[waizuichougou]

Overview

Three tests fail on main: test_partial_tail_store_uses_attention_and_recurrent_cow_sources and both test_eagle_partial_tail_round_trip_serves_mamba_checkpoint cases. After #62613 made Mamba events report the checkpoint block size and predecessor hash, the first test still expected placeholder values. After #63249 changed load-job keys to an ordered tuple, the EAGLE tests still compared that tuple with a set. This PR updates those stale expectations.

Claims

• Assert the Mamba checkpoint event's block size and parent hash as emitted by the event tracker.
• Check loaded offload keys in prefix order, matching the scheduler's tuple contract.

Validation

pytest -v -s tests/v1/kv_connector/unit/offloading_connector/test_scheduler.py \
  -k 'partial_tail_store_uses_attention_and_recurrent_cow_sources or eagle_partial_tail_round_trip_serves_mamba_checkpoint'
# 3 passed

pytest -v -s tests/v1/kv_connector/unit/offloading_connector/test_scheduler.py
# 467 passed

pytest -v -s tests/v1/kv_connector/unit/offloading_connector/test_events.py
# 55 passed

On latest main (5d55ef1590), the focused run reports 3 failed, 464 deselected; with these expectation updates it reports 3 passed, 464 deselected. Ruff, formatting, and git diff --check pass.

Details

The Mamba event tracker emits the checkpoint block size (4) and links the checkpoint hash to its predecessor; the scheduler test still expected the older placeholder values. A separate EAGLE round-trip assertion compared the load job's ordered key tuple with a set, although the scheduler preserves prefix order. This PR changes test assertions only; it does not change CI configuration or test markers.

Duplicate-work check: open PR searches found no change for these stale scheduler assertions. Open PR #63550 adds KV event layout coverage in tests/v1/core/test_kv_cache_events_layouts.py; it does not modify or cover these offloading scheduler expectations. Open PR #59177 addresses a separate flaky XPU event test.

---

Pull Request Checklist

• I used vLLM's /pr-checklist skill. (Mandatory for agents, optional for humans).

• AI assistance was used during the creation of this PR.

• Design Fit: Keeps changes within the affected scheduler tests and follows current event and ordering contracts.

• Testing and Validation: Reproduces the failures on main and passes the focused and full test files.

• Code Quality and Style: Updates only stale test assertions; no production code or user documentation changes are needed.

• Pull Request Contents: Follows the repository template, documents root cause, validation, and duplicate-work checks.

[claude]

Claude Code Review

This pull request is from a fork — automated review is disabled. A repository maintainer can comment @claude review to run a one-time review.

[orozery]

/ci run

[github-actions]

✅ @waizuichougou, CI is now available for this PR.

• /ci run starts upstream CI; /amd-ci run starts AMD CI only.
• Run commands test your branch with every commit currently on its upstream target branch. If it is behind, the bot merges the target branch into it first (pull before pushing again). Append --allow-stale to a run command to test an outdated branch as is, at your own risk.
• /ci retry retries failed jobs in the CI build for the current PR head. If the current head has no CI build, it starts a new CI build for the current head containing only jobs that failed in the latest earlier CI build for this PR.
• /amd-ci retry retries failed jobs in AMD CI for the current PR head. Use /amd-ci run when the current head has no AMD CI build.
• /ci cancel cancels scheduled or running CI builds for this PR branch; /amd-ci cancel does the same for AMD CI only.

[github-actions]

✅ Triggered Buildkite CI #99183 for commit b7cec6c89a2b.

ℹ️ This PR is 43 commits behind upstream main. Merged upstream main into this branch as b7cec6c89a2b before starting CI. Pull the branch before pushing more changes.