fix: controlling issues

cmd/tcl/testworkflow-toolkit/commands/parallel.go

@@ -263,6 +263,7 @@ func NewParallelCmd() *cobra.Command {
 
 				prevStatus := testkube.QUEUED_TestWorkflowStatus
 				prevStep := ""
+				prevIsFinished := false
 				scheduled := false
 				for v := range ctrl.WatchLightweight(ctx) {
 					// Handle error
@@ -283,14 +284,17 @@ func NewParallelCmd() *cobra.Command {
 					}
 
 					// Handle result change
-					if v.Status != prevStatus || v.Current != prevStep {
+					// TODO: the final status should always have the finishedAt too,
+					//       there should be no need for checking isFinished diff
+					if v.Status != prevStatus || lastResult.IsFinished() != prevIsFinished || v.Current != prevStep {
 						if v.Status != prevStatus {
 							log(string(v.Status))
 						}
 						updates <- Update{index: index, result: v.Result}
 						prevStep = v.Current
 						prevStatus = v.Status
-						if v.Result.IsFinished() {
+						prevIsFinished = lastResult.IsFinished()
+						if lastResult.IsFinished() {
 							instructions.PrintOutput(env.Ref(), "parallel", ParallelStatus{Index: int(index), Status: v.Status, Result: v.Result})
 							ctxCancel()
 							return v.Result.IsPassed()

cmd/testworkflow-init/main.go

@@ -110,6 +110,7 @@ func main() {
 		orchestration.Pause(step, *step.StartedAt)
 		for _, parentRef := range step.Parents {
 			parent := state.GetStep(parentRef)
+			// TODO: What about parents of the parents?
 			orchestration.Pause(parent, *step.StartedAt)
 		}
 		return err
@@ -125,6 +126,7 @@ func main() {
 		orchestration.Resume(step, ts)
 		for _, parentRef := range step.Parents {
 			parent := state.GetStep(parentRef)
+			// TODO: What about parents of the parents?
 			orchestration.Resume(parent, ts)
 		}
 		return err

cmd/testworkflow-init/orchestration/executions.go

@@ -133,6 +133,7 @@ func (e *executionGroup) Kill() (err error) {
 func (e *executionGroup) Abort() {
 	e.aborted.Store(true)
 	_ = e.Kill()
+	_ = e.Resume()
 }
 
 func (e *executionGroup) IsAborted() bool {

pkg/api/v1/testkube/model_test_workflow_result_extended.go

@@ -8,7 +8,7 @@ import (
 )
 
 func (r *TestWorkflowResult) IsFinished() bool {
-	return !r.IsStatus(QUEUED_TestWorkflowStatus) && !r.IsStatus(RUNNING_TestWorkflowStatus) && !r.IsStatus(PAUSED_TestWorkflowStatus)
+	return !r.FinishedAt.IsZero() && !r.IsStatus(QUEUED_TestWorkflowStatus) && !r.IsStatus(RUNNING_TestWorkflowStatus) && !r.IsStatus(PAUSED_TestWorkflowStatus)
 }
 
 func (r *TestWorkflowResult) IsStatus(s TestWorkflowStatus) bool {
@@ -90,7 +90,7 @@ func (r *TestWorkflowResult) Fatal(err error, aborted bool, ts time.Time) {
 	if r.FinishedAt.IsZero() {
 		r.FinishedAt = ts.UTC()
 	}
-	if r.Initialization.Status == nil || (*r.Initialization.Status == QUEUED_TestWorkflowStepStatus) || (*r.Initialization.Status == RUNNING_TestWorkflowStepStatus) {
+	if r.Initialization.Status == nil || !(*r.Initialization.Status).Finished() {
 		r.Initialization.Status = common.Ptr(FAILED_TestWorkflowStepStatus)
 		if aborted {
 			r.Initialization.Status = common.Ptr(ABORTED_TestWorkflowStepStatus)
@@ -158,16 +158,23 @@ func (r *TestWorkflowResult) RecomputeDuration() {
 	if !r.FinishedAt.IsZero() {
 		r.PausedMs = 0
 
