Strange repeated errors

Maxim Dounin mdounin at
Wed May 23 08:41:04 UTC 2012


On Wed, May 23, 2012 at 09:17:11AM +0200, Greg wrote:

> Hi,
> Le 22/05/2012 18:21, Maxim Dounin a écrit :
> >Looks like your upstream servers fail to respond in time for some
> >reason and nginx tries upstream servers configured as per
> >proxy_next_upstream (see
> >
> >The reason why your backends time out is probably related to a
> >particular request properties, likely something user/cookie
> >related as you can't reproduce it with just url.  Try looking into
> >backend logs.
> You're right !
> But is that normal that next upstreams are tried infinitely ? How
> many tries did NginX before considere that the request is really
> failed ?

Normally nginx will try all servers specified in an upstream{} 
block before giving up.

Before 1.3.0 there were infinite loop over upstream servers 
possible if there were "backup" servers specified in an upstream 
block.  It normally affected only setups with "proxy_next_upstream 
http_404" used as well, but might also affect normal setups if 
only specific requests fail and iteration of the loop over 
backends takes more than fail_timeout.

Maxim Dounin

More information about the nginx mailing list