[ejabberd] Idle C2S processes

Anton Belyaev anton.belyaev at gmail.com
Tue Sep 22 00:03:22 MSD 2009


2009/9/21 Carlos Abalde <carlos.abalde at gmail.com>:
> Hi all,
>
> I'm experiencing a strange behavior with ejabberd. I'm running a
> clustered deployment with a lot of users accessing to the server through
> http-bind connections (same client implementation for everybody).
> Everything works fine, however, I've detected that, once in a while,
> some connections are 'blocked'.
>
> The symptoms are C2S processed with growing message queues, always in
> the same state (waiting) and in the same current function ({prim_inet,
> recv0, 3}). Looking for info of the C2S session in mnesia (IP and port)
> and running a netstat I've discovered that the connection does not
> exist. I don't know how to get associated http-bind pid, but I suspect
> that that process is dead.
>
> Anyone has experienced the same symptoms? Any idea on how to avoid this
> problem?

Do you experience this only on HTTP-BIND connections?


More information about the ejabberd mailing list