[None][test] default the disagg cache-transceiver precheck to off

[chuangz0]

Description

Waive the disaggregated perf-sanity cache-transceiver precheck by flipping its default to opt-in, until the gate is validated on the post-merge stages.

The precheck is a launch-script gate (slurm_ct_precheck_gate.sh, driven by ctPrecheckEnabled), not a pytest case, so it cannot be waived via tests/integration/test_lists/waives.txt — this default flip is the waive. None of the 90 disagg perf-sanity yamls set cache_transceiver_precheck, so with this change the precheck is disabled on every stage and the real tests run unaffected.

Re-enabling paths:

• per-test yaml: cache_transceiver_precheck: enabled: true (pilot a config)
• globally for a validation run: TRTLLM_DISAGG_CT_PRECHECK=1 (env overrides yaml either way)
• once validated, a follow-up PR flips the default back to on

Test Coverage

• tests/unittest/others/test_cache_transceiver_precheck_config.py (32 passed): added an assertion that an empty config resolves to ctPrecheckEnabled=0; explicit yaml opt-in and the env kill-switch behavior are unchanged and still covered.

PR Checklist

• PR title and description are self-explanatory
• Test coverage added/updated
• Docs updated (precheck README)

🤖 Generated with Claude Code

Dev Engineer Review

• Changed ctPrecheckEnabled to default to 0.
• Preserved YAML opt-in through cache_transceiver_precheck.enabled: true.
• Preserved TRTLLM_DISAGG_CT_PRECHECK environment-variable overrides.
• Updated the README to document the default-disabled behavior.
• No public API changes or unrelated configuration changes were identified.

QA Engineer Review

• Modified tests/unittest/others/test_cache_transceiver_precheck_config.py.
• Updated coverage to verify that an empty configuration disables the precheck.
• Existing coverage continues to verify explicit YAML opt-in and environment-variable behavior.
• No files under tests/integration/test_lists/, test-db/, or qa/ were modified.
• The unit test is not listed in the integration test lists.
• Verdict: sufficient.

[coderabbitai]

Walkthrough

The cache-transceiver precheck is disabled by default. YAML can enable it with enabled: true. The environment variable remains an override. Documentation and unit tests now cover the new default and explicit opt-in.

Changes

Cache transceiver precheck opt-in

Layer / File(s)	Summary
Default precheck policy tests/scripts/perf-sanity/cache_transceiver_precheck/precheck_config.py	The YAML fallback now disables the precheck. The environment variable continues to override YAML settings.
Opt-in documentation and validation tests/scripts/perf-sanity/cache_transceiver_precheck/README.md, tests/unittest/others/test_cache_transceiver_precheck_config.py	The README documents explicit YAML opt-in. Tests cover YAML enablement and the disabled empty-configuration default.

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

Possibly related PRs

• NVIDIA/TensorRT-LLM#17137: Modifies the same precheck configuration, README, and tests for model-root propagation and transfer logic.
• NVIDIA/TensorRT-LLM#17223: Modifies the same precheck configuration and tests for model-root propagation and KV ownership.

Suggested reviewers: bowenfu, dhansen-nvidia, hyukn

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and concisely states that the disaggregated cache-transceiver precheck now defaults to disabled.
Description check	✅ Passed	The description explains the change, affected behavior, re-enabling paths, test coverage, and documentation updates.
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.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@tests/scripts/perf-sanity/cache_transceiver_precheck/README.md`:
- Around line 29-37: Add the standard NVIDIA copyright header at the beginning
of the modified README, before the existing title, and set its copyright year to
the latest meaningful modification year.

🪄 Autofix

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 92a33b94-5c7c-4886-86a7-4040e5bd6d7b

📥 Commits

Reviewing files that changed from the base of the PR and between b775e80 and 5937d4b.

📒 Files selected for processing (3)

• tests/scripts/perf-sanity/cache_transceiver_precheck/README.md
• tests/scripts/perf-sanity/cache_transceiver_precheck/precheck_config.py
• tests/unittest/others/test_cache_transceiver_precheck_config.py

[chuangz0]

/bot skip --comment "waive some test"

[tensorrt-cicd]

PR_Github #64206 [ skip ] triggered by Bot. Commit: 5937d4b Link to invocation

[tensorrt-cicd]

PR_Github #64206 [ skip ] completed with state SUCCESS. Commit: 5937d4b
Skipping testing for commit 5937d4b

Link to invocation