fix(stovepipe): stop wrapping the buildsignal re-poll publish as retryable - #469
Open
behinddwalls wants to merge 1 commit into
Open
Conversation
This was referenced Jul 29, 2026
behinddwalls
force-pushed
the
preetam/stovepipe-publish-classification
branch
from
July 29, 2026 22:41
16f5390 to
2ae86d2
Compare
behinddwalls
marked this pull request as ready for review
July 29, 2026 23:10
…yable ## Summary ### Why? `buildsignal` wrapped a failed reschedule publish in `errs.NewRetryableError`. `platform/errs` rule 4 says not to do that: a failed queue publish must not be made retryable just because replaying it is convenient, because that turns a permanent enqueue failure into an infinite retry instead of dead-lettering. It was also inconsistent with the sibling path — the publish to `record`, one branch above, already returns raw. ### What? Returns the raw error, leaving the classifier's verdict to stand, and drops the now-unused `errs` dependency. ### Merge ordering **This should merge only after the build and buildsignal stages have DLQ consumers.** Today only `stovepipe-process-dlq` is registered, and `MoveToDLQ` routes by topic-suffix convention, so anything dead-lettered from `buildsignal` lands on `buildsignal_dlq` and is never drained — leaving the request non-terminal with its build slot held, which is the wedge the reconciler exists to prevent. This change does not create that hole; every other non-retryable failure in the stage already falls into it. But it does add one more path in, so the ordering is worth respecting rather than discovering later. If the DLQ work slips, the alternative is to land this anyway and accept parity with the existing failure modes. ## Test Plan ✅ `bazel test //stovepipe/...` — the reschedule-publish-failure case now asserts a non-retryable classification. ✅ `bazel test //test/e2e/stovepipe/...` — the poll loop is unaffected on the success path.
behinddwalls
force-pushed
the
preetam/stovepipe-publish-classification
branch
from
July 30, 2026 07:51
2ae86d2 to
71366e3
Compare
mnoah1
approved these changes
Jul 30, 2026
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.
Summary
Why?
buildsignalwrapped a failed reschedule publish inerrs.NewRetryableError.platform/errsrule 4 says not to do that: a failed queue publish must not be made retryable just because replaying it is convenient, because that turns a permanent enqueue failure into an infinite retry instead of dead-lettering. It was also inconsistent with the sibling path — the publish torecord, one branch above, already returns raw.What?
Returns the raw error, leaving the classifier's verdict to stand, and drops the now-unused
errsdependency.Merge ordering
This should merge only after the build and buildsignal stages have DLQ consumers. Today only
stovepipe-process-dlqis registered, andMoveToDLQroutes by topic-suffix convention, so anything dead-lettered frombuildsignallands onbuildsignal_dlqand is never drained — leaving the request non-terminal with its build slot held, which is the wedge the reconciler exists to prevent.This change does not create that hole; every other non-retryable failure in the stage already falls into it. But it does add one more path in, so the ordering is worth respecting rather than discovering later. If the DLQ work slips, the alternative is to land this anyway and accept parity with the existing failure modes.
Test Plan
✅
bazel test //stovepipe/...— the reschedule-publish-failure case now asserts a non-retryable classification.✅
bazel test //test/e2e/stovepipe/...— the poll loop is unaffected on the success path.Stack