Other Pod(s) Status Clean up & Pod(s) stuck in Terminating state #89
Conversation
| @@ -24,13 +24,39 @@ func podRelatedToCronJob(pod *corev1.Pod, jobStore cache.Store) bool { | |||
| return false | |||
| } | |||
|
|
|||
| func shouldDeletePod(pod *corev1.Pod, orphaned, pending, evicted, successful, failed time.Duration) bool { | |||
| func shouldDeleteTerminatingPod(pod *corev1.Pod, orphaned, pending, evicted, terminating, successful, failed time.Duration) bool { | |||
There was a problem hiding this comment.
why do you need to pass orphaned, pending, evicted,successful, failed durations if you only use terminating?
| if !podFinishTime.IsZero() { | ||
| age := time.Since(podFinishTime) | ||
| if terminating > 0 && age >= terminating { | ||
| log.Println("Pod(s) Which Are In Terminating State") |
There was a problem hiding this comment.
@lwolf for troubleshooting I have added these logs and I thought I have removed those lines.will update the same.
| // evicted pods, those with or without owner references, but in Evicted state | ||
| // - uses c.deleteEvictedAfter, this one is tricky, because there is no timestamp of eviction. | ||
| // So, basically it will be removed as soon as discovered | ||
| if pod.Status.Phase == corev1.PodFailed && pod.Status.Reason == "Evicted" && evicted > 0 { | ||
| return true | ||
| } | ||
| if pod.Status.Phase == corev1.PodFailed && pod.Status.Reason == "OutOfpods" && evicted > 0 { |
There was a problem hiding this comment.
please add comment about outOfpods and outOfcpu reason or a link to the docs describing it's behavior.
| @@ -155,7 +156,7 @@ func TestKleaner_DeletePod(t *testing.T) { | |||
| } | |||
| for name, tc := range testCases { | |||
| t.Run(name, func(t *testing.T) { | |||
| result := shouldDeletePod(tc.podSpec, tc.orphaned, tc.pending, tc.evicted, tc.successful, tc.failed) | |||
| result := shouldDeletePod(tc.podSpec, tc.orphaned, tc.pending, tc.evicted, tc.terminated, tc.successful, tc.failed) | |||
There was a problem hiding this comment.
please add test scenario for every case that you're adding
There was a problem hiding this comment.
@lwolf yes. I will be adding the test cases for each scenario and pushing one by one change for this PR
| func shouldDeletePod(pod *corev1.Pod, orphaned, pending, evicted, successful, failed time.Duration) bool { | ||
| func shouldDeleteTerminatingPod(pod *corev1.Pod, orphaned, pending, evicted, terminating, successful, failed time.Duration) bool { | ||
| // terminating pods which got hanged, those with or without owner references, but in Evicted state | ||
| // - uses c.deleteEvictedAfter, this one is tricky, because there is no timestamp of eviction. |
There was a problem hiding this comment.
I assume this comment is just a copy-paste from the shouldDeletePod, cause it doesn't make any sense
|
|
||
| // In Case If Pod(s) is Stuck in Terminating state just in case to delete with force | ||
| // Not goood way to fid the root cause of the issue why pod is stuck in terminating state | ||
| func (c *Kleaner) DeletePodWithForce(pod *corev1.Pod) { |
There was a problem hiding this comment.
as we discussed in the issue I'd prefere not to have force-deletion in the codebase
There was a problem hiding this comment.
@lwolf yes but we can keep by enabling the flag. it will be useful when there is an issue calico and workload on this node will be blocked due to this issue so to avoid those pod(s) which are stuck in terminating state until the real issue is fixed
There was a problem hiding this comment.
no, the issue should be solved by node draining, not force-deletion.
Added the below changes:
Issues #88 :
#88