[rt-users] SOS:: many mysql processes are killing RAM, why?

[ktm at rice.edu]

On Tue, Jul 23, 2013 at 10:14:39AM +0800, Craig Ringer wrote:
> On 07/19/2013 11:42 PM, Alex Decalli wrote:
> > 
> > I have RT 3.8 on mysql 5.1 on ubuntu 11.04. Why do I have many mysql
> > processes they are bringing down RAM.
> 
> Quite likely you have too many Apache workers with mod_perl, or too many
> fcgi workers if you're using fcgi.
> 
> You will need to limit the number of concurrent RT workers - assuming
> that the MySQL workers are even being used by RT, which we don't know
> based on the information you have provided.
> 
> Alternately, you can put a *connection pool* in place between RT and the
> database. If you were using PostgreSQL I would advise you to use
> PgBouncer in transaction pooling mode, but I cannot offer much advice
> for MySQL. I'm sure there's a similar pooling proxy.
> 
> This sort of question is probably better suited to Serverfault.com than
> the RT list. BTW, text output is usually way more useful than
> screenshots; eg "top -b -n 1".
> 
Hi Alex,

I will second Craig's suggestion about a connection pooler. We use fcgi/pgbouncer
with PostgreSQL here and it works well. The only difference is that we use "session"
mode and not "transaction" mode with pgbouncer. We had cases of ticket updates being
interspersed. Maybe it was a local configuration problem but changing to
"session" mode fixed it.

Regards,
Ken