Skip to content

Rebalance PostgreSQL feature shards from measured timings - #1046

Merged
rmcdaniel merged 1 commit into
mainfrom
workflow/pg-shard-headroom-20261105
Nov 5, 2026
Merged

rmcdaniel merged 1 commit into
mainfrom
workflow/pg-shard-headroom-20261105

Conversation

@rmcdaniel

Copy link
Copy Markdown
Member

The original main run 42085517168 timed out PostgreSQL shard 0 at its unchanged 600-second limit, despite its 607-test JUnit report having no assertion failures. The new inbox regression defaulted to weight 1 but measured 11.7 seconds; stale relative hints concentrated 601/563 seconds of observed work in the first two shards while the other two needed 404/395 seconds.

Refresh only PostgreSQL hints from the four original JUnit artifacts, with 10% padding and every larger existing hint retained. All 156 files remain assigned once across four shards; MySQL data, test bodies, workflow configuration and the 600-second limit are unchanged. Data replay projects 507/511/487/458 seconds of measured work after redistribution. These are allocation estimates, not a new backend pass.

Validation: reproduced all four original file manifests with the unchanged splitter algorithm; candidate inventory/profile/positive-weight/diff/boundary/secret checks pass. Ordinary exact-head CI and actual resulting-main backend qualification remain required. Original timeout is preserved; no rerun or deadline extension was made.

Refs #1039. This is a measured follow-up to #1041, separate from the accepted immutable 2.20.4 installed proof in #1043.

@sonarqubecloud

sonarqubecloud Bot commented Nov 5, 2026

Copy link
Copy Markdown

@rmcdaniel
rmcdaniel marked this pull request as ready for review November 5, 2026 19:11
@rmcdaniel
rmcdaniel merged commit cd8831d into main Nov 5, 2026
24 checks passed
@rmcdaniel

Copy link
Copy Markdown
Member Author

Normally merged at 19:11:14Z as cd8831d5167d9c514ec39921c49e43bb04e4ba29, tree 12559c0679224829c37a77461afb4373a833dcd8, ordered parents [d5ce3f024d68dcd5d7f1e3785f70971c2c4d1587, 49eec310da8d35fe499b8cd049a039f96d2f686c]. Fresh exact-head ordinary qualification was terminal 17 success / seven declared PR-profile skips, zero failed/pending; protected merge used no admin/bypass or altered check.

Original automatic resulting-main 42091475038/1 is running the full target-branch Quality Cycle. This is validation of changed allocation source, not an unchanged-source retry; original failed 42085517168/1 remains intact. Workflow owns actual PostgreSQL clock/exit results and full quality attribution, then incorporated branch cleanup. No timeout, shard count, MySQL hint, test body, runtime code or publication changed.

@rmcdaniel
rmcdaniel deleted the workflow/pg-shard-headroom-20261105 branch November 5, 2026 19:37
@rmcdaniel

Copy link
Copy Markdown
Member Author

Resulting-main qualification is complete for cd8831d5167d9c514ec39921c49e43bb04e4ba29 / tree 12559c0679224829c37a77461afb4373a833dcd8, ordered parents [d5ce3f02, 49eec310]. Original run 42091475038, attempt 1 finished SUCCESS: 22 successful jobs, three declared PR-only skips, no failing or pending jobs. This carries the incorporated #1039 inbox regression and #1046 measured PostgreSQL allocation correction.

Actual Quality Cycle attribution:

  • ECS and PHPStan passed in quality.
  • Unit/PCOV: PHP 8.3.36, PHPUnit 12.5.43, PCOV 1.0.12; 5,871 tests / 78,806 assertions, 49 skips and 15 PHPUnit notices.
  • Coverage: 75,663 / 80,595 lines = 93.88%; the 93.08% ratchet passes, the 100% target is not reached. Summary ZIP SHA-256 52bb97840040deda8fc56225014a999273f4f87b1aa285ce695217742f9bcd25.
  • MySQL features: 2,827 tests / 28,151 assertions / three skips; all four actual commands exited zero at 729/699/706/661 seconds, within the existing 900-second bound.
  • PostgreSQL features: 2,827 tests / 27,756 assertions / 32 skips; all four actual commands exited zero at 554/401/524/425 seconds, within the unchanged 600-second bound. No assertion errors/failures in either backend.
  • MariaDB feature job and all four Laravel 10–13 embedded upgrade cells passed. Contract, corpus, tracing HTTP, Sonar and aggregate build also passed.

The original d5ce3f02 run 42085517168/1 remains FAILURE: PostgreSQL shard 0 exited 124 at the hard limit despite passing-looking JUnit. It is not retrospectively qualified. The successor adjusts measured allocation weights only; no timeout, shard count or product/test body changed.

Both incorporated remote branches (workflow/1021-parallel-receive-regression and workflow/pg-shard-headroom-20261105) were deleted normally after ancestry/current-head/dependency checks, and their absence was verified. No additional dispatch or retry was made. Waterline owns the normal #1036 refresh over this accepted main; Workflow owns combined review and resulting-main acceptance.

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

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant