> But don't forget that the 32 and 64 bit nodes will have to look at
> separate experiment software spaces, and thus you will have to set
> GlueHostApplicationSoftwareRunTimeEnvironment appropriately (i.e.
> not at all for queues serving 32 bit machines, if only the 64-bit
> version's been installed, etc.) - do the present SGM tools cover that?
AFAIK, if you have separate CE for each SubCluster, than no problem even with
currently available sgm tools. (If you have more that one CE per SubCluster,
they should share this info). From site administration point of view, this is
probably the preferred hardware configuration.
If you want one CE to serve more SubClusters, probably sgm tools need to be
improved to allow for a specification of a SubCluster when new
GlueHostApplicationSoftwareRunTimeEnvironment is installed on such a CE, but
this is just a guess and I don't know if this is doable and/or useful at all...
Best regards, Antun
>
> I have this sneaky feeling that this is heading down the route of
> both lots of hacking of the info system by sites and micromanagement
> of software installs by VOs, and will turn out to be a false economy
> overall.
>
> Thanks
>
> Henry
>
> --
> Dr. Henry Nebrensky [log in to unmask]
> http://people.brunel.ac.uk/~eesrjjn
> "The opossum is a very sophisticated animal.
> It doesn't even get up until 5 or 6 p.m."
------- End of Original Message -------
|