Hello!

On Sat, Oct 03, 2009 at 05:19:04PM -0400, Davy Campano wrote:

> Does anyone what the defaults are for fastcgi_connect_timeout and how it
> works? It seems like at first when one of my upstream backends was down,
> that it waited 10-11 seconds before trying an alternate, but then after a
> few failures it started only waiting 3 seconds to try the alternate.

fastcgi_connect_timeout defaults to 60 seconds

It works exactly as other timeouts do - it's armed on connect
start, and fires after specified time.

Change in timings "after a few failures" likely to be caused by
max_fails / fail_timeout, see

http://wiki.nginx.org/NginxHttpUpstreamModule#server

for details.

Exact numbers you specify seems to be unrelated to timeouts
though. It's likely to be caused by overloaded backends instead.
For pure timeouts you should see something like 60 seconds before
max_fails reached, and 0 seconds after.

Maxim Dounin