python-web-troubleshooter
Installation
SKILL.md
Python Web Service Troubleshooter
Structured diagnosis for Python web service (Gunicorn / uWSGI) performance and capacity issues.
Core principle: workers are buffers, replicas are scale-out, and latency is what actually determines capacity.
Step 1: Break Down Latency First
Before changing any config, collect these metrics to build a complete picture of where time is being spent:
| Layer | Key Metrics |
|---|---|
| Gunicorn | idle worker count vs working worker count |
| Nginx | upstream_response_time, upstream_connect_time (check via APM or access log dashboard) |
| Application | total request duration, queue wait time as % of total |
| Database | DB CPU, query latency, DB load, waiting on IO? |