Closed DilumAluthge closed 2 years ago
Yeah I dunno. I tried reloading it from scratch, and it seems to have worked.
What is "reloading it from scratch"?
Killing the buildbot and restarting it, rather than reloading while the old process is still running.
Okay, watching the console as you just caused it to reconfig; it appears that it successfully finishes reconfiguring..... after 21 seconds. So because it takes so long, it always reports it as a failure.
Ah I see.
So, the tricky part here is that even after you killed and restarted the Buildbot, the auto-reload still fails. E.g. https://build.julialang.org/#/builders/34/builds/68
So whenever we make future changes to this repo, we'll need to manually kill and restart the Buildbot to deploy the changes. (Because the auto-reload won't work.)
I think that mild inconvenience is still worth the benefit of the better resource management we get with auto-canceling old PR jobs. But it is slightly inconvenient.
Although the auto reload job reports failure, I believe it's actually working. The logs say "reconfigure successful after 21s", so I think it's just that the webui shows the timeout. We can probably adjust the timeout as well.
We can probably adjust the timeout as well.
That seems like the path of least resistance!
Maybe make it 30 seconds?
Is this timeout something that I can do with a PR to this repo? Or does it need to be done on the Buildbot machine itself?
We can probably adjust the timeout as well.
That seems like the path of least resistance!
Maybe make it 30 seconds?
Is this timeout something that I can do with a PR to this repo? Or does it need to be done on the Buildbot machine itself?
See #250 and #251
@staticfloat It is with this change that the auto-reload fails. See e.g. https://build.julialang.org/#/builders/34/builds/66
The failure is deterministic, in the sense that I keep re-running it, and it fails every time.