+		// Finalize pauses
+		for i := range r.Pauses {
+			step := r.Steps[r.Pauses[i].Ref]
+			if !step.FinishedAt.IsZero() {
+				if r.Pauses[i].ResumedAt.IsZero() {
+					r.Pauses[i].ResumedAt = step.FinishedAt
+				}
+				if r.Pauses[i].PausedAt.Before(step.StartedAt) {
+					r.Pauses[i].PausedAt = step.StartedAt
+				}
+			}
+		}
+
 		// Get unique pause periods
 		pauses := make([]TestWorkflowPause, 0)
 	loop:
 		for _, p := range r.Pauses {
-			// Finalize the pause if it's not
-			step := r.Steps[p.Ref]
-			if !step.FinishedAt.IsZero() && p.ResumedAt.IsZero() {
-				p.ResumedAt = step.FinishedAt
-			}
-
 			for i := range pauses {
 				// They don't overlap
 				if p.PausedAt.After(pauses[i].ResumedAt) || p.ResumedAt.Before(pauses[i].PausedAt) {
@@ -326,6 +333,14 @@ func (r *TestWorkflowResult) Recompute(sig []TestWorkflowSignature, scheduledAt
 		r.Steps[s.ref] = s.result
 	}
 
+	// Ensure initialization timestamps
+	if !r.Initialization.FinishedAt.IsZero() && r.Initialization.StartedAt.IsZero() {
+		r.Initialization.StartedAt = r.Initialization.FinishedAt
+	}
+	if !r.Initialization.StartedAt.IsZero() && r.Initialization.QueuedAt.IsZero() {
+		r.Initialization.QueuedAt = r.Initialization.StartedAt
+	}
+
 	// Calibrate the clock for group steps
 	walkSteps(sig, func(s TestWorkflowSignature) {
 		if len(s.Children) == 0 {
@@ -543,7 +558,16 @@ func recomputeTestWorkflowStepResult(v TestWorkflowStepResult, sig TestWorkflowS
 
 	// Ensure there is a start time if the step is already finished
 	if v.StartedAt.IsZero() && !v.FinishedAt.IsZero() {
-		v.StartedAt = v.QueuedAt
+		if v.QueuedAt.IsZero() {
+			v.StartedAt = v.FinishedAt
+		} else {
+			v.StartedAt = v.QueuedAt
+		}
+	}
+
+	// Ensure there is a queued time if the step is already finished
+	if v.QueuedAt.IsZero() && !v.StartedAt.IsZero() {
+		v.QueuedAt = v.StartedAt
 	}
 
 	// Compute children

pkg/testworkflows/testworkflowcontroller/controller.go

@@ -231,6 +231,7 @@ func (c *controller) WatchLightweight(parentCtx context.Context) <-chan Lightwei
 	prevNodeName := ""
 	prevPodIP := ""
 	prevStatus := testkube.QUEUED_TestWorkflowStatus
+	prevIsFinished := false
 	sig := stage.MapSignatureListToInternal(c.signature)
 	ch := make(chan LightweightNotification)
 	go func() {
@@ -245,20 +246,25 @@ func (c *controller) WatchLightweight(parentCtx context.Context) <-chan Lightwei
 			podIP, _ := c.PodIP(parentCtx)
 			current := prevCurrent
 			status := prevStatus
+			isFinished := prevIsFinished
 			if v.Value.Result != nil {
 				if v.Value.Result.Status != nil {
 					status = *v.Value.Result.Status
 				} else {
 					status = testkube.QUEUED_TestWorkflowStatus
 				}
 				current = v.Value.Result.Current(sig)
+				isFinished = v.Value.Result.IsFinished()
 			}
 
-			if nodeName != prevNodeName || podIP != prevPodIP || prevStatus != status || prevCurrent != current {
+			// TODO: the final status should always have the finishedAt too,
+			//       there should be no need for checking isFinished diff
+			if nodeName != prevNodeName || isFinished != prevIsFinished || podIP != prevPodIP || prevStatus != status || prevCurrent != current {
 				prevNodeName = nodeName
 				prevPodIP = podIP
 				prevStatus = status
 				prevCurrent = current
+				prevIsFinished = isFinished
 				ch <- LightweightNotification{NodeName: nodeName, PodIP: podIP, Status: status, Current: current, Result: v.Value.Result}
 			}
 		}
@@ -271,16 +277,6 @@ func (c *controller) Logs(parentCtx context.Context, follow bool) io.Reader {
 	go func() {
 		defer writer.Close()
 		ref := ""
-		// Wait until there will be events fetched first
-		alignTimeoutCh := time.After(alignmentTimeout)
-		select {
-		case <-c.jobEvents.Peek(parentCtx):
-		case <-alignTimeoutCh:
-		}
-		select {
-		case <-c.podEvents.Peek(parentCtx):
-		case <-alignTimeoutCh:
-		}
 		ch, err := WatchInstrumentedPod(parentCtx, c.clientSet, c.signature, c.scheduledAt, c.pod, c.podEvents, WatchInstrumentedPodOptions{
 			JobEvents: c.jobEvents,
 			Job:       c.job,

pkg/testworkflows/testworkflowcontroller/logs.go

@@ -40,6 +40,23 @@ type ContainerLog struct {
 	Output *instructions.Instruction
 }
 
+type ContainerLogType string
+
+const (
+	ContainerLogTypeHint   ContainerLogType = "hint"
+	ContainerLogTypeOutput ContainerLogType = "output"
+	ContainerLogTypeLog    ContainerLogType = ""
+)
+
+func (c *ContainerLog) Type() ContainerLogType {
+	if c.Hint != nil {
+		return ContainerLogTypeHint
+	} else if c.Output != nil {
+		return ContainerLogTypeOutput
+	}
+	return ContainerLogTypeLog
+}
+
 // getContainerLogsStream is getting logs stream, and tries to reinitialize the stream on EOF.
 // EOF may happen not only on the actual container end, but also in case of the log rotation.
 // @see {@link https://stackoverflow.com/a/68673451}

pkg/testworkflows/testworkflowcontroller/notifier.go

@@ -312,7 +312,7 @@ func (n *notifier) UpdateStepStatus(ref string, status testkube.TestWorkflowStep
 }
 
 func (n *notifier) finishInit(status ContainerResultStep) {
-	if n.result.Initialization.FinishedAt.Equal(status.FinishedAt) && n.result.Initialization.Status != nil && *n.result.Initialization.Status == status.Status {
+	if n.result.Initialization.FinishedAt.Equal(status.FinishedAt) && n.result.Initialization.Status != nil && *n.result.Initialization.Status == status.Status && (status.Status != testkube.ABORTED_TestWorkflowStepStatus || n.result.Initialization.ErrorMessage == status.Details) {
 		return
 	}
 	n.result.Initialization.FinishedAt = status.FinishedAt.UTC()
@@ -352,7 +352,7 @@ func (n *notifier) FinishStep(ref string, status ContainerResultStep) {
 		n.finishInit(status)
 		return
 	}
-	if n.result.Steps[ref].FinishedAt.Equal(status.FinishedAt) && n.result.Steps[ref].Status != nil && *n.result.Steps[ref].Status == status.Status {
+	if n.result.Steps[ref].FinishedAt.Equal(status.FinishedAt) && n.result.Steps[ref].Status != nil && *n.result.Steps[ref].Status == status.Status && (status.Status != testkube.ABORTED_TestWorkflowStepStatus || n.result.Steps[ref].ErrorMessage == status.Details) {
 		return
 	}
 	s := n.result.Steps[ref]