[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Condor-users] Looking for suggestions for inhomogeneous pool

Good morning,

I was lucky to get some abandoned hardware to integrate into my (up to now,
homogeneous) pool, and am looking for a clever way to setup the new nodes.

Up to now, I have been using a PRE_JOB_RANK to prefer smaller slots over
bigger ones:

NEGOTIATOR_PRE_JOB_RANK = 1000000 - 100 * Memory - $(SWEEP_DIFF)

(SWEEP_DIFF is a small fudge factor to sweep through all nodes, using
Julian date and node number, and doesn't matter here)

Since the new nodes are multi-core, I'd like to use them for MPI jobs, but
not to limit them to parallel universe.
Is there a means to overrule the current NEGOTIATOR_PRE_JOB_RANK setting
for parallel universe jobs only? Another term in the expression, like
 1000000000*(Target.JobUniverse == 11)
? (BTW, can I somehow avoid those "magic numbers"?)
Anything I would