fix(observer): a held-out member may fail the listing loudly, and that is still maintenance - #340
Merged
Merged
Conversation
…t is still maintenance The live proof stopped a member's daemon for real and found the second failure mode: where a poolless daemon had answered partially and silently, a stopped one fails the whole cluster listing with an error, and the suppression did not cover loudness -- the platform went unhealthy over a member that was exactly where its operator put it. While any member is held out, the listing error moves to ListingUnavailable instead of the collection errors; without a hold it pages exactly as before, and the degraded-visibility rule still guards the window either way. Claude-Session: https://claude.ai/code/session_01LsGid6U5RrQdFvJmvYdGCF
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The live proof of #338 stopped a member's incus for real (socket masked, 80s hold) and found the second failure mode: a truly-down daemon fails the whole cluster listing with an error — where the poolless daemon of the original incident had answered partially and silently. Detection worked (
held_out: [gha-runner-4], both gauges fired) but the platform went unhealthy over the loud error.While any member is held out, the listing error now moves to
ListingUnavailable(attributed, not paged); without a hold it pages exactly as before. Both directions tested.internal/fleetobserveonly — no provider release.https://claude.ai/code/session_01LsGid6U5RrQdFvJmvYdGCF