This repository has been archived by the owner on Nov 1, 2022. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 1.1k
automated
and ignore
annotations clash
#1748
Labels
Comments
hiddeco
added
bug
☠ high user impact
review
Issues that need a review
and removed
review
Issues that need a review
labels
Feb 19, 2019
Thanks a lot for your high quality bug report and proposed solution. 🥇 I validated the bug (valid) and your proposed solution (correct) and turned it into #1749. I am aiming to get a patch release out as soon as possible after this has been merged into |
Awesome! To be honest, if this was still open by the weekend I was going to use it to cut my teeth on Go testing a bit more. 😁 Thanks for verifying so quickly! |
Our |
@whitlaaa until we have settled and prepped for release |
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Having a resource annotated as both automated and ignored appears to be treated as both and seems to break automation of all cluster resources.
In our case, as long as we had some HelmRelease manifests annotated with both, none of the automated resources in our cluster were getting their images updated. Note that we were seeing the latest images available (via
fluxctl list-images
), they were just never used.Upon removing one of the two, everything started flowing as normal.
One of our HelmReleases looked like this:
With both of these annotations in place, we'd have log entries that seem to indicate it being both ignored and automated:
ts=2019-02-18T22:02:34.277718708Z caller=sync.go:66 component=sync-loop resource=default:helmrelease/my-webapi ignore=apply ts=2019-02-18T22:07:27.170619954Z caller=images.go:33 component=sync-loop error="checking services for new images: helmreleases.flux.weave.works \"my-webapi\" not found"
After doing some digging, I think this occurs due to an issue in pollForNewImages.
It looks like GetUnlockedAutomatedResources doesn't take the
policy.Ignore
into account, so the image check fails when comparing with a non-existent flux controller on line 33. Since that error then causes a return, it short circuits image updates for everything else.The text was updated successfully, but these errors were encountered: