[None][test] Waive qwen3_30b_a3b_fp8 pd_disagg mm-encoder test on single-GPU (NIXL setup failure)

[venkywonka]

Summary

Waive test_mm_encoder_standalone.py::test_single_request_chat_multiple_images[pd_disagg-qwen3_30b_a3b_fp8], which fails fleet-wide on the pre-merge DGX_B200-PyTorch single-GPU stage.

NVBug: https://nvbugs/6269683

Failure

At test setup, constructing the pd_disagg LLM's disaggregated KV-cache transfer agent hits a hard NIXL assertion:

RuntimeError: [TensorRT-LLM][ERROR] Assertion failed: status == NIXL_SUCCESS
  (.../executor/cache_transmission/nixl_utils/transferAgent.cpp:614)
  AgentConnectionManager::AgentConnectionManager(...) -> CacheTransceiver::CacheTransceiver(...)
RuntimeError: Executor worker returned error

i.e. NIXL registerMem fails while pinning the transceiver's transfer buffers during executor-worker construction on a single GPU. The sibling no_pd_disagg / raw_inputs sub-cases pass (they never build a NIXL CacheTransceiver).

Exact failing builds (CI Report)

Not specific to any one PR — the same case fails identically across many unrelated pipelines (stage DGX_B200-PyTorch, single-GPU):

• PR [TRTLLM-11457][feat] Async Ulysses pipeline (Enabled for LTX-2 + WAN) #13978, build 41402 — http://tensorrt-llm.tensorrt-llm-ci-report.sc2-paas.nvidia.com/?job=%2FLLM%2Fmain%2FL0_MergeRequest_PR&build=41402
• PR [None][fix] Fix AutoDeploy accuracy tests #13925, build 41352 — http://tensorrt-llm.tensorrt-llm-ci-report.sc2-paas.nvidia.com/?job=%2FLLM%2Fmain%2FL0_MergeRequest_PR&build=41352
• PR [None][fix] Stabilize Mamba replay state update #14841, build 41381 — http://tensorrt-llm.tensorrt-llm-ci-report.sc2-paas.nvidia.com/?job=%2FLLM%2Fmain%2FL0_MergeRequest_PR&build=41381
• PR [TRTLLM-12842][feat] Maximal LLMAPI capture in usage telemetry #14398, build 41375 — http://tensorrt-llm.tensorrt-llm-ci-report.sc2-paas.nvidia.com/?job=%2FLLM%2Fmain%2FL0_MergeRequest_PR&build=41375
• Post-merge (main), L0_PostMerge build 2759 — http://tensorrt-llm.tensorrt-llm-ci-report.sc2-paas.nvidia.com/?job=%2FLLM%2Fmain%2FL0_PostMerge&build=2759
• This PR ([TRTLLM-12467][feat] EPD improvements #13864), build 41351 — http://tensorrt-llm.tensorrt-llm-ci-report.sc2-paas.nvidia.com/?job=%2FLLM%2Fmain%2FL0_MergeRequest_PR&build=41351

Also observed on #14599 (41365), #14524 (41373), #14941 (41337), #14908 (41360), #14948 (41357), #12958 (41363); related Hang detected on rank 0 in PyExecutor on #14812/#14891. Passes only intermittently (~1/3, node-dependent) — consistent with single-GPU NIXL EPD-disagg not being reliably supported on the DGX_B200 single-GPU nodes (no IB/RoCE fabric and/or memory pressure from the multi-instance disagg fixture).

Scope

• One-line waives.txt entry; no source/product changes.
• Stopgap until the single-GPU NIXL EPD-disagg path is fixed, or the pd_disagg variant is gated to multi-GPU (tracked in https://nvbugs/6269683).

Summary by CodeRabbit

• Tests
  • Updated test configuration to skip a specific test case.

[venkywonka]

/bot run

[coderabbitai]

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: efeb623d-b322-4a2c-a8de-fa463083cb1a

📥 Commits

Reviewing files that changed from the base of the PR and between 33b0a32 and 0e10683.

📒 Files selected for processing (1)

• tests/integration/test_lists/waives.txt

---

📝 Walkthrough

Walkthrough

This PR adds a single waiver entry to the test skip list for a failing multimodal encoder test case parameterized with pd_disagg-qwen3_30b_a3b_fp8, tracked by nvbug 6269683.

Changes

Multimodal encoder test waiver

Layer / File(s)	Summary
Test waiver entry for multimodal encoder tests/integration/test_lists/waives.txt	Added a SKIP waiver entry for the test_single_request_chat_multiple_images[pd_disagg-qwen3_30b_a3b_fp8] test case in the multimodal encoder unit tests, referencing nvbug 6269683.

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~2 minutes

Possibly related PRs

• NVIDIA/TensorRT-LLM#14854: Updates tests/integration/test_lists/waives.txt by adding SKIP/waiver entries for failing tests using the same mechanism.
• NVIDIA/TensorRT-LLM#14883: Modifies tests/integration/test_lists/waives.txt by adding new SKIP/waiver entries for failing tests, including waiving nvbug 6269683-related cases.
• NVIDIA/TensorRT-LLM#14789: Modifies the same tests/integration/test_lists/waives.txt file by adding SKIP waiver entries for failing multimodal/openai-disagg/multigpu test cases.

Suggested reviewers

• crazydemo
• jieli-matrix
• LarryXFly
• StanleySun639
• xinhe-nv

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The pull request title clearly and specifically describes the main change: waiving a specific test case for qwen3_30b_a3b_fp8 with pd_disagg on single-GPU due to NIXL setup failure.
Description check	✅ Passed	The pull request description provides a comprehensive explanation of the issue, failure details, exact failing builds with CI links, and scope, exceeding the template requirements for clarity and justification.
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.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

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

[tensorrt-cicd]

PR_Github #52139 [ run ] triggered by Bot. Commit: 0e10683 Link to invocation

[tensorrt-cicd]

PR_Github #52139 [ run ] completed with state FAILURE. Commit: 0e10683
/LLM/main/L0_MergeRequest_PR pipeline #41463 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

• Please check the failed tests and fix your PR
• If you cannot view the failures, ask the CI triggerer to share details
• Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

[venkywonka]

/bot run

[tensorrt-cicd]

PR_Github #52157 [ run ] triggered by Bot. Commit: 0e10683 Link to invocation

[tensorrt-cicd]

PR_Github #52157 [ run ] completed with state SUCCESS. Commit: 0e10683
/LLM/main/L0_MergeRequest_PR pipeline #41479 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

• Please check the failed tests and fix your PR
• If you cannot view the failures, ask the CI triggerer to share details
• Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation