Trouble understanding why Varnish retries on backend timeouts

. gmoniey at
Mon Nov 12 20:30:11 CET 2012

Thanks for the explanation.

In regards to the race condition, would it make sense to be able to turn
off this functionality using some config parameter? I guess with the patch
applied, it shouldn't really be much of a problem as it will only occur for
GETs, but ideally, it would be nice to control this type of behavior.

On Mon, Nov 12, 2012 at 2:01 AM, Tollef Fog Heen <
tfheen at> wrote:

> ]] .
> > I recently got bit by this functionality, and I was hoping that someone
> > could shed some light on why this is the default behavior for varnish.
> Because there's a race between when we submit a GET on a connection and
> when the backend closes it on us because of a timeout.
> > From everything I have read online,
> > the retry should only occur once, but I could have sworn I saw it
> > happen multiple times.
> If you restart in vcl_error, it can happen multiple times.
> > and even a patch (which doesn't seem to have been applied):
> >
> We should probably apply this.  Martin, do you want to resubmit it
> (possibly after cleaning it up)?
> --
> Tollef Fog Heen
> Technical lead | Varnish Software AS
> t: +47 21 98 92 64
> We Make Websites Fly!
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <>

More information about the varnish-dev mailing list