Dave Protasowski
Dave Protasowski
> However: As it stands, it's possible that a revision is never touched by the labeler, for example if Configuration and Route are created independently. Such a Revision stays in...
> For example if you create a Route + Config and specify specific revisions in the Route. Am I missing something I feel like that's not the case cause the...
/triage accepted
> @dprotaso should this PR be closed as well based on closing the issue for inactivity - [#12908 (comment)](https://github.com/knative/serving/issues/12908#issuecomment-1203086546)
The general problem we have is knowing whether a failure is transient or permanent. ie. image pulls fail due to blips in the network. Even in your example an exodus...
I'm a bit stumped because the selection of the next latest ready revision only looks at revisions newer than the current latest ready revision see [findAndSetLatestReadyRevision](https://github.com/knative/serving/blob/7c6d4bc5f2947c9c488d125685f2243785689e8b/pkg/reconciler/configuration/configuration.go#L135). There's one an exception...
Actually this looks a bit off - this is looking the revision's generation and _not_ the configuration generation label that's part of the revision https://github.com/knative/serving/blob/7c6d4bc5f2947c9c488d125685f2243785689e8b/pkg/reconciler/configuration/configuration.go#L172 Can you post the yaml...
It's also unclear how the generation would change given the spec can't be changed https://github.com/knative/serving/blob/7c6d4bc5f2947c9c488d125685f2243785689e8b/pkg/apis/serving/v1/revision_validation.go#L39-L54
*working theory* If we assume the second revision's deployment failed to progress because of some event. ie. a pod was rolled/node drained - we would need logs/k8s events to confirm....