatus is succeeded which is inherited
+//    from the Task which triggers the completion.
 // 4. If multiple above conditions are satisfied at the same time, the behavior can
 //    be any one of these satisfied conditions.
 // 5. If none of above conditions are satisfied until all Tasks of the Framework are
-//    completed (including a special case that the Framework does even not have any
-//    Task), immediately complete the FrameworkAttempt and the CompletionStatus is
-//    succeeded which is not inherited from any Task.
+//    completed and the Framework has at least one Task, immediately complete the
+//    FrameworkAttempt and the CompletionStatus is succeeded which is not inherited
+//    from any Task.
 //
 // Notes:
 // 1. When the FrameworkAttempt is completed, the FrameworkState is transitioned to

pkg/controller/controller.go

@@ -1535,53 +1535,57 @@ func (c *FrameworkController) syncFrameworkAttemptCompletionPolicy(
 		return true
 	}
 
-	// The Framework must not Completing or Completed, so TaskRoles/Tasks in
-	// f.Spec must fully contain not DeletionPending (ScaleDown) TaskRoles/Tasks
-	// in f.Status, thus completedTaskCount must <= totalTaskCount.
 	totalTaskCount := f.GetTotalTaskCountSpec()
-	completedTaskCount := f.GetTaskCountStatus(completedTaskSelector)
-	if completedTaskCount >= totalTaskCount {
-		var lastCompletedTaskStatus *ci.TaskStatus
-		var lastCompletedTaskRoleName string
-		for _, taskRoleSpec := range f.Spec.TaskRoles {
-			taskRoleName := taskRoleSpec.Name
-			taskRoleStatus := f.GetTaskRoleStatus(taskRoleName)
-			if taskRoleStatus == nil {
-				// Unreachable
-				continue
+	// At least one completed Task is needed to trigger its
+	// FrameworkAttemptCompletionPolicy.
+	if totalTaskCount >= 1 {
+		completedTaskCount := f.GetTaskCountStatus(completedTaskSelector)
+		// The Framework must not Completing or Completed, so TaskRoles/Tasks in
+		// f.Spec must fully contain not DeletionPending (ScaleDown) TaskRoles/Tasks
+		// in f.Status, thus completedTaskCount must <= totalTaskCount.
+		if completedTaskCount >= totalTaskCount {
+			var lastCompletedTaskStatus *ci.TaskStatus
+			var lastCompletedTaskRoleName string
+			for _, taskRoleSpec := range f.Spec.TaskRoles {
+				taskRoleName := taskRoleSpec.Name
+				taskRoleStatus := f.GetTaskRoleStatus(taskRoleName)
+				if taskRoleStatus == nil {
+					// Unreachable
+					continue
+				}
+
+				roleTotalTaskCount := taskRoleSpec.TaskNumber
+				if roleTotalTaskCount == 0 {
+					continue
+				}
+
+				roleLastCompletedTask := taskRoleStatus.CompletionTimeOrderedTaskStatus(
+					completedTaskSelector, roleTotalTaskCount-1)
+
+				if lastCompletedTaskStatus == nil ||
+					roleLastCompletedTask.CompletionTime.Time.After(
+						lastCompletedTaskStatus.CompletionTime.Time) {
+					lastCompletedTaskStatus = roleLastCompletedTask
+					lastCompletedTaskRoleName = taskRoleName
+				}
 			}
 
-			roleTotalTaskCount := taskRoleSpec.TaskNumber
-			if roleTotalTaskCount == 0 {
-				continue
-			}
+			firstTriggerCompletionStatus = ci.NewCompletedTaskTriggeredCompletionStatus(
+				lastCompletedTaskStatus, lastCompletedTaskRoleName,
+				completedTaskCount, totalTaskCount)
 
