| *** elodilles is now known as elodilles_pto | 10:34 | |
| *** haleyb_ is now known as haleyb | 13:49 | |
| ykarel | gmann, hi | 15:26 |
|---|---|---|
| ykarel | https://review.opendev.org/c/openstack/tempest/+/887237 brow slow jobs in releases before xena | 15:26 |
| ykarel | s/brow/broke | 15:27 |
| gmann | ykarel: hi | 15:27 |
| ykarel | with ERROR: unknown environment 'slow' | 15:27 |
| gmann | ykarel: ohk, we use old tempest < xena but job definition are picked from master that is why | 15:28 |
| ykarel | yeap right | 15:28 |
| gmann | ykarel: let me fix is quickly | 15:28 |
| ykarel | Thanks | 15:28 |
| ykarel | reported https://bugs.launchpad.net/neutron/+bug/2027817 for tracking | 15:34 |
| gmann | ykarel: thanks. | 15:34 |
| opendevreview | Ghanshyam proposed openstack/tempest master: Fix tempest-slow-py3 job for < stable/xena https://review.opendev.org/c/openstack/tempest/+/888590 | 15:52 |
| mtreinish | melwitt: yeah that's still a desireable feature, but I was struggling with issues in the interprocess communication in that PR. | 16:51 |
| mtreinish | Although the last time I tried it, I think it was working better than I remember, But it's been a long time since I tried it | 16:52 |
| mtreinish | I'll try to find some time to dust off that PR if there is an increased interest in having the feature again | 16:55 |
| melwitt | mtreinish: oh yeah, we've been having things where the time spent by each process running tests can be very unbalanced. like one process spends over an hour running tests and another spent only 10 min | 18:01 |
| melwitt | so jobs can timeout while there were processes sitting idle that could have been running stuff | 18:02 |
| mtreinish | hmm, it's been a really long time since I've looked at the ci config. But at least the last time I did look we were pre-seeding the stestr history before starting the tests to give the stestr scheduler some timing data to better balance the workers. Are the jobs still doing that? | 18:53 |
| mtreinish | but, yeah that balance seems really bad | 18:53 |
| *** haleyb_ is now known as haleyb | 21:26 | |
| melwitt | mtreinish: I think so? I couldn't figure out what the history was based on but I do think I saw something related to using history to decide how to distribute the tests | 22:19 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!