Gunicorn workers restart by connection reset

Question

It seems to me that my gunicorn workers are restarted each time there is connection reset by a browser (eg by reloading a page while a request is still progress or as a result of connectivity problems).

This doesn't seem to be a sensible behaviour. Effectively I can bring down all the workers just by refreshing a page in a browser a few times.

Questions:

• What are the possible causes for a gunicorn worker restart?
• What would be the right way to debug this behaviour?

I'm using Pyramid 1.4, Gunicorn (tried eventlet, gevent and sync workers - all demonstrate the same behaviour). The server runs behind nginx.

Answer 1

I mis-diagnosed the problem. It seems that both Firefox and Chrome perform certain optimizations when loading the same page address multiple times. I thought the server was becoming irresponsive, but in fact there were no requests generated to serve.