[TRTLLM-15264][doc] Kimi K3 disagg: stop recommending a UCX_TLS pin by default

[brnguyen2]

Description

Follow-up to review feedback on #17334 (#17334 (comment)).

The K3 disagg README exported UCX_TLS=tcp,self,sm,cuda_copy,cuda_ipc as a launch prerequisite, and the benchmark config hard-coded the same pin in worker_env_var. That transport list excludes RDMA/verbs, so carrying it to clusters whose inter-node communication depends on verbs can fail initialization or silently force traffic onto TCP.

This change makes no pin the default:

• The launch-sequence prerequisite is now "leave UCX_TLS unset" (UCX selects transports itself, RDMA/verbs included), with a pointer to the troubleshooting caveats.
• benchmark_kimi_k3_dep16.yaml no longer carries TRTLLM_WORKER_UCX_TLS in worker_env_var; the harness's start_worker.sh already clears any container-provided UCX_TLS (a container-default UCX_TLS=tcp breaks V2 NIXL VRAM registration), so workers default to UCX's own selection.
• The transport override is documented in the troubleshooting caveats only, scoped to clusters whose verbs transports cannot initialize, with an explicit warning that the list excludes RDMA/verbs.

Test Coverage

Documentation/example-config change only; no code paths affected.

PR Checklist

• PR title follows the [TRTLLM-15264][doc] convention
• pre-commit run on the touched files

Dev Engineer Review

• Updated Kimi K3 disaggregated deployment guidance to leave UCX_TLS unset by default.
• Removed TRTLLM_WORKER_UCX_TLS from the example worker environment.
• Documented the fallback override for clusters where verbs transports cannot initialize.
• Configuration changes are scoped to documentation and an example YAML file.
• No code, API, performance, or error-handling changes were made.
• No test-list changes were made.

QA Engineer Review

No test changes.

[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: 192090cb-4336-455d-9d52-bd77ca1034cf

📥 Commits

Reviewing files that changed from the base of the PR and between c67879c and b5fac8a.

📒 Files selected for processing (2)

• examples/kimi_k3/disagg/README.md
• examples/kimi_k3/disagg/benchmark_kimi_k3_dep16.yaml

---

Walkthrough

The Kimi K3 disaggregated deployment guidance now leaves UCX transport selection automatic by default. It documents clearing container-provided settings and using TRTLLM_WORKER_UCX_TLS only when verbs initialization fails.

Changes

UCX transport guidance

Layer / File(s)	Summary
Transport selection and fallback policy examples/kimi_k3/disagg/README.md	Manual and cluster guidance now leaves UCX_TLS unset by default, rejects container-provided UCX_TLS=tcp for V2 NIXL VRAM registration, and documents the TCP/CUDA fallback for verbs initialization failures.
Worker override wiring examples/kimi_k3/disagg/README.md, examples/kimi_k3/disagg/benchmark_kimi_k3_dep16.yaml	SLURM startup clears container-provided UCX_TLS and reapplies an optional TRTLLM_WORKER_UCX_TLS value. The benchmark configuration no longer sets the fallback override by default.

Estimated code review effort: 2 (Simple) | ~10 minutes

Suggested reviewers: bowenfu

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title follows the required format and clearly summarizes the main documentation change.
Description check	✅ Passed	The description includes the required sections and clearly explains the change and 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.}

[brnguyen2]

/bot run

[tensorrt-cicd]

PR_Github #65139 [ run ] triggered by Bot. Commit: b5fac8a Link to invocation

[tensorrt-cicd]

PR_Github #65139 [ run ] completed with state FAILURE. Commit: b5fac8a
/LLM/main/L0_MergeRequest_PR pipeline #52935 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

[brnguyen2]

/bot run

[tensorrt-cicd]

PR_Github #65190 [ run ] triggered by Bot. Commit: b5fac8a Link to invocation

[tensorrt-cicd]

PR_Github #65190 [ run ] completed with state SUCCESS. Commit: b5fac8a
/LLM/main/L0_MergeRequest_PR pipeline #52979 completed with status: 'SUCCESS'

CI Report

Link to invocation

[nv-guomingz]

LGTM