fix(grpc_servicer): handle vllm log forwarding on servicer side

[CatherineSue]

Description

Problem

vllm's grpc_server.py manually attaches the smg_grpc_servicer logger to the vllm logging hierarchy at module level. This means vllm has to know about and configure third-party loggers, which was flagged in vllm-project/vllm#38333 (comment).

Solution

Move the logger attachment into the smg_grpc_servicer.vllm subpackage __init__.py so it happens automatically when the servicer is imported. Also switch health_servicer.py from logging.getLogger to vllm.logger.init_logger for consistency with servicer.py.

Changes

• grpc_servicer/smg_grpc_servicer/vllm/__init__.py: Attach the top-level smg_grpc_servicer logger to the vllm logging hierarchy on import.
• grpc_servicer/smg_grpc_servicer/vllm/health_servicer.py: Use vllm.logger.init_logger instead of logging.getLogger.

Test Plan

• Start vLLM with --grpc and send requests
• Verify servicer INFO messages (e.g. Generate request ...) still appear with vLLM's log format
• Verify health check log messages also use vLLM's format

Checklist

• cargo +nightly fmt passes
• cargo clippy --all-targets --all-features -- -D warnings passes
• (Optional) Documentation updated
• (Optional) Please join us on Slack #sig-smg to discuss, review, and merge PRs

Summary by CodeRabbit

• Chores
  • Synchronized logging configuration so package logs use the same handlers and log level as the vllm service, ensuring consistent formatting and output across components.
  • Aligned the health-check service logger with vllm's logging setup to improve observability and unify log behavior.

[coderabbitai]

📝 Walkthrough

Walkthrough

Centralized logging setup added: the package logger for smg_grpc_servicer is reconfigured to mirror vllm logger handlers and level; propagation disabled. One health servicer now initializes its logger via vllm.logger.init_logger(__name__) instead of logging.getLogger(__name__).

Changes

Cohort / File(s)	Summary
Logging configuration setup grpc_servicer/smg_grpc_servicer/vllm/__init__.py	Adds logging import and copies vllm logger handlers and level to the smg_grpc_servicer package logger, and sets propagate = False.
Logger initialization update grpc_servicer/smg_grpc_servicer/vllm/health_servicer.py	Replaces logging.getLogger(__name__) with vllm.logger.init_logger(__name__) for the module logger.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Possibly related PRs

• feat(grpc): add VllmHealthServicer for standard gRPC health checking (grpc.health.v1) #885 — Modifies the same vllm module files and introduces a VllmHealthServicer; closely related code locations.

Suggested reviewers

• slin1237

Poem

🐇 Hop by hop, I checked the log trace,
Handlers aligned in a cozy place.
vllm whispers, package obeys,
Quiet, tidy log-filled days.
🎉

🚥 Pre-merge checks | ✅ 3

✅ Passed checks (3 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title 'fix(grpc_servicer): handle vllm log forwarding on servicer side' directly and clearly describes the main change: implementing vLLM log forwarding configuration on the servicer side rather than in vLLM itself.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

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

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/servicer-log-forwarding

---

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

[gemini-code-assist]

Code Review

This pull request integrates the smg_grpc_servicer logging with the vllm logging hierarchy and updates the health servicer to use vllm.logger.init_logger. A suggestion was made to copy the logger handlers instead of referencing them directly to avoid unintended side effects from future modifications to the shared handler list.

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@grpc_servicer/smg_grpc_servicer/vllm/__init__.py`:
- Around line 11-15: The current logger setup aliases _pkg_logger.handlers to
_vllm_logger.handlers and unconditionally sets _pkg_logger.propagate = False;
instead, copy _vllm_logger.handlers into _pkg_logger.handlers (so changes to
vllm handlers later don't affect our logger) and only disable propagation when
there are handlers present on _pkg_logger (i.e., set propagate=False if the
copied handlers list is non-empty, otherwise leave propagation unchanged) —
update the initialization that references _vllm_logger and _pkg_logger to
perform a shallow copy of handlers and a conditional propagate assignment.

🪄 Autofix (Beta)

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: Organization UI

Review profile: ASSERTIVE

Plan: Pro

Run ID: 458bc0f1-e452-4604-b4c2-60e0704334eb

📥 Commits

Reviewing files that changed from the base of the PR and between b8a8be9 and 750d9ca.

📒 Files selected for processing (2)

• grpc_servicer/smg_grpc_servicer/vllm/__init__.py
• grpc_servicer/smg_grpc_servicer/vllm/health_servicer.py

[coderabbitai]

♻️ Duplicate comments (1)

grpc_servicer/smg_grpc_servicer/vllm/__init__.py (1)

15-15: ⚠️ Potential issue | 🟠 Major

Guard propagate based on handler availability.

_pkg_logger.propagate = False is still unconditional. If vllm has zero handlers at import time, smg_grpc_servicer logs can be suppressed. Set propagate to False only when copied handlers are present.

Suggested fix

 _pkg_logger.handlers = list(_vllm_logger.handlers)
 _pkg_logger.setLevel(_vllm_logger.level)
-_pkg_logger.propagate = False
+if _pkg_logger.handlers:
+    _pkg_logger.propagate = False

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@grpc_servicer/smg_grpc_servicer/vllm/__init__.py` at line 15, The package
logger currently unconditionally sets _pkg_logger.propagate = False which can
suppress host app logs if vllm has no handlers; change the logic in
vllm/__init__.py where _pkg_logger handlers are copied so that you only set
_pkg_logger.propagate = False when there are copied/attached handlers (e.g.,
check if _pkg_logger.handlers or the list returned by copy of handlers is
non-empty) and leave propagate True otherwise; update the block that copies
handlers and the propagate assignment (referencing _pkg_logger and the
handler-copying code) so propagate is conditional on handlers being present.

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Duplicate comments:
In `@grpc_servicer/smg_grpc_servicer/vllm/__init__.py`:
- Line 15: The package logger currently unconditionally sets
_pkg_logger.propagate = False which can suppress host app logs if vllm has no
handlers; change the logic in vllm/__init__.py where _pkg_logger handlers are
copied so that you only set _pkg_logger.propagate = False when there are
copied/attached handlers (e.g., check if _pkg_logger.handlers or the list
returned by copy of handlers is non-empty) and leave propagate True otherwise;
update the block that copies handlers and the propagate assignment (referencing
_pkg_logger and the handler-copying code) so propagate is conditional on
handlers being present.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Pro

Run ID: e4b058ff-98f8-418f-93fa-436a6f29b599

📥 Commits

Reviewing files that changed from the base of the PR and between 750d9ca and 82fbaf2.

📒 Files selected for processing (1)

• grpc_servicer/smg_grpc_servicer/vllm/__init__.py