Hi Cyril,

On Mon, Jun 20, 2011 at 07:42:11PM +0200, Cyril Bonté wrote:
> Hi again, 
> 
> I answer to myself ;-)
> 
> Le lundi 20 juin 2011 19:12:25, Cyril Bonté a écrit :
> > While auditing a server today, I encountered an issue with the HTTP
> > keep-alive timeout. It just didn't work and connections stayed alive for 5
> > minutes (their client/server timeout).
> > The fact is that they decided to split their proxy in frontend and backends
> > and moved some options in the backends parts. While this is a valid
> > configuration if we refer to the documentation, the http-keep-alive timeout
> > is ignored because at the time the value is checked in the code (in the
> > early request process), the backend is not known yet.
> > (...)
> > I've started to work on a fix and can try to send a patch soon.
> > 
> > In the same idea, they don't use "timeout http-request" but it looks like
> > the same issue exists.
> 
> After reading more carefully the documentation, maybe another solution is in 
> a 
> short update in it, by replacing *should* by *must*.
> 
> "
> If this parameter is not set, the client timeout still applies between each
> chunk of the incoming request. It *must* be set in the frontend to take
> effect, unless the frontend is in TCP mode, in which case the HTTP backend's
> timeout will be used.
> "

That's what I wanted to suggest, in fact it "works as designed" as some like
to call it. Yes we may improve the doc. I checked why we supported it at all
in the backend, and indeed it's for cases where we switch from a TCP frontend
to an HTTP backend.

Timeouts are clearly the type of setting that should lie in a defaults section,
that's the best way not to forget any of them and to have them coherent.

Regards,
Willy


Reply via email to