[Pacemaker] Shouldn't colocation -inf: be mandatory?

Dejan Muhamedagic dejanmm at fastmail.fm
Thu Jun 17 11:18:42 UTC 2010


Hi,

On Wed, Jun 16, 2010 at 09:07:10AM -0400, Vadym Chepkov wrote:
> 
> On Jun 16, 2010, at 2:55 AM, Andrew Beekhof wrote:
> 
> > On Tue, Jun 15, 2010 at 9:41 PM, Dejan Muhamedagic <dejanmm at fastmail.fm> wrote:
> > 
> >> colocation not-together -inf: d1 d2 d3
> > 
> > I think there is a problem with this syntax, particularly for +inf.
> > 
> > Consider:
> >  colocation together1 inf: d1 d2
> > 
> > This means d1 must run where d2 is.
> > 
> > But if I add d3:
> >  colocation together1 inf: d1 d2 d3
> > 
> > Now the original constraint is reversed and d2 must run where d1 is
> > (think of how groups work).
> > (Unless you're modifying the order).
> > 
> > I think we need:
> >   no brackets: exactly 2 resources must be specified
> >   () brackets: a non-sequential set
> >   [] brackets: a sequential set
> > 
> > 
> 
> Would something like this be a legitimate syntax then?
> 
> colocation together-but-do-not-die 500: [ d1 d2 d3 ] anchor
> 
> To combine different types in on constraint, basically?

You'd need either of the two:

[ d1 d2 d3 ] ( anchor )
[ d1 d2 d3 ] [ anchor ]

> The reason I am asking, I don't think it's possible to do
> something like this now, isn't it?
> 
> colocation together1 500: d1 d2 d3
> colocation together2 500: d4 d5 d6
> colocation together0 inf: together2 together1

That wouldn't be possible with the new notation either.

Thanks,

Dejan

> Thanks,
> Vadym
> 
> 
> _______________________________________________
> 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




More information about the Pacemaker mailing list