Dave,
Our clusters are homogeneous, so we are trying to do a first-level sort
on "load" where the load formula is simply "slots". If it did this much
correctly I could happily ignore it's second-level behavior. Have you
ever made this work as a first-level sort?
We have two clusters, one running 6.1u5 and the other running 6.2u2-1.
The 6.2u2-1 system appears to sort by sequence number in defiance of my
instructions to use "load" with a load formula of "slots". I have not
actually been able to deduce a coherent pattern of behavior for the
6.1u5 system, but it is clearly not packing the jobs.
What is the "Univa repo" you refer to?
Jim
On 6/2/2011 10:01 AM, Dave Love wrote:
Reuti<[email protected]> writes:
There is none. It's like outlined here:
http://blogs.oracle.com/sgrell/entry/grid_engine_scheduler_hacks_least
Apologies for the confusion. However, that doesn't deaql with
multi-level sorting on seqno and load. In an inhomogeneous cluster,
you typically want to sort by seqno (node type) first. At
least in that case, the second-level packing using slots for the load
doesn't work for us in the serial case either. There's a
one-liner fix for the parallel case in the Univa repo, but I don't know
what's the problem in the serial case.
_______________________________________________
users mailing list
[email protected]
https://gridengine.org/mailman/listinfo/users
