Retry pod update on version conflict error in e2e test. #7297
Conversation
ghost
assigned 
|
|
||
| expectNoError(waitForPodRunning(c, pod.Name)) | ||
|
|
||
| By("verifying the updated pod is in kubernetes") | ||
| pods, err = podClient.List(labels.SelectorFromSet(labels.Set(map[string]string{"time": value})), fields.Everything()) | ||
| Expect(len(pods.Items)).To(Equal(1)) | ||
| fmt.Println("pod update OK") | ||
| By("pod update OK") |
There was a problem hiding this comment.
Maybe a Logf here instead of By()?
|
All feedback addressed, and tidied up a few other related things hanging around from the previous version of this test. PTAL @lavalamp |
| } | ||
| if errors.IsConflict(err) { | ||
| Logf("Conflicting update to pod, re-get and re-update: %v", err) | ||
| pod, err = podClient.Get((*pod).Name) |
There was a problem hiding this comment.
Doing the update/get in this order gives an entire 5 seconds for someone to modify the pod and force you into another loop... I'd recommend changing the order back. :)
There was a problem hiding this comment.
Yes, sorry, and I just remembered I forgot to change the time periods.
Let me change those and get back to you.
|
OK, I reduced the retry interval to 500ms for 30s, and changed the code so that the first update attempt (which usually succeeds) uses the original return value of the Create call. Only retries do another Get. Hope you like it. I could fairly easily be convinced to just do a Get before every Update to get rid of the "if" statement - but that would add a completely unnecessary API call in the most common case. |
|
LGTM |
lavalamp
added
the
lgtm
…pdate-retry Retry pod update on version conflict error in e2e test.
ghost
merged commit 
Fixes #7013.