-			roleLastCompletedTask := taskRoleStatus.CompletionTimeOrderedTaskStatus(
-				completedTaskSelector, roleTotalTaskCount-1)
-
-			if lastCompletedTaskStatus == nil ||
-				roleLastCompletedTask.CompletionTime.Time.After(
-					lastCompletedTaskStatus.CompletionTime.Time) {
-				lastCompletedTaskStatus = roleLastCompletedTask
-				lastCompletedTaskRoleName = taskRoleName
+			if firstTriggerCompletionStatus.Trigger == nil {
+				klog.Infof("[%v]: syncFrameworkAttemptCompletionPolicy: %v", f.Key(),
+					firstTriggerCompletionStatus.Diagnostics)
+			} else {
+				klog.Infof("[%v][%v][%v]: syncFrameworkAttemptCompletionPolicy: %v", f.Key(),
+					firstTriggerCompletionStatus.Trigger.TaskRoleName,
+					firstTriggerCompletionStatus.Trigger.TaskIndex,
+					firstTriggerCompletionStatus.Trigger.Message)
 			}
+			c.completeFrameworkAttempt(f, false, firstTriggerCompletionStatus)
+			return true
 		}
-
-		firstTriggerCompletionStatus = ci.NewCompletedTaskTriggeredCompletionStatus(
-			lastCompletedTaskStatus, lastCompletedTaskRoleName,
-			completedTaskCount, totalTaskCount)
-
-		if firstTriggerCompletionStatus.Trigger == nil {
-			klog.Infof("[%v]: syncFrameworkAttemptCompletionPolicy: %v", f.Key(),
-				firstTriggerCompletionStatus.Diagnostics)
-		} else {
-			klog.Infof("[%v][%v][%v]: syncFrameworkAttemptCompletionPolicy: %v", f.Key(),
-				firstTriggerCompletionStatus.Trigger.TaskRoleName,
-				firstTriggerCompletionStatus.Trigger.TaskIndex,
-				firstTriggerCompletionStatus.Trigger.Message)
-		}
-		c.completeFrameworkAttempt(f, false, firstTriggerCompletionStatus)
-		return true
 	}
 
 	return false
@@ -1998,18 +2002,22 @@ func (c *FrameworkController) syncTaskState(
 			return nil
 		}
 
-		// The Framework must not Completing or Completed, so TaskRoles/Tasks in
-		// f.Spec must fully contain not DeletionPending (ScaleDown) TaskRoles/Tasks
-		// in f.Status, thus completedTaskCount must <= totalTaskCount.
 		totalTaskCount := f.GetTotalTaskCountSpec()
-		completedTaskCount := f.GetTaskCountStatus(completedTaskSelector)
-		if completedTaskCount >= totalTaskCount {
-			triggerCompletionStatus = ci.NewCompletedTaskTriggeredCompletionStatus(
-				taskStatus, taskRoleName, completedTaskCount, totalTaskCount)
+		// At least one completed Task is needed to trigger its
+		// FrameworkAttemptCompletionPolicy.
+		if taskStatus.IsCompleted(true) && totalTaskCount >= 1 {
+			completedTaskCount := f.GetTaskCountStatus(completedTaskSelector)
+			// The Framework must not Completing or Completed, so TaskRoles/Tasks in
+			// f.Spec must fully contain not DeletionPending (ScaleDown) TaskRoles/Tasks
+			// in f.Status, thus completedTaskCount must <= totalTaskCount.
+			if completedTaskCount >= totalTaskCount {
+				triggerCompletionStatus = ci.NewCompletedTaskTriggeredCompletionStatus(
+					taskStatus, taskRoleName, completedTaskCount, totalTaskCount)
 
-			klog.Info(logPfx + triggerCompletionStatus.Trigger.Message)
-			c.completeFrameworkAttempt(f, false, triggerCompletionStatus)
-			return nil
+				klog.Info(logPfx + triggerCompletionStatus.Trigger.Message)
+				c.completeFrameworkAttempt(f, false, triggerCompletionStatus)
+				return nil
+			}
 		}
 
 		return nil