The application-services mac workers keep data around between builds
Categories
(Infrastructure & Operations :: RelOps: Posix OS, task)
Tracking
(Not tracked)
People
(Reporter: bdk, Unassigned)
References
Details
(Whiteboard: [relops-mac])
Our mac worker seems to keep around directories between builds, which causes taskcluster errors like this one. The task worked once, but the second time around the artifacts
directory was still there from the last run which made it fail.
Is there a way to get a fresh build directory each time? Alternatively, is there something we could be doing differently in our taskcluster code?
Updated•2 years ago
|
Updated•2 years ago
|
Comment 1•2 years ago
|
||
This looks like a l3 worker, so it uses generic-worker multiuser (which creates a new user per task) so we shouldn't be seeing task interference.
Looking at the output, it seems like the task is working inside it's task directory (so even without multiuser tasks shouldn't interfere with each other).
Checking out the worker, I see that /var/opt/generic-worker/tasks/task_168661858129484/
is being used for multiple tasks... not sure why.
:masterwayz or :pmoore, any ideas?
Comment 2•2 years ago
|
||
It could be caused by something like this:
https://github.com/mozilla-platform-ops/ronin_puppet/pull/477/files
If that isn't it, I can take a closer look.
Comment 3•2 years ago
|
||
I can't find worker logs in papertrail - do they go somewhere else?
To be clear, numberOfTasksToRun
should be 0, not 1.
Comment 4•2 years ago
|
||
I think that was a troubleshooting step for something unrelated, I'll fix that in a PR and quick rollout to just those workers.
No idea what could cause that folder directory issue.
Description
•