Andrew Beekhof schrieb:
> 
> On May 19, 2008, at 4:39 PM, Michael Schwartzkopff wrote:
> 
>> Hi,
>>
>> we had the possibility to play with a large Linux-HA installation and
>> we tried
>> to set up a 16-node-cluster. We somehow manged it. Please see the
>> result at:
>>
>> http://clusterbau.blogspot.com/2008/05/how-many-nodes-does-cluster-tolerate.html
>>
>>
>> Bascially we found out that such large installations are hard to
>> manage and
>> get really slow.
> 
> I'd be very interesting in finding out (and fixing) why the cluster was
> so slow.
> 
> Some things I can think of that may help...
> * which version were you running?

Latest pacemaker available from SuSE build service.

> * did you set globally_unique=false for the clone(s)?
Default value. We did not change anything

> * how much monitoring were you doing?
once every 60 seconds.

> * what log level did you have in ha.cf?
Default. No changes on ha.cf here.

> 
> Which process(es) did top report to be the biggest CPU hogs?

Can't remember.

It was just playing around. Nothing for serious investigations. Perhaps
next time I have access to that machine(s) we can do a debugging
session. I will ask.

Greetings,

Michael.
_______________________________________________
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