Repository navigation
chore(deps): provider-runtime v1.4.0 — recover interrupted creates instead of refusing forever (#22) - #25
Merged
Conversation
…stead of refusing forever (#22) v1.4.0 gives an unresolved create handshake a bounded path back. Previously, an interrupted create whose external resource Observe could not find was terminal: the reconciler refused with Requeue:false and nothing ever resolved the ambiguity, so the only documented remedy was a human deleting krateo.io/external-create-pending by hand. Six Repo CRs have been in exactly that state on krateo-057 since 2026-09-21, wedged by the go-git transport race that 0.15.4 fixed. That fix stopped NEW wedges; it could not recover the existing ones. This bump is what recovers them. The runtime now confirms the negative rather than holding on the ambiguity: within a 2-minute grace it re-observes without recreating, so a create that did land is adopted rather than duplicated; once the grace has fully elapsed with the resource still absent, the create did not land and the pending marker is cleared so the normal create path runs again. All six markers are days old, so they resolve on the first reconcile after rollout. No git-provider code change is needed. An earlier reading of #22 proposed making Observe consult the forge when the external name is empty; that was wrong. Returning ResourceExists:false there is CORRECT — it is how a new Repo gets created — and changing it could stop Create ever running. The recovery is time-based precisely because the local signal cannot be made meaningful. Verified: unit suite green (9 packages, -race). Integration tests need a Docker daemon and did not run locally; CI covers them. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
braghettos
force-pushed
the
chore/provider-runtime-v1.4.0
branch
from
September 28, 2026 08:56
d859e49 to
ba6cd61
Compare
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.
Closes #22.
What this recovers
Six
RepoCRs have been wedged on krateo-057 since 2026-09-21:They were wedged by the go-git transport race that 0.15.4 fixed. That fix stopped new wedges; it could not recover existing ones. This bump is what recovers them.
provider-runtime v1.4.0 confirms the negative rather than holding on the ambiguity:
All six markers are days old, so they resolve on the first reconcile after rollout.
No git-provider code change, and that is a correction
An earlier reading of #22 — mine — proposed making
Observeconsult the forge when the external name is empty, on the grounds that it currently reports absence from local bookkeeping.That was wrong. Returning
ResourceExists: falsethere is correct: it is how a newRepogets created, and changing it could stopCreateever running. The local signal cannot be made meaningful for the recovery question, which is exactly why the runtime fix is time-based instead.Tests
Unit suite green: 9 packages,
-race, 0 failures.Integration tests did not run locally — they need a Docker daemon for kind and it was not available. CI covers them, and this PR should not merge until it does.
🤖 Generated with Claude Code