> -----Original Message-----
> From: Testbed Support for GridPP member institutes [mailto:TB-
> [log in to unmask]] On Behalf Of Stephen Burke
> Sent: 11 December 2012 19:36
> To: [log in to unmask]
> Subject: Re: Ops meeting at 11am
>
> Testbed Support for GridPP member institutes [mailto:TB-
> > [log in to unmask]] On Behalf Of Ewan MacMahon said:
> > I don't think it's a bug as such - it appears to be designed in. The
> > failure arises from the QUEUE_xxx_CLUSTER_UniqueID YAIM variable[1]
>
> Hmm ... but that isn't a variable for the cluster node configuration, it's
> a variable for the CREAM node.
>
No, it's not. It's parsed by YAIM on the cluster publisher, specifically
by the 'config_cluster' and 'config_cluster_gip_glue2' functions. Plus,
it doesn't specify a CE anywhere, so given that YAIM input is just an
unstructured list of variable definitions, there's no way to take something
like 'QUEUE_LONGFIVE_CLUSTER_UniqueID' and work out which CE it does or
doesn't apply to.
Plus, empirically, I have the errors about these variables in the YAIM
logs on my cluster publisher node. Which I'm not using, so you're welcome
to have a log in and look at it if you'd like.
> Clearly you can't have queues with the same
> name on the same CREAM node feeding different clusters, whether in cluster
> mode or not, but if you have separate CREAM nodes you can define the same
> queue names with different cluster IDs,
>
I'm pretty sure you can't. Clearly, there's no in principle reason that
you shouldn't be able to, but (at least as far as YAIM is concerned) you
don't actually seem to be able to do it.
Ewan
|