Repository navigation
Loosen the embedding gate's elapsed-time bound so a loaded runner cannot fail it - #881
Merged
Merged
Conversation
…not fail it Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Signed-off-by: WaylandYang <145302500+WaylandYang@users.noreply.github.com>
WaylandYang
force-pushed
the
fix/embedding-gate-timing
branch
from
September 23, 2026 09:26
408b721 to
54a1186
Compare
This was referenced Sep 23, 2026
Merged
Merged
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Why
The first
pushrun ondevafter #879 (run 35840994347) failed intestonpipeline::pipeline_tests::the_embedding_gate_is_never_held_beyond_its_ceiling:The test drives twelve 150 ms fake-embedding batches through the 4-wide gate and asserted total elapsed
< serial / 2(900 ms). Ideal is ~450 ms, but on a 4-vCPU runner with ~390 other tests running in the same process a 1.03 s wall time is ordinary scheduling noise, not a gate problem: the same run's peak-in-flight measurement was within the ceiling and batches overlapped. Every other job in that run was green. The test comes from #539 and none of the recent CI changes touch what it exercises.What
Only the elapsed-time bound changes, from
serial / 2toserial * 3 / 4(1.35 s). It remains a backstop against the gate fully serialising the batches (1.8 s); the properties the test is really about, peak in-flight ≤EMBED_JOBSand peak ≥ 2, are untouched and still strict. Comment explains why the bound is where it is.Verification
Local: the test run 10× in a row against the docker compose Postgres, all passing (times in the PR thread once the run finishes).
🤖 Generated with Claude Code