[Pacemaker] [BUG] Clone + group = orphan(s) ?

[Dejan Muhamedagic]

Hi,

On Wed, Jan 06, 2010 at 03:35:10PM +1000, Thomas Guthmann wrote:
> Hi Andrew
> 
> >Any chance you could download the latest tarball and try that?
> >    http://hg.clusterlabs.org/pacemaker/stable-1.0/archive/1aa7cf1450e4.tar.bz2
> 
> Awesome! No more orphans ! \o/
> Everything is working now, it's clean and tidy.
> 
> *But* ... there is a small regression in the crm tool. When I clone
> a group, crm used to update the location constraints by the name of
> the clone instead of the group name. Does "updated" mean deleted now
> ? :)

"update" doesn't sound like "delete" to me :)

> So for the moment I re-add all constraints manually after the
> clone command.
> 
> crm(live)configure# show loc-tom-lo-vip-ns1
> location loc-tom-lo-vip-ns1 tom-DNS \
>         rule $id="rule-tom-lo-vip-ns1" +inf: anycast-server-location
> eq tom and anycast-server-type eq dns
> crm(live)configure# clone tom-DNS-clone tom-DNS meta clone-max=2
> INFO: resource references in location:loc-tom-lo-vip-ns1 updated
> INFO: resource references in location:loc-tom-lo-vip-dns1 updated
> INFO: resource references in location:loc-tom-named updated
> INFO: resource references in location:loc-tom-lo-vip-dns2 updated
> INFO: resource references in location:loc-tom-lo-vip-ns2 updated
> crm(live)configure# show loc-tom-lo-vip-ns1
> crm(live)configure#

That's really strange since crm doesn't depend on any internal
business of CRM.

Can you please file a bugzilla for Shell and attach your CIB
before creating the clone.

Thanks,

Dejan

> Many thanks for your great support Andrew.
> 
> Last request, could you add a vocabulary page in the wiki (or can
> you create me an account) and add the orphan description you gave in
> a email 1 months ago because otherwise it's written no where.
> 
> "Orphan resources are those for which there are entry in the status
> section, but no definition in the resources section.
> The cluster basically thinks you haven't defined the vip resource for
> some reason."
> 
> Cheers,
> -Thomas
> 
> 
> _______________________________________________
> Pacemaker mailing list
> Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker