Testbed Support for GridPP member institutes [mailto:TB-
> [log in to unmask]] On Behalf Of Andrew Lahiff said:
> That's what I've never understood about glue - it seems to be designed to
> NOT be able to describe correctly 99% of grid sites, where there is a single
> cluster made up of distinct groups of worker nodes with similar properties.
The limit doesn't come from glue, it's imposed by the WMS - it's related to the endless question about passing parameters to the batch system. Basically if you have multiple subclusters behind a single CE you would need to be able to tell the CE which subcluster(s) you matched, and the WMS has never been able to do that. I'm not sure why that would affect Rebus, but it may have embedded the assumption in some way whereas gstat just adds everything up.
Stephen
|