Skip to content

[https://nvbugs/6463814][chore] Unwaive passing perf-sanity disagg tests - #16655

Merged
chenfeiz0326 merged 1 commit into
NVIDIA:mainfrom
chenfeiz0326:unwaive-6463814-perf-sanity
Jul 21, 2026
Merged

[https://nvbugs/6463814][chore] Unwaive passing perf-sanity disagg tests#16655
chenfeiz0326 merged 1 commit into
NVIDIA:mainfrom
chenfeiz0326:unwaive-6463814-perf-sanity

Conversation

@chenfeiz0326

@chenfeiz0326 chenfeiz0326 commented Jul 21, 2026

Copy link
Copy Markdown
Collaborator

Summary by CodeRabbit

  • Tests
    • Updated performance test waiver coverage for selected B200 and GB200 DeepSeek V32 and Kimi K25 configurations.
    • Removed obsolete waiver entries and refreshed references for retained scenarios.

Description

Bug 6463814 waived 7 GB200 disaggregated perf-sanity cases after LLM/main/L0_PostMerge #2839. Re-triage found the only real failure was aggr_upload-qwen3_5_397b_fp4_blackwell-qwen3_5_397b_fp4_tep8_1k1k (GDN decode "Tensor data pointer is not aligned to 32 bytes" error), which is now tracked separately in sub-bug 6477124 and is not part of this bug's waiver set.

The 7 disagg_upload GB200 NIXL cases waived under 6463814 were auto-flagged flaky ("Test terminated unexpectedly") and are passing again, so this PR unwaives them:

  • disagg_upload-e2e-gb200_deepseek-r1-fp4_8k1k_con4096_ctx1_dep4_gen1_dep16_eplb0_mtp1_ccb-NIXL
  • disagg_upload-e2e-gb200_deepseek-v32-fp4_32k4k_con2048_ctx1_dep4_gen1_dep32_eplb288_mtp1_ccb-NIXL
  • disagg_upload-e2e-gb200_deepseek-v32-fp4_8k1k_con4096_ctx1_dep4_gen1_dep32_eplb256_mtp0_ccb-NIXL
  • disagg_upload-e2e-gb200_kimi-k25-thinking-fp4_8k1k_con4096_ctx1_dep4_gen1_dep16_eplb0_mtp0_ccb-NIXL
  • disagg_upload-gen_only-gb200_deepseek-r1-fp4_128k8k_con128_ctx1_pp8_gen1_dep16_eplb0_mtp1_ccb-NIXL
  • disagg_upload-gen_only-gb200_deepseek-v32-fp4_8k1k_con1024_ctx1_dep4_gen1_dep32_eplb256_mtp3_ccb-NIXL
  • disagg_upload-gen_only-gb200_kimi-k25-thinking-fp4_8k1k_con4096_ctx1_dep4_gen1_dep16_eplb0_mtp0_ccb-NIXL

Test Coverage

Removing a waiver must not skip the post-merge stages that cover these cases. The unwaived cases live on these GB200 disagg perf-sanity post-merge stages (from #2839):

  • GB200-8_GPUs-2_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE1-GPU4-Post-Merge-2
  • GB200-36_GPUs-9_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE8-GPU32-Post-Merge-1
  • GB200-36_GPUs-9_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE8-GPU32-Post-Merge-2
  • GB200-36_GPUs-9_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE8-GPU32-Post-Merge-9
  • GB200-24_GPUs-6_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE2-GPU8-GEN1-NODE4-GPU16-Post-Merge-2
  • GB200-20_GPUs-5_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE4-GPU16-Post-Merge-1
  • GB200-20_GPUs-5_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE4-GPU16-Post-Merge-2
  • GB200-20_GPUs-5_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE4-GPU16-Post-Merge-4

Maintainer: please trigger the corresponding post-merge stages when running CI, e.g.:

/bot run --stage-list "GB200-8_GPUs-2_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE1-GPU4-Post-Merge-2, GB200-36_GPUs-9_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE8-GPU32-Post-Merge-1, GB200-36_GPUs-9_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE8-GPU32-Post-Merge-2, GB200-36_GPUs-9_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE8-GPU32-Post-Merge-9, GB200-24_GPUs-6_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE2-GPU8-GEN1-NODE4-GPU16-Post-Merge-2, GB200-20_GPUs-5_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE4-GPU16-Post-Merge-1, GB200-20_GPUs-5_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE4-GPU16-Post-Merge-2, GB200-20_GPUs-5_Nodes-PyTorch-Disagg-PerfSanity-CTX1-NODE1-GPU4-GEN1-NODE4-GPU16-Post-Merge-4"

PR Checklist

  • PR description clearly explains what and why.

