fix(bench): give sqs its own drain corpus instead of a 51-hour pass - #163
Merged
Conversation
The drain corpus is a message *count*, and six million was sized for the fastest cell — in-process at ~4.7 M msg/s, where it buys a window of about a second. SQS measures LocalStack at ~900 msg/s and fills at ~660, so the same count buys a 1.7-hour window to measure what a minute measures, and the full pass runs about 51 hours. One count means two different things on backends three orders of magnitude apart. SQS now drains 60 000, as `SQS_DRAIN_MESSAGES`, substituted into the matrix for that target alone. Substituted rather than appended after `--`, because the harness rejects a knob given twice — the runbook already says so. Nothing that makes a drain row a rate changes at this size: the window runs tens of seconds against chartgen's 1 s floor, the group assembles in the first handful of messages, and `--drain-max-bytes` still binds the 64 KiB leg at 49 152, so that leg is not deviated at all. Every other knob is shared, so this is a per-backend corpus, not a second matrix. The deviation is recorded where a reader will meet it rather than only in the script: the script prints a `deviation:` line into the log, every drain row already carries `drain.corpus`, and the charts name the backend whose corpus differs. Also replaces "about an hour per Docker backend" with the measured per-backend rates. That budget was derived from the only two backends ever published, and RabbitMQ at ~18 000 msg/s wants closer to two hours.
Deploying with
|
| Status | Name | Latest Commit | Preview URL | Updated (UTC) |
|---|---|---|---|---|
| ✅ Deployment successful! View logs |
shove-docs | 9b53491 | Commit Preview URL Branch Preview URL |
Sep 07 2026, 09:22 PM |
Codecov Report✅ All modified and coverable lines are covered by tests. 📢 Thoughts on this report? Let us know! |
This was referenced Sep 8, 2026
zannis
added a commit
that referenced
this pull request
Sep 8, 2026
Measure SQS against LocalStack on the same host as the other published runs, with both SQS deviations in place: the 60 000-message drain corpus from #163 and the 100-per-worker FIFO corpus from #173, plus the SNS byte-limit fix from #171 so the 64 KiB batch publishes go through. 60 rows, no failures; consumer_group and broadcast are recorded as unsupported, as the capability traits say. The document now holds every backend. The README and performance page say so, and say what the SQS rows measure: LocalStack, at a recorded corpus, not the AWS service.
zannis
added a commit
that referenced
this pull request
Sep 8, 2026
* bench(results): add the sqs run to the results document Measure SQS against LocalStack on the same host as the other published runs, with both SQS deviations in place: the 60 000-message drain corpus from #163 and the 100-per-worker FIFO corpus from #173, plus the SNS byte-limit fix from #171 so the 64 KiB batch publishes go through. 60 rows, no failures; consumer_group and broadcast are recorded as unsupported, as the capability traits say. The document now holds every backend. The README and performance page say so, and say what the SQS rows measure: LocalStack, at a recorded corpus, not the AWS service. * test(chartgen): assert the grown canvas's plot minimum from the y axis, not the ticks Tick labels stop short of the plot boundaries, so a tick span at 0.8 of the minimum could pass an undersized plot. The y axis is a two-point vertical polyline and bounds the plot body exactly; assert the full minimum from it.
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
1 participant
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.
Depends-on: none — this touches only
scripts/bench.shandbenches/README.md. It changes no Rust, no committed results document and no committed SVG, so nothing gates it and no merge order with #162 misrenders an intermediate state.