Skip to content

fix(relay): bound storage conformance admission time - #24

Draft
mfethe1 wants to merge 4 commits into
product/mainfrom
codex/buzz-conformance-timeout
Draft

fix(relay): bound storage conformance admission time#24
mfethe1 wants to merge 4 commits into
product/mainfrom
codex/buzz-conformance-timeout

Conversation

@mfethe1

@mfethe1 mfethe1 commented Sep 7, 2026

Copy link
Copy Markdown
Owner

Summary

A storage backend that stalls during the relay's conformance probe can leave startup waiting indefinitely. Bound the complete production probe to 120 seconds and fail startup when it expires. An unfinished conditional-write race is never admitted.

Select the regression tests in both unit-test runners and remove redundant ACP test invocations while preserving the existing full-crate lane.

Related issue

Related machine-home work: #15. Discovered while recovering the Mack relay.

Testing

Three real HTTP probe tests passed: a stalled backend was denied after 250 ms, a responsive backend passed, and a zero deadline failed before issuing a request. Removing the timeout made the regression fail; restoring it passed. The exact final head passed the relay binary-target check, formatting, and six runner routing/failure-propagation scenarios. Those runner scenarios use a captured subprocess stub; the subsequent full local CI also ran all 968 ACP tests successfully. Independent source review passed at 195c72764c65bf2a1ae3ff87576aca1874e3db3d.

The actual native relay executable was tested with unchanged production settings of 32 writers, three rounds, and a 120-second deadline. Stalled storage caused exit 1 after 120.153 seconds from probe start, without admission or readiness. Responsive storage was admitted with zero transport drops, returned HTTP 200 {"status":"ready"}, and shut down cleanly. Root independently checked the actual binary hash, raw logs, and closed fixture ports. Two preliminary fixture setup/guard failures are retained; owned test databases, services, and private credentials were removed.

Full just ci failed only at its final mobile stage: 2,241 passed and eight golden comparisons failed. All eight actual images matched the independently verified unchanged-base images on Airy. All preceding stages passed, including 6,583 desktop tests. Full just test passed its unit stages, then failed the Docker Compose infrastructure launcher with unknown shorthand flag: 'd' in -d and a service wait timeout. The isolated native startup checks passed separately.

This change has not been deployed as part of this work. Hosted checks for this head have completed with image-build, Windows Rust, PostgreSQL, desktop smoke, and desktop integration failures. Hosted qualification and protected release approval remain outstanding; successful local checks do not qualify this PR for merge.

Signed-off-by: Michael Feth <mfethe1@gmail.com>
Signed-off-by: Michael Feth <mfethe1@gmail.com>
Signed-off-by: Michael Feth <mfethe1@gmail.com>
@coderabbitai

coderabbitai Bot commented Sep 7, 2026

Copy link
Copy Markdown

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

  • Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

Comment @coderabbitai help to get the list of available commands.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant