studio/flamenco
studio/flamenco
12
30
Fork 38
Code Issues 63 Pull Requests 5 Releases Wiki Activity

WIP: Allow jobs to be submitted in paused status #104318

Closed
David Zhang wants to merge 32 commits from David-Zhang-10/flamenco:submit-as-paused into main
pull from: David-Zhang-10/flamenco:submit-as-paused
merge into: studio:main

When changing the target branch, be careful to rebase the branch in your fork to match. See documentation.
Conversation 0 Commits 32 Files Changed 23 +1 -33
1 changed files with 1 additions and 33 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files
Showing only changes of commit 716f86c0f4 - Show all commits

34
internal/manager/task_state_machine/task_state_machine.go
View File

@ -119,7 +119,7 @@ func (sm *StateMachine) updateJobAfterTaskStatusChange(
return sm.jobStatusIfAThenB(ctx, logger, job, api.JobStatusCompleted, api.JobStatusRequeueing, "task was queued") return sm.jobStatusIfAThenB(ctx, logger, job, api.JobStatusCompleted, api.JobStatusRequeueing, "task was queued")
case api.TaskStatusPaused: case api.TaskStatusPaused:
return sm.updateJobOnTaskStatusPaused(ctx, logger, job) return nil
case api.TaskStatusCanceled: case api.TaskStatusCanceled:
return sm.updateJobOnTaskStatusCanceled(ctx, logger, job) return sm.updateJobOnTaskStatusCanceled(ctx, logger, job)
@ -182,38 +182,6 @@ func (sm *StateMachine) updateJobOnTaskStatusCanceled(ctx context.Context, logge
return nil return nil
} }
// updateJobOnTaskStatusPaused conditionally escalates the pausing of a task to pause the job.
func (sm *StateMachine) updateJobOnTaskStatusPaused(ctx context.Context, logger zerolog.Logger, job *persistence.Job) error {
// If no more tasks can run, pause the job.
numRunnable, _, err := sm.persist.CountTasksOfJobInStatus(ctx, job,
api.TaskStatusActive, api.TaskStatusQueued, api.TaskStatusSoftFailed)
if err != nil {
return err
}
if numRunnable == 0 {
logger.Info().Msg("paused task was last runnable task of job, pausing job")
return sm.JobStatusChange(ctx, job, api.JobStatusPaused, "paused task was last runnable task of job, pausing job")
}
if job.Status == api.JobStatusPauseRequested {
// if the job is in pause-requested state, and all other tasks are paused,
// then the job can be paused.
numPaused, numTotal, err := sm.persist.CountTasksOfJobInStatus(ctx, job, api.TaskStatusPaused)
if err != nil {
return err
}
if numPaused == numTotal {
logger.Info().Msg("all tasks of job are paused, job is paused")
return sm.JobStatusChange(ctx, job, api.JobStatusPaused, "all tasks paused")
}
} else {
// if the job is not in pause-requested state, then some error occurred and the job should be failed.
logger.Info().Msg("task cannot be changed to paused when job is not in pause-requested state")
}
return nil
}
// updateJobOnTaskStatusFailed conditionally escalates the failure of a task to fail the entire job. // updateJobOnTaskStatusFailed conditionally escalates the failure of a task to fail the entire job.
func (sm *StateMachine) updateJobOnTaskStatusFailed(ctx context.Context, logger zerolog.Logger, job *persistence.Job) error { func (sm *StateMachine) updateJobOnTaskStatusFailed(ctx context.Context, logger zerolog.Logger, job *persistence.Job) error {
// Count the number of failed tasks. If it is over the threshold, fail the job. // Count the number of failed tasks. If it is over the threshold, fail the job.