Okay, thanks. The server on the other end is a proprietary thing, it's
possible that the developer didn't think about keep-alives at all.
I'm only noticing it as an issue because the session times seem very long (like, much longer than they should be).
On Mon, Sep 29, 2008 at 9:30 PM, Willy Tarreau <w#1wt.eu> wrote:
> Hi Chris,
>
> On Mon, Sep 29, 2008 at 10:04:33AM -0700, Chris Charman wrote:
> > Hey folks, quick question for anyone who might know.
> >
> > I currently use option httpclose in both the defaults and frontend
> > definition on my instance. It seems like backend connections are still
> being
> > kept alive though. Would httpclose be required in the backend definition
> as
> > well?
>
> No, it is not needed. A connection switches to httpclose if at least one of
> the frontend or backend which serves it is configured with httpclose. And
> since
> your defaults also specify the option, both your frontend and backend are
> set
> to httpclose. So you should double-check, normally you are not getting any
> keep-alive.
>
> If it would happen that your server ignores "connection: close", you can
> use
> "option forceclose", which I developped because I once encountered this
> problem (I don't remember on what server though). It enforces connection
> close
> by closing the connection to the server once the server starts to respond.
>
> But generally you don't need this.
>
> Regards,
> Willy
>
>
Received on 2008/09/30 06:34
This archive was generated by hypermail 2.2.0 : 2008/09/30 06:48 CEST