support lifecycle rules in ci-config
Categories
(Taskcluster :: Operations and Service Requests, enhancement)
Tracking
(Not tracked)
People
(Reporter: bstack, Assigned: bstack)
Details
Attachments
(4 files)
As it says on the tin.
Assignee | ||
Updated•4 years ago
|
Assignee | ||
Comment 1•4 years ago
|
||
Not sure if this is the best way to go about it or what else I should do!
Assignee | ||
Comment 2•4 years ago
|
||
Pushed by mozilla@hocat.ca: https://hg.mozilla.org/ci/ci-admin/rev/725e7504edd7 Allow configuring lifecycle rules for workerpools r=tomprince
Pushed by mozilla@hocat.ca: https://hg.mozilla.org/ci/ci-configuration/rev/40b3c40ac64a Set shorter reregistration timeout for workers r=tomprince
Comment 5•4 years ago
|
||
It turns out that most of our workers are on legacy docker-worker images, which have not been updated to
support registration, so I don't think we to limit the worker life-time this way.
Comment 6•4 years ago
|
||
MANUAL PUSH: bypassing broken lando restrictions
Backout by mozilla@hocat.ca: https://hg.mozilla.org/ci/ci-configuration/rev/d64633cbcd67 Backed out changeset 40b3c40ac64a for workers not supporting registration.
Assignee | ||
Comment 8•4 years ago
|
||
So we saw some workers that lived long enough to be terminated due to not reregistering?
Comment 9•4 years ago
|
||
Yep. In particular infra/build-decision
has a single worker that doesn't end up ever being idle for >15m, so does not shutdown. I noticed because worker-manger still showed it as running, and didn't start a worker, when it hit credential timeout.
Comment 10•4 years ago
|
||
Otherwise, ci-admin will think that the value always needs changing.
Comment 11•4 years ago
|
||
Pushed by mozilla@hocat.ca: https://hg.mozilla.org/ci/ci-configuration/rev/7419afc30827 Set default lifecycle to taskcluster default; r=Callek
Comment 12•3 years ago
|
||
I think this is done :)
Description
•