github/kubernetes
1
0
Fork 0
mirror of https://github.com/k3s-io/kubernetes.git synced 2026-01-13 11:25:19 +00:00
Code Issues Projects Releases Wiki Activity

Improves backoff policy in JobController

Browse Source 
issues: https://github.com/kubernetes/kubernetes/issues/56853

Add check if the number of pods succeeded increased since the last
check. If yes the backoff delay is cleared. This logic improves the Job
backoff policy when parallelism > 1 and few pods's Job failed but others
succeed.
This commit is contained in:
cedric lamoriniere
2018-02-05 10:06:19 +01:00
parent 551f73d324
commit c6e8bd62ad
2 changed files with 17 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
pkg/controller/job/job_controller.go
View File

@@ -553,6 +553,14 @@ func (jm *JobController) syncJob(key string) (bool, error) {
}
forget := false
// Check if the number of jobs succeeded increased since the last check. If yes "forget" should be true
// This logic is linked to the issue: https://github.com/kubernetes/kubernetes/issues/56853 that aims to
// improve the Job backoff policy when parallelism > 1 and few Jobs failed but others succeed.
// In this case, we should clear the backoff delay.
if job.Status.Succeeded < succeeded {
forget = true
}
// no need to update the job if the status hasn't changed since last time
if job.Status.Active != active || job.Status.Succeeded != succeeded || job.Status.Failed != failed || len(job.Status.Conditions) != conditions {
job.Status.Active = active
@@ -560,12 +568,12 @@ func (jm *JobController) syncJob(key string) (bool, error) {
job.Status.Failed = failed
if err := jm.updateHandler(&job); err != nil {
return false, err
return forget, err
}
if jobHaveNewFailure && !IsJobFinished(&job) {
// returning an error will re-enqueue Job after the backoff period
return false, fmt.Errorf("failed pod(s) detected for job key %q", key)
return forget, fmt.Errorf("failed pod(s) detected for job key %q", key)
}
forget = true