On Thu, Nov 24, 2011 at 11:56:01AM +0100, [email protected] wrote:
> Hi Dejan,
> 
> yep ! the modification is efficient now with lrmdadmin, I've tested with 
> following values :
> lrmadmin -p max-children 24
> and the 24 dummy resources are started in quite one shot and also the 24 
> op monitor are executed in one shot.
> lrmadmin -p max-children 2
> and the 24 dummy resources are started by successive pairs, and same thing 
> for op monitor.
> So that is really efficient.
> 
> Is this value taken in account for any action in Pacemaker ? meaning : 
> configure, delete, start, stop, etc. ?

No. Pacemaker has a similar attribute (global cluster property)
batch-limit which limits the number of actions (operations) that
the upper layer (crmd/TE) is going to request in parallel from
lrmd.

Thanks,

Dejan

> Thanks a lot
> Regards
> Alain
> 
> 
> 
> 
> 
> De :    Dejan Muhamedagic <[email protected]>
> A :     General Linux-HA mailing list <[email protected]>
> Date :  24/11/2011 10:53
> Objet : Re: [Linux-HA] Antw: What about "start-delay" attribute status ?
> Envoyé par :    [email protected]
> 
> 
> 
> Hi,
> 
> On Wed, Nov 23, 2011 at 08:52:43AM +1100, Andrew Beekhof wrote:
> > On Wed, Nov 23, 2011 at 4:05 AM, Dejan Muhamedagic <[email protected]> 
> wrote:
> > > On Tue, Nov 22, 2011 at 04:44:50PM +0100, [email protected] wrote:
> > >> Hi again,
> > >>
> > >> that's strange because I did tests around this parameter
> > >> LRMD_MAX_CHILDREN,
> > >> with 24 Dummy resources, therefore resources which do quite nothing 
> and so
> > >> Pacemaker
> > >> should start all resources at quite the same time one after the 
> other.
> > >> Then monitor op
> > >> should also be quite at the same time one after the other.
> > >> First, I test with no  LRMD_MAX_CHILDREN in /etc/sysconfig/pacemaker 
> so
> > >> default value
> > >> which is probably 4 as you told me, then  I set it to 2, restart 
> Pacemaker
> > >> and did same test,
> > >> and finally set it to 24 (just for a school case) and did the same 
> test .
> > >> And the result is the same for the three tests :
> > >> when all the 24 Dummy resources are started , as you can see below,
> > >> the op monitor seems to be gathered by 4, whatever is the
> > >> LRMD_MAX_CHILDREN value,
> > >> whereas my understanding was the monitor operations should have been
> > >> parallelized for
> > >> almost the 24 resources as the monitor takes a very short while to be
> > >> completed ...
> > >
> > >> Where am I wrong ?
> > >
> > > It could be that the init script on your platform doesn't
> > > support this parameter. You should talk to your vendor.
> > 
> > Init script?  Doesn't the lrmd just look for that variable in its 
> environment?
> 
> No, it doesn't. lrmadmin has to be used to set the parameter:
> 
> lrmadmin -p max-children $LRMD_MAX_CHILDREN
> 
> Thanks,
> 
> Dejan
> 
> P.S. Attached suse openais init script.
> 
> > _______________________________________________
> > Linux-HA mailing list
> > [email protected]
> > http://lists.linux-ha.org/mailman/listinfo/linux-ha
> > See also: http://linux-ha.org/ReportingProblems
> [pièce jointe "openais" supprimée par Alain Moulle/FR/BULL] 
> _______________________________________________
> 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