Since we moved to the Queue API (bug 1259627) we don't retry the flaky docker image generation (bug 1367491). In announcing the move rail said: There is one thing to watch out. The old API supports `reruns`, which helped us rerunning failed tasks automatically. I tried to work around the lack of this feature in the Queue API by using docker-worker's `onExitStatus`, but it may behave a bit differently.
Created attachment 8902131 [details] [review] retry on 255
(In reply to Rail Aliiev [:rail] ⌚️UTC+3 from comment #2) > Created attachment 8902131 [details] [review] > retry on 255 Replied in PR with requested changes. Will change the flags here as well once we've merged to avoid re-setting flag for review.
Comment on attachment 8902131 [details] [review] retry on 255 Deployed
56.0b8 is not helping, there were not failures! :)
Strongly suspect https://hg.mozilla.org/mozilla-central/rev/84fd52d2832a#l4.14 is the reason for that, and fixes bug 1367491.
That hasn't been uplifted to beta though, so maybe something else/coincidence.
We got retries for hg errors in tasks 0, 1, and 2 in https://tools.taskcluster.net/groups/PFi2U7q2SCWNvW-ud7TkWw/tasks/csCjgMfVQxqOyBV6aKAB3w/details. Then it failed in task 3 on a clamav error, where we get an exit status of -1.
Added -1 to the list in https://github.com/mozilla-releng/releasetasks/pull/276 and deployed
Closing this. Bug 1398964 is a good to have.