Re: [SOLVED] Rails app with maxconn 1 receiving more than 1 request at a time

On Sun, Sep 14, 2008 at 6:11 PM, Willy Tarreau <w#1wt.eu> wrote:
> The first attached patch fixes the problem. Not only I could not
> reproduce the symptom anymore, but it also makes sense.

Thanks, Willy, for coming up with a fix so quickly. We will be updating our production servers with the patches tomorrow.

> It still indicates something strange in your setups, because you
> all have been facing retries to trigger this bug. That means that
> your servers are sometimes rejecting connections or not accepting
> them. One possible cause might be because you would have some TCP
> connection tracking enabled on either the proxy or the server (or
> both), such as netfilter with or without NAT.

Would this not occur whenever the Mongrel processes are being stopped/restarted? In our setup, we restart those processes many times a day using Monit, in order to neutralize some pesky memory leaks. And obviously they are killed and restarted whenever we deploy a new code release.

Alexander. Received on 2008/09/14 18:51