This is a waiver-list-only change (tests/integration/test_lists/waives.txt); no source or API changes.

Bug 6463814 waived 7 GB200 disagg perf-sanity cases after L0_PostMerge
NVIDIA#2839. Re-triage found the only real failure was
aggr_upload-qwen3_5_397b_fp4_blackwell-qwen3_5_397b_fp4_tep8_1k1k
(GDN decode tensor-alignment error), now tracked separately in
sub-bug 6477124. The 7 disagg_upload GB200 NIXL cases waived under
6463814 are passing again, so unwaive them:

- disagg_upload-e2e-gb200_deepseek-r1-fp4_8k1k_con4096_ctx1_dep4_gen1_dep16_eplb0_mtp1_ccb-NIXL
- disagg_upload-e2e-gb200_deepseek-v32-fp4_32k4k_con2048_ctx1_dep4_gen1_dep32_eplb288_mtp1_ccb-NIXL
- disagg_upload-e2e-gb200_deepseek-v32-fp4_8k1k_con4096_ctx1_dep4_gen1_dep32_eplb256_mtp0_ccb-NIXL
- disagg_upload-e2e-gb200_kimi-k25-thinking-fp4_8k1k_con4096_ctx1_dep4_gen1_dep16_eplb0_mtp0_ccb-NIXL
- disagg_upload-gen_only-gb200_deepseek-r1-fp4_128k8k_con128_ctx1_pp8_gen1_dep16_eplb0_mtp1_ccb-NIXL
- disagg_upload-gen_only-gb200_deepseek-v32-fp4_8k1k_con1024_ctx1_dep4_gen1_dep32_eplb256_mtp3_ccb-NIXL
- disagg_upload-gen_only-gb200_kimi-k25-thinking-fp4_8k1k_con4096_ctx1_dep4_gen1_dep16_eplb0_mtp0_ccb-NIXL

Signed-off-by: Chenfei Zhang <chenfeiz@nvidia.com>
@coderabbitai

coderabbitai Bot commented Jul 21, 2026

Copy link
Copy Markdown
Contributor

Review Change Stack

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info
⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: baab0aef-0140-4061-b91a-06b943e3ac12

📥 Commits

Reviewing files that changed from the base of the PR and between ff976d8 and 1a57471.

📒 Files selected for processing (1)
  • tests/integration/test_lists/waives.txt
💤 Files with no reviewable changes (1)
  • tests/integration/test_lists/waives.txt

📝 Walkthrough

Walkthrough

Updated integration test waivers by removing obsolete disaggregated upload entries and adding four configuration-specific waivers with updated nvbugs references.

Changes

Integration waiver updates

Layer / File(s) Summary
Refresh disaggregated upload waivers
tests/integration/test_lists/waives.txt
Replaces obsolete entries with four narrower DeepSeek V32 and Kimi K25 configuration waivers, line-number markers, and updated nvbugs IDs.

Estimated code review effort: 1 (Trivial) | ~2 minutes

Possibly related PRs

Suggested reviewers: crazydemo

🚥 Pre-merge checks | ✅ 5
✅ Passed checks (5 passed)
Check name Status Explanation
Title check ✅ Passed The title matches the waiver-list cleanup and clearly states the main change with a valid ticket and type.
Description check ✅ Passed The description includes the required sections, explains the reason and change, and provides relevant test coverage.
Docstring Coverage ✅ Passed No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check ✅ Passed Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check ✅ Passed Check skipped because no linked issues were found for this pull request.
✨ Finishing Touches
🧪 Generate unit tests (beta)
  • Create PR with unit tests

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

@chenfeiz0326
chenfeiz0326 requested a review from mzweilz July 21, 2026 03:48
@chenfeiz0326

Copy link
Copy Markdown
Collaborator Author

/bot skip --comment "Only unwaive perf tests, no need to run the whole CI pipeline"

@chenfeiz0326
chenfeiz0326 enabled auto-merge (squash) July 21, 2026 03:49
@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #60564 [ skip ] triggered by Bot. Commit: 1a57471 Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #60564 [ skip ] completed with state SUCCESS. Commit: 1a57471
Skipping testing for commit 1a57471

Link to invocation

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.

3 participants