Summary

Track desired granian features that exist in gunicorn but are missing in granian. These would help with SSE workloads and prevent resource leaks.

Requested Features

1. --max-requests (worker recycling after N requests)

Gunicorn equivalent: --max-requests

Restart workers after handling N requests. Useful for preventing memory leaks and clearing stale connections.

# Gunicorn
gunicorn --max-requests 10000 app:app

# Desired in Granian
granian --max-requests 10000 app:app

2. --max-requests-jitter (randomized restart timing)

Gunicorn equivalent: --max-requests-jitter

Add random jitter to --max-requests or --workers-lifetime to prevent all workers from restarting simultaneously (thundering herd).

# Gunicorn - restarts between 9000-11000 requests
gunicorn --max-requests 10000 --max-requests-jitter 1000 app:app

# Desired in Granian
granian --workers-lifetime 3600 --workers-lifetime-jitter 600 app:app
# Or
granian --max-requests 10000 --max-requests-jitter 1000 app:app

3. --graceful-timeout for worker shutdown

Gunicorn equivalent: --graceful-timeout

Granian has --workers-kill-timeout but clearer graceful shutdown semantics would help.

Context

• Related to SSE connection leak issue: [BUG]: (sse): Granian CPU spikes to 800% after load stops, recovers when load resumes #2357
• Granian upstream issue: Disconnect event is not propogated to the middleware while a request is processing emmett-framework/granian#286
• These features would help work around granian not properly closing SSE connections after client disconnect

References

• Gunicorn settings docs
• Granian GitHub