Surviving Digg?

[Rt Ibmer]

>Ok, tried that, no notice printed out.  What is a good value for
>worker_connections?

http://wiki.codemongers.com/NginxMainModule#worker_processes talks about these settings.

>proxy_connect_timeout      90;
>proxy_send_timeout         90;
>proxy_read_timeout         90;

>Are these reasonable?

Depends on what is normal for your app. IMO these are WAY to high for defaults. It means that nginx will wait 1.5 minutes(!) for a downed server to respond before moving to your other box. You should also look at max_fails and fail_timeout.

>Yup, just trying to figure out what exactly it is that we need!  We've
>hit the front page of digg before, though not as big last time.

To me it seems nginx did its job. Glad you solved the FDs with ulimit -n that is the only obvious thing that jumps out at me on the nginx side. I highly suspect an application bottleneck from what you describe, such as a slow db or something.

>Thanks for the help!

My pleasure. I am just a few weeks into nginx and this list so it feels good to help out and be in a position to start giving back.  HTH.


      ____________________________________________________________________________________
Be a better friend, newshound, and 
know-it-all with Yahoo! Mobile.  Try it now.  http://mobile.yahoo.com/;_ylt=Ahu06i62sR8HDtDypao8Wcj9tAcJ