[GE users] Cluster Partitioning via PEs

aeszter Ansgar.Esztermann at mpi-bpc.mpg.de
Mon Mar 15 12:54:34 GMT 2010

    [ The following text is in the "utf-8" character set. ]
    [ Your display is set for the "ISO-8859-10" character set.  ]
    [ Some characters may be displayed incorrectly. ]

Hello everyone,

migrating from a Torque/Maui installation, I am following along the lines of this message to implement somthing akin to "partitions":

The idea is that parallel jobs must be scheduled in a way that conforms to our Infiniband topology -- isolated sets of 24 nodes each, so slot combinations belonging to different switches are not allowed.
I was under the impression that I could use host groups for nodes connected to a common switch, then define somthing like implicit parallel environments in the queue config like this:

pe_list intelmpi,[@ib1=intelmpi_ib1],[@ib2=intelmpi_ib2]

However, SGE (6.2u3, to be exact), does not seem to share that opinion:

# qconf -Aq short 
Parallel environment "intelmpi_ib1" does not exist

Am I overlooking something obvious?



Ansgar Esztermann
Max-Planck-Institut für biophysikalische Chemie, Abteilung 105


To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].

More information about the gridengine-users mailing list