Motivation

Uvicorn access logs can get flooded by high-frequency polling endpoints (especially /metrics which is printed every 5 seconds, leads to an overload of logs), making operational logs noisy and harder to use for debugging.

Similar work is also being worked on in vLLM, see #29023.

Modifications

Adds configurable path-based filtering for Uvicorn access logs to suppress noisy endpoints.
Introduces/uses ServerArgs knobs:

• --uvicorn-access-log-exclude-prefixes ...: a list of path prefixes to exclude from access logs (space-separated).
• --disable-uvicorn-access-log-filter: disables the feature entirely and restores the original behavior.

(By default, --uvicorn-access-log-exclude-prefixes is set to /metrics)

Usage examples

• Exclude both /metrics(by default):
  --uvicorn-access-log-exclude-prefixes /metrics
• Exclude both /metrics and /health:
  --uvicorn-access-log-exclude-prefixes /metrics /health
• Disable filtering:
  --disable-uvicorn-access-log-filter

Accuracy Tests

Benchmarking and Profiling

Checklist

• Format your code according to the Format code with pre-commit.
• Add unit tests according to the Run and add unit tests.
• Update documentation according to Write documentations.
• Provide accuracy and speed benchmark results according to Test the accuracy and Benchmark the speed.
• Follow the SGLang code style guidance.
• Work with maintainers to merge your PR. See the PR Merge Process