test(amber): trim LoopIntegrationSpec nested cases to 2x2 and tighten the run deadline - #7053
Merged
Merged
Conversation
… the run deadline The two nested-loop cases used a 3-row input, so each ran 3 outer x 3 inner iterations; every iteration re-executes its region and respawns every worker in it (~2s per Python worker), making the two nested cases dominate the suite's CI cost. A 2-row input (2x2) keeps the coverage that matters -- nesting depth drives the loop_counter/loop_start_id envelope, and each loop level still takes its back-edge at least twice across the run -- while roughly halving the suite's worker respawns. Also tighten the per-workflow completion deadline from 3 minutes to 90 seconds (~2.5x the slowest healthy case on the 3-core macOS runner). On macOS the suite intermittently hangs mid-run and burns the entire deadline before withRetry re-runs the test, so the loose deadline was costing ~3 minutes per occurrence on top of the retry. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Contributor
Automated Reviewer SuggestionsBased on the
|
Contributor
|
| config | throughput | MB/s | latency | max Δ latest / 7d | |
|---|---|---|---|---|---|
| 🔴 | bs=10 sw=10 sl=64 | 393 | 0.24 | 23,826/35,436/35,436 us | 🔴 +10.1% / 🔴 +124.6% |
| ⚪ | bs=100 sw=10 sl=64 | 795 | 0.485 | 125,414/147,122/147,122 us | ⚪ within ±5% / 🔴 +37.0% |
| ⚪ | bs=1000 sw=10 sl=64 | 903 | 0.551 | 1,110,980/1,155,097/1,155,097 us | ⚪ within ±5% / 🔴 +13.3% |
Baseline details
Latest main 791d7df from same runner
| config | metric | PR | latest main | 7d avg | Δ latest | Δ 7d |
|---|---|---|---|---|---|---|
| bs=10 sw=10 sl=64 | throughput | 393 tuples/sec | 403 tuples/sec | 786.12 tuples/sec | -2.5% | -50.0% |
| bs=10 sw=10 sl=64 | MB/s | 0.24 MB/s | 0.246 MB/s | 0.48 MB/s | -2.4% | -50.0% |
| bs=10 sw=10 sl=64 | p50 | 23,826 us | 24,571 us | 12,305 us | -3.0% | +93.6% |
| bs=10 sw=10 sl=64 | p95 | 35,436 us | 32,176 us | 15,774 us | +10.1% | +124.6% |
| bs=10 sw=10 sl=64 | p99 | 35,436 us | 32,176 us | 18,978 us | +10.1% | +86.7% |
| bs=100 sw=10 sl=64 | throughput | 795 tuples/sec | 816 tuples/sec | 999.71 tuples/sec | -2.6% | -20.5% |
| bs=100 sw=10 sl=64 | MB/s | 0.485 MB/s | 0.498 MB/s | 0.61 MB/s | -2.6% | -20.5% |
| bs=100 sw=10 sl=64 | p50 | 125,414 us | 122,191 us | 100,616 us | +2.6% | +24.6% |
| bs=100 sw=10 sl=64 | p95 | 147,122 us | 140,649 us | 107,356 us | +4.6% | +37.0% |
| bs=100 sw=10 sl=64 | p99 | 147,122 us | 140,649 us | 113,255 us | +4.6% | +29.9% |
| bs=1000 sw=10 sl=64 | throughput | 903 tuples/sec | 908 tuples/sec | 1,031 tuples/sec | -0.6% | -12.4% |
| bs=1000 sw=10 sl=64 | MB/s | 0.551 MB/s | 0.554 MB/s | 0.63 MB/s | -0.5% | -12.5% |
| bs=1000 sw=10 sl=64 | p50 | 1,110,980 us | 1,107,345 us | 980,328 us | +0.3% | +13.3% |
| bs=1000 sw=10 sl=64 | p95 | 1,155,097 us | 1,136,262 us | 1,027,528 us | +1.7% | +12.4% |
| bs=1000 sw=10 sl=64 | p99 | 1,155,097 us | 1,136,262 us | 1,054,298 us | +1.7% | +9.6% |
Raw CSV
config_idx,batch_size,schema_width,string_len,num_batches,total_ms,total_tuples,total_bytes,tuples_per_sec,mb_per_sec,lat_p50_us,lat_p95_us,lat_p99_us
0,10,10,64,20,508.41,200,128000,393,0.240,23825.77,35435.71,35435.71
1,100,10,64,20,2516.12,2000,1280000,795,0.485,125414.36,147122.46,147122.46
2,1000,10,64,20,22144.52,20000,12800000,903,0.551,1110980.15,1155097.24,1155097.24
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
2 participants
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.
What changes were proposed in this PR?
LoopIntegrationSpecis the most expensive suite in theamber-integrationCI job. Two changes, both spec-only (no engine change):i < len(table))Every loop iteration re-executes its region and respawns every worker in it (~2s per Python worker), so the two nested cases dominate the suite's cost; 2x2 halves their iterations while keeping the coverage that matters — nesting depth drives the
loop_counter/loop_start_idenvelope, and each loop level still takes its back-edge at least twice across the run. The single-loop cases keep the 3-row input so one loop takes the same back-edge twice consecutively.The deadline matters on macOS: the suite intermittently hangs mid-run there (observed in 3 of 4 attempts across two 2026-07-29 runs, e.g. this failed job), burning the whole deadline before
withRetryre-runs the test. 90s is ~2.5x the slowest healthy case on the 3-core macOS runner, so a hang now costs ~1.5 min less per occurrence. Root-causing the hang itself is tracked in #7056.Any related issues, documentation, discussions?
Closes #7052. Suite added in #5700; the macOS hang is tracked in #7056.
How was this PR tested?
LoopIntegrationSpecassertions, updated for the 2x2 counts (outer 9→4, inner 3→2), pass in this PR'samber-integrationCI job on both ubuntu and macOS.sbt WorkflowExecutionService/scalafmtCheckAllpasses locally.Was this PR authored or co-authored using generative AI tooling?
Generated-by: Claude Code (Claude Fable 5)