[CI] fix blend-server venv #3099
Conversation
Signed-off-by: deng451e <838677410@qq.com>
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 1 potential issue.
❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.
Reviewed by Cursor Bugbot for commit ea9f10a. Configure here.
There was a problem hiding this comment.
Code Review
This pull request updates the CI environment by relocating logs to the repository root for better visibility and adjusting the Python execution environment for the blend server. It also introduces a handler for REPORT_BLOCK_ALLOCATION requests. A review comment identifies that this request type is incorrectly assigned to the affinity thread pool; it should be moved to the normal thread pool to avoid consuming resources intended for GPU-bound tasks and to ensure architectural consistency.
What this PR does / why we need it:
Special notes for your reviewers:
If applicable:
Note
Medium Risk
Moderate risk because it changes CI dependency resolution (drops explicit torch pin) and the runtime environment used to launch the blend server, which could surface version/compatibility issues during integration runs.
Overview
Runs the LMCache blend server using the nightly/test venv (
TEST_PYTHON) instead of the default image venv, aligning the blend test components with the intended vLLM/LMCache environment.Updates the blend CI environment setup to install nightly
vllmwithout an explicittorch==...pin, and enhances the blend test harness to write per-process logs under a repo-rootlogs_<BUILD_ID>/directory (with a merged artifact still produced for CI).Wires
RequestType.REPORT_BLOCK_ALLOCATIONintoblend_server_v2(handler registration + normal thread pool) so observability reports are accepted by the blend server.Reviewed by Cursor Bugbot for commit a96556c. Bugbot is set up for automated code reviews on this repo. Configure here.