Fix metabox (apt profile + daemon respawn) (BugFix)#866
Conversation
Hook25
left a comment
Hook25
left a comment
There was a problem hiding this comment.
If you want to change them all here, I'm fine with it but:
- change the title of the PR
- change also the deb service
yphus
left a comment
yphus
left a comment
There was a problem hiding this comment.
The service file provided by the checkbox-ng deb package also needs the same fix:
https://github.com/canonical/checkbox/blob/main/checkbox-ng/debian/checkbox-ng.service
kissiel
changed the title
Updated in a3c509e |
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #866 +/- ##
=======================================
Coverage 35.70% 35.70%
=======================================
Files 303 303
Lines 34250 34250
Branches 5917 5917
=======================================
Hits 12230 12230
Misses 21458 21458
Partials 562 562
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. |
Hook25
left a comment
Hook25
left a comment
There was a problem hiding this comment.
+1, ty for updating them all!
|
I also removed the glxgears test that was silently failing before, but now it started failing properly and we really don't want to have it tested. |
The previous "on-failure" meant that sending a signal to agent (for instance SIGKILL) would completely terminate the agent and make systemd not restart it. This is very bad for emulating agent crashes. Thus I'm proposing here to `always` do it.
Considering that Checkbox is a tool that's intended to be sometimes killed, it needs to always be brought back online (its agent). Immediately this fixes the agent respawn metabox tests and goes in line with the changes being made to the auto agent respawn stories.
This is an relic from the times when we wanted to improve container provisioning speed, but now, when testing from source we provision everything with a normal apt execution, so this can go away. Also this list had a potential to quickly go stale.
kissiel
force-pushed
the
metabox-improvements
branch
from
3d83da9 to
8b2e784
Compare
* in metabox always restart the agent The previous "on-failure" meant that sending a signal to agent (for instance SIGKILL) would completely terminate the agent and make systemd not restart it. This is very bad for emulating agent crashes. Thus I'm proposing here to `always` do it. * make agents from snaps **always** respawn Considering that Checkbox is a tool that's intended to be sometimes killed, it needs to always be brought back online (its agent). Immediately this fixes the agent respawn metabox tests and goes in line with the changes being made to the auto agent respawn stories. * always respawn deb-delivered checkbox agents * remove some of the debs in the pre-provisioning from lxd profiles in Metabox
3 participants
Description
efe7665
The previous "on-failure" meant that sending a signal to agent (for instance SIGKILL) would completely terminate the agent and make systemd not restart it.
This is very bad for emulating agent crashes. Thus I'm proposing here to
alwaysdo it.50e3f43 and 33f5ecc
Make agents from snaps always respawn
Considering that Checkbox is a tool that's intended to be sometimes killed, it needs to always be brought back online (its agent).
Immediately this fixes the agent respawn metabox tests and goes in line with the changes being made to the auto agent respawn stories.
ec27e41
Remove apt-related pre-provisioning from lxd profiles in Metabox
This is an relic from the times when we wanted to improve container provisioning speed, but now, when testing from source we provision everything with a normal apt execution, so this can go away. Also this list had a potential to quickly go stale.
Testing
Tested with MB runs using an agent-killing scenario.