[Pacemaker] [PATCH]Omitted STONITH of useless broadcast.(only 2 nodes configuration)

[renayama19661014 at ybb.ne.jp]

Hi Andrew,

Thank you for comment.

> It should be able to calculate the expected number of votes/replies
> from the ais/heartbeat membership list directly.
> The crmd shouldn't need to pass that info in IMHO.

OK.

I think about a patch of the form to acquire memberlist from stonithd.

Best Regards,
Hideo Yamauchi.

--- Andrew Beekhof <andrew at beekhof.net> wrote:

> On Wed, Jul 7, 2010 at 3:02 AM,  <renayama19661014 at ybb.ne.jp> wrote:
> > Hi Andrew,
> >
> > Thank you for comment.
> >
> >> stonithd should also have access to the memebership list, so i dont
> >> think clients like the crmd need to be involved here.
> >
> > My understanding may be wrong.
> >
> > When stonithd accesses memberlist, stonithd can know node configuration.
> > It is from memberlist and does not understand which node can carry out STONITH.
> > Only stonithd of each node knows an STONITH list of nodes to have.
> >
> > Or would you commented in a meaning to judge the number of the configuration of the cluster
> from
> > memberlist?
> 
> It should be able to calculate the expected number of votes/replies
> from the ais/heartbeat membership list directly.
> The crmd shouldn't need to pass that info in IMHO.
> 
> >
> > Best Regards,
> > Hideo Yamauchi.
> >
> > --- Andrew Beekhof <andrew at beekhof.net> wrote:
> >
> >> 2010/7/6 &#65533;<renayama19661014 at ybb.ne.jp>:
> >> > Hi All,
> >> >
> >> > I wrote a patch.
> >> >
> >> > This patch limited it to two node configuration and wrote it.
> >> > When failed in STONITH by the configuration of two nodes, the request of STONITH to the
> other
> >> node is
> >> > useless.
> >> >
> >> > Because the reason is because it can carry out STONITH only from one node.
> >> > We should omit the request of STONITH of Broadcast.
> >> >
> >> > It is not necessary to wait for time-out of broadcast of STONITH by omitting it.
> >> >
> >> > &#65533;* We use the configuration of two nodes in many cases.
> >> > &#65533;* It is very effective to omit STONITH of useless broadcast.
> >>
> >> stonithd should also have access to the memebership list, so i dont
> >> think clients like the crmd need to be involved here.
> >>
> >> >
> >> > I am confirming the movement of this patch now.
> >> > Please tell me an opinion of all of you.
> >> >
> >> > Best Regards,
> >> > Hideo Yamauchi.
> >> >
> >> >
> >> > _______________________________________________
> >> > Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> >> > http://oss.clusterlabs.org/mailman/listinfo/pacemaker
> >> >
> >> > Project Home: http://www.clusterlabs.org
> >> > Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> >> > Bugs: http://developerbugs.linux-foundation.org/enter_bug.cgi?product=Pacemaker
> >> >
> >> >
> >>
> >> _______________________________________________
> >> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> >> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
> >>
> >> Project Home: http://www.clusterlabs.org
> >> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> >> Bugs: http://developerbugs.linux-foundation.org/enter_bug.cgi?product=Pacemaker
> >>
> >
> >
> > _______________________________________________
> > Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> > http://oss.clusterlabs.org/mailman/listinfo/pacemaker
> >
> > Project Home: http://www.clusterlabs.org
> > Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> > Bugs: http://developerbugs.linux-foundation.org/enter_bug.cgi?product=Pacemaker
> >
> 
> _______________________________________________
> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
> 
> Project Home: http://www.clusterlabs.org
> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> Bugs: http://developerbugs.linux-foundation.org/enter_bug.cgi?product=Pacemaker
>