Sixteen MiB log segments

[martian56]

Third attempt at measuring this; the first two were blocked by the harness rather than by the change.

Postgres flushes sixteen MiB log segments and ours were sixty four. Our durable commit costs about 2.5x theirs at the same group size and the same flush count, so if flush cost tracks the file extent rather than the bytes dirtied, this is a direct explanation.

Under the old harness it was the only candidate positive on both runs, linux-x64-full plus 11.4 then plus 35.8 percent, while every other profile disagreed with itself by up to 89 points. With both builds now benched on one runner, and a misbehaving baseline no longer failing the job, it can finally be judged.

[coderabbitai]

Warning

Review limit reached

@martian56, you've reached your PR review limit, so we couldn't start this review.

Next review available in: 9 minutes

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: a40cb49b-9534-4c96-9cc6-f29be1f0634d

📥 Commits

Reviewing files that changed from the base of the PR and between 5718650 and 9a54ce0.

📒 Files selected for processing (2)

• storage/src/group_commit.rs
• storage/src/wal.rs

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch segments-final

---

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

• X
• Mastodon
• Reddit
• LinkedIn

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

[github-actions]

Bench report

⚠️ Worst regression: windows-x64-full writes at -33.9%.

Baseline: main @ same runn. Cells read baseline → this PR. Verdicts need more than ±15% to leave the noise band.

linux-arm64-full

workload	ops/s	Δ ops/s	read p99 ms	write p99 ms	write p99.9 ms
reads	30655 → 30923	⚪ +0.9%	0.23 → 0.23	-	-
mixed	14527 → 12300	🔴 -15.3%	0.50 → 0.49	11.20 → 19.35	32.47 → 35.89
writes	5471 → 4903	⚪ -10.4%	-	9.07 → 11.02	28.71 → 29.67

linux-x64-1cpu-1gb

workload	ops/s	Δ ops/s	read p99 ms	write p99 ms	write p99.9 ms
reads	13561	new	0.42	-	-
mixed	8171	new	0.62	19.40	28.53
writes	4476	new	-	8.21	24.00

linux-x64-2cpu-2gb

workload	ops/s	Δ ops/s	read p99 ms	write p99 ms	write p99.9 ms
reads	17517	new	0.39	-	-
mixed	12099	new	0.64	3.20	14.00
writes	6267	new	-	2.28	8.19

linux-x64-full

workload	ops/s	Δ ops/s	read p99 ms	write p99 ms	write p99.9 ms
reads	45409 → 45507	⚪ +0.2%	0.15 → 0.16	-	-
mixed	10848 → 7626	🔴 -29.7%	0.23 → 0.25	29.69 → 56.73	115.37 → 125.78
writes	3711 → 3606	⚪ -2.8%	-	15.47 → 15.76	59.97 → 82.96

windows-x64-full

workload	ops/s	Δ ops/s	read p99 ms	write p99 ms	write p99.9 ms
reads	22727 → 22792	⚪ +0.3%	0.36 → 0.34	-	-
mixed	5452 → 4706	⚪ -13.7%	0.48 → 0.33	30.48 → 31.09	41.87 → 42.08
writes	873 → 577	🔴 -33.9%	-	15.61 → 31.76	24.45 → 42.49

Field reference

All engines in identical containers, same driver, measured 2026-07-24 @ ba0ec39. A different harness than the tables above, so read this as standing, not as this PR's delta.

linux-arm64-full

engine	reads	mixed	writes
quantadb	17978	12420 (27ms wp99)	6449 (19ms wp99)
postgres18	15227	14723 (2ms wp99)	9923 (1ms wp99)
mariadb114	4221	4437 (8ms wp99)	6666 (3ms wp99)
mysql84	4258	4394 (9ms wp99)	4274 (4ms wp99)

linux-x64-1cpu-1gb

engine	reads	mixed	writes
quantadb	19419	10854 (29ms wp99)	5170 (39ms wp99)
postgres18	10196	8580 (43ms wp99)	5912 (46ms wp99)
mariadb114	3455	3732 (4ms wp99)	5626 (8ms wp99)
mysql84	3497	3628 (4ms wp99)	3259 (36ms wp99)

linux-x64-2cpu-2gb

engine	reads	mixed	writes
quantadb	22528	16774 (5ms wp99)	8374 (4ms wp99)
postgres18	17060	17006 (2ms wp99)	11404 (1ms wp99)
mariadb114	3735	4006 (3ms wp99)	6741 (3ms wp99)
mysql84	3804	3845 (4ms wp99)	4847 (3ms wp99)

linux-x64-full

engine	reads	mixed	writes
quantadb	18144	15075 (6ms wp99)	8025 (5ms wp99)
postgres18	15402	15154 (2ms wp99)	10872 (1ms wp99)
mariadb114	3437	3699 (4ms wp99)	6138 (3ms wp99)
mysql84	3490	3615 (4ms wp99)	3996 (5ms wp99)

Shared CI runners: treat single-profile swings as suggestive, agreement across profiles as real.

[martian56]

Measured and rejected, this time on a baseline from the same runner.

• linux-x64-full: minus 2.8 percent, inside the band
• linux-arm64-full: minus 10.4 percent, inside the band
• windows-x64-full: minus 33.9 percent, a clear regression
• the docker profiles have no paired baseline yet

The plus 11.4 and plus 35.8 percent this showed on x64-full under the old harness was an artifact. Paired against the same runner it is neutral to negative everywhere.

Worth stating plainly, because it is the point of the harness work: this is a change I believed in on the strength of two runs agreeing, and it was wrong. Segment size is not why our commit costs about 2.5x a Postgres commit.