Hi,

On Thu, Feb 07, 2008 at 05:50:20PM +0100, Andreas Kurz wrote:
> On Feb 7, 2008 10:08 AM, Lino Moragon <[EMAIL PROTECTED]> wrote:
> > Hi list,
> > I'm running a HA 2 node active / passive cluster with drbd and mysql.
> > I was told about the ocf monitoring function of the yet available mysql
> > ocf script.
> 
> The base monitor function in the mysql RA does only a check if there
> is a process with the PID of your mysql instance running.
> 
> > Could you tell me the advantages or disadvantages of implementing this
> > monitoring function?
> 
> We have a quite similar cluster in our production environment and I
> encountered no problems with a "deeper" mysql check. I installed a
> small test db (the 'City' db from the mysql site) that's exclusive for
> heartbeat's monitor selects so no table locking problems should occur.
> Of course you should not run out of available mysql connections and
> choose timeout values appropriate to your system load.
> 
> A monitor example:
> <op id="op_DB_community-monitor" interval="10s" name="monitor"
> timeout="10s" start_delay="30s">
>                <instance_attributes id="ia_op_DB_community-monitor">
>                  <attributes>
>                    <nvpair id="op_DB_community-monitor-check_level"
> name="OCF_CHECK_LEVEL" value="10"/>
>                  </attributes>
>                </instance_attributes>
> </op>
> 
> > I mean surely it does have benefits, but what if I my php
> > webapplications e.g. generate too much load, too large and complex
> > queries and the timeout value exceeds?
> > Then the server would restart or failover and then that would repeat
> > itself on and on and my ha-cluster wouldn't be a hacluster anymore.
> > Have you any experience what possible things could happen to a mysql
> > server and what would be the reasons for hang-up, crash, etc...?
> > Problem is, I can't estimate the load or amount of queries the
> > application will generate, as I don't have much experience in this field
> > and the DBs will grow over time.
> > Should I implement ocf monitoring or rather not?
> 
> I'm convinced you should ;-)

I'd agree. The oracle RA includes a similar check within the
default, 0-level, monitor operation.

Thanks,

Dejan

> Regards,
> Andreas
> 
> >
> > I'm very thankful for any advice or hints about this matter, where to
> > get more information, or personal experiences with the stability of
> > mysql servers.
> > Thanks
> >  Lino
> > _______________________________________________
> > Linux-HA mailing list
> > [email protected]
> > http://lists.linux-ha.org/mailman/listinfo/linux-ha
> > See also: http://linux-ha.org/ReportingProblems
> >
> _______________________________________________
> Linux-HA mailing list
> [email protected]
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to