test(users): force 10 seconds sleep to improve reliability #1215
Conversation
| time.sleep(10) | ||
| new_user.delete() |
There was a problem hiding this comment.
I think this should go after, because the delete itself is taking too long:
| time.sleep(10) | |
| new_user.delete() | |
| new_user.delete() | |
| time.sleep(10) |
Looks like we really need this. I thought my little sidekiq wait hack would work though 🤕 I based it off of this, but I'm not sure if we can extract from the API when this exact sidekiq deletion job is finished:
We can't guarantee when the user will actually be deleted by Sidekiq; this could take a while if the user has a lot of projects/events/etc.
I'll keep looking into this later, maybe we could also just wrap these asserts in retries until the condition is met (with a timeout), since it seems to me like GitLab gets more bloated with new features and so things can take longer with new releases on these little Travis VMs?
They do perform better during the day/office hours though, just retried all the jobs and they're going green 🤣
There was a problem hiding this comment.
Maybe use something like flaky for this?
|
Maybe we don't need this anymore. Seems like #1316 might have resolved the issue. Not 100% sure, but I haven't noticed this error since it got merged. |
True, I think your PR fixed it! We can probably close this now and re-enable the other flaky test as well. |
I'm welcome for better ideas