[torqueusers] timeshared nodes

[David Johnson (MetOcean Solutions)]

Can someone explain to me how timeshared nodes should be specified in Torque???

I have been trying ':ts' in the nodes file (e.g. 'node1:ts'  - previous message below) without success.
Should I instead be specifying multiple "virtual" processors with e.g. 'node1 np=2'  ??
I have tried this, it kinda works, but then can't seem to be able to get 1 process per node (the scheduler always puts 2 processes on each node).

There is no mention of ':ts' in the online documentation for Torque node configuration:
http://www.clusterresources.com/wiki/doku.php?id=torque:3.2_configuring_node_properties

Thoroughly confused

Thanks in advance any enlightenment


------------------------------------------------------------------
Previous message:

I am trying to set up a small cluster of single processor machines. I want to have all configures as timeshared, so as to run long jobs, but also to occasionally push through jobs fast across the whole cluster.

I have Torque/Maui set up. Everything works fine if nodes are configured as 'cluster'
However when I add the :ts to nodes in nodes config file it goes bad.

pbsnodes -a reports everything correctly, all nodes as 'time-shared' and in 'free' state.

However jobs in the queue do not start with error:

job is deferred.  Reason:  RMFailure  (cannot start job - RM failure, rc: 15062, msg: 'Unknown node ')
Holds:    Defer  (hold reason:  RMFailure)

The only difference is the addition of the ':ts' subscript.

Any ideas?




--------------------------------------------------------------------------------


_______________________________________________
torqueusers mailing list
torqueusers at supercluster.org
http://www.supercluster.org/mailman/listinfo/torqueusers
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.supercluster.org/pipermail/torqueusers/attachments/20060422/1bf20d15/attachment.html