internal/manager/task_state_machine/task_state_machine.go

@@ -5,8 +5,6 @@ package task_state_machine
 import (
 	"context"
 	"fmt"
-	"projects.blender.org/studio/flamenco/pkg/website"
-
 	"github.com/rs/zerolog"
 	"github.com/rs/zerolog/log"
 
@@ -120,6 +118,7 @@ func (sm *StateMachine) updateJobAfterTaskStatusChange(
 		return sm.jobStatusIfAThenB(ctx, logger, job, api.JobStatusCompleted, api.JobStatusRequeueing, "task was queued")
 
 	case api.TaskStatusPaused:
+		// Pausing a task has no impact on the job.
 		return nil
 
 	case api.TaskStatusCanceled:
@@ -214,10 +213,19 @@ func (sm *StateMachine) updateJobOnTaskStatusCompleted(ctx context.Context, logg
 	if err != nil {
 		return err
 	}
+	numActive, _, err := sm.persist.CountTasksOfJobInStatus(ctx, job, api.TaskStatusActive)
+	if err != nil {
+		return err
+	}
 	if numComplete == numTotal {
 		logger.Info().Msg("all tasks of job are completed, job is completed")
 		return sm.JobStatusChange(ctx, job, api.JobStatusCompleted, "all tasks completed")
 	}
+	if numActive == 0 && job.Status == api.JobStatusPauseRequested {
+		// there is no active task, and the job is in pause-requested status, so we can pause the job
+		logger.Info().Msg("all tasks of job are completed, job is paused")
+		return sm.JobStatusChange(ctx, job, api.JobStatusPaused, "all tasks completed")
+	}
 	logger.Info().
 		Int("taskNumTotal", numTotal).
 		Int("taskNumComplete", numComplete).
@@ -453,7 +461,6 @@ func (sm *StateMachine) pauseTasks(
 	// Any task that might run in the future should get paused.
 	// Active jobs should remain active until finished
 	taskStatusesToPause := []api.TaskStatus{
-		api.TaskStatusActive,
 		api.TaskStatusQueued,
 		api.TaskStatusCanceled,
 		api.TaskStatusSoftFailed,
@@ -466,15 +473,7 @@ func (sm *StateMachine) pauseTasks(
 		return "", fmt.Errorf("pausing tasks of job %s: %w", job.UUID, err)
 	}
 
-	// If pause was requested, it has now happened, so the job can transition.
-	if job.Status == api.JobStatusPauseRequested {
-		logger.Info().Msg("all tasks of job paused, job can go to 'paused' status")
-		return api.JobStatusPaused, nil
-	}
-
-	// This could mean state transition entered a non-recoverable error state.
-	log.Warn().Str("jobStatus", string(job.Status)).Msgf("unexpected job status in StateMachine::pauseTasks(), please report this at %s", website.BugReportURL)
-	return "", fmt.Errorf("unexpected job status %q in StateMachine::pauseTasks()", job.Status)
+	return api.JobStatusPauseRequested, nil
 }
 
 // requeueTasks re-queues all tasks of the job.