Testbed Support for GridPP member institutes
> [mailto:[log in to unmask]] On Behalf Of Matt Doidge said:
> So I was wondering if some kind individual could please make
> available
> their working subclustered CE ldifs so I can blatently copy
> off of them?
You can always get some LDIF with an ldapsearch to e.g. the tier-1.
> For reference he's my latest fumbly attempts at customising
> our ldifs if
> anyone wants to see how *not* to do it:
> www.hep.lancs.ac.uk/~msd/static-file-CE.ldif
> www.hep.lancs.ac.uk/~msd/static-file-Cluster.ldif
I suspect that the thing you've missed is that you have to split the
clusters as well as the subclusters, i.e. with the current WMS matching
you can only have one subcluster per cluster. The linkage goes CE <->
cluster <-> subcluster, so you have to make sure that all the uniqueid
references for the cluster and subcluster change to something which is
not the CE headnode hostname (and which is unique!).
Also although I haven't checked the ldif in detail I happen to notice
this:
dn: GlueSubClusterUniqueID=fal-pygrid-44-twins.lancs.ac.uk,
GlueClusterUniqueID=fal-pygrid-44.lancs.ac.uk,mds-vo-name=resource,o=gri
d
[...]
GlueChunkKey: GlueSubClusterUniqueID=fal-pygrid-44-twins.lancs.ac.uk
That ChunkKey should contain the Cluster ID, not the SubCluster.
Stephen
--
Scanned by iCritical.
|