Skip to content

fix(test-suites): add missing --pipeline 10 to mget-1KiB-5keys-pipeline-10 (v0.3.18)#394

Merged
fcostaoliveira merged 2 commits into
redis:mainfrom
paulorsousa:fix-mget-1KiB-5keys-pipeline-10-missing-pipeline-arg
Jun 26, 2026
Merged

fix(test-suites): add missing --pipeline 10 to mget-1KiB-5keys-pipeline-10 (v0.3.18)#394
fcostaoliveira merged 2 commits into
redis:mainfrom
paulorsousa:fix-mget-1KiB-5keys-pipeline-10-missing-pipeline-arg

Conversation

@paulorsousa

@paulorsousa paulorsousa commented May 21, 2026

Copy link
Copy Markdown
Collaborator

Summary

memtier_benchmark-5Mkeys-string-mget-1KiB-5keys-pipeline-10.yml was missing --pipeline 10 in its clientconfig.arguments. Every sibling pipeline-10 MGET spec on the same matrix (100B, 512B, 1KiB-* with various key counts) ships the flag — this one was the only outlier.

Without the flag, memtier_benchmark defaults to pipeline=1, so the test was running un-pipelined despite the name advertising pipeline-10.

Why it matters

A recent regression-investigation run on redis/redis#14907 (per-IO-thread stat counters) flagged mget-1KiB-5keys-pipeline-10 as the only σ-confirmed regression in the io-threads=8 string suite (−12.6 %). Every other pipeline-10 MGET sibling on the same axis (+1.5 % to +6.5 %) moved the opposite direction, which didn't fit any false-sharing-fix mechanism. Spec inspection showed the --pipeline 10 was missing — the regression was being measured at effective pipeline=1.

Changes

  • Add --pipeline 10 to the clientconfig.arguments line.
  • Bump to v0.3.18.

Test plan

  • Re-run mget-1KiB-5keys-pipeline-10 against PR head + unstable baseline on oss-standalone-08-io-threads; confirm both numbers shift up to the pipelined operating point and the PR delta no longer shows the spurious −12 %.

Note

Low Risk
Benchmark-spec correction and packaging metadata only; no runtime application logic changes.

Overview
Aligns memtier_benchmark-5Mkeys-string-mget-1KiB-5keys-pipeline-10 with other pipeline-10 MGET specs by adding --pipeline 10 to clientconfig.arguments. Without it, memtier ran at pipeline=1, so benchmark deltas (e.g. spurious regressions on io-threads suites) did not match the intended pipelined workload.

pyproject.toml bumps the package to 0.3.23, explicitly **include**s files-to-groups.json and core-specs.json in sdist/wheel, and raises poetry_core to >=1.5.0 so dict-form include with format is honored at build time.

Reviewed by Cursor Bugbot for commit 53d2f03. Bugbot is set up for automated code reviews on this repo. Configure here.

…ne-10

The clientconfig.arguments for `memtier_benchmark-5Mkeys-string-mget-1KiB-5keys-pipeline-10`
was missing the `--pipeline 10` flag (sibling specs at other value sizes — 100B, 512B —
all carry it). Without the flag, memtier_benchmark defaulted to pipeline=1, so the test
was not exercising the pipelined-MGET path its name advertises.

This made the run produced for redis/redis#14907 misleading: the −12.6 % "regression"
on this spec was at effective pipeline=1, while every other pipeline-10 MGET sibling
on the same matrix showed +1–7 % wins under proper pipelining.

Bump to 0.3.18.
The branch bumped pyproject 0.3.17->0.3.18, which now conflicts with main
(already at 0.3.23). Restore pyproject.toml to main so only the test-suite
fix (adding the missing --pipeline 10 to the pipeline-10 mget spec) remains.
@fcostaoliveira fcostaoliveira merged commit 518f28e into redis:main Jun 26, 2026
9 checks passed
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.

2 participants