I activated $request_time in nginx logs and indeed I'm reaching the haproxy timeout (30s), whereas harakiri is set to a lower value (29s), but no worker is killed. I allocated two times more workers, and the problem is still present.
It seems like the requests are blocked somewhere in uwsgi, without being killed by harakiri... and it's not because of slow clients, I cannot reproduce with curl --limit-rate 1. I don't believe the workers are overwhelmed either as the haproxy session stats do not show high numbers (only a few connections active at the time of the failures). Please tell me if setting a breakpoint somewhere in the code would be useful. Thanks for your help, -- Jean-Baptiste Quenot _______________________________________________ uWSGI mailing list [email protected] http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
