Hola Pablo,
> We have here at CSCS two CEs of every kind, including Cream. (I don't know
> anything about BES, though).
>
> The setup was not special, the only thing you have to care about is how to
> publish information. Just one of your CEs can publish the information on the
> GlueSubCluster, the other ones should say they have 0 CPUs behind.
Well, there is another thing you should care about. CEs sharing the same
batch system should implement 1 of these options:
1. share the gridmapdir (via NFS, GPFS, Lustre, ...);
2. have disjoint sets of pool accounts.
Otherwise they will map users inconsistently, thereby allowing users to
steal each other's proxies on the WN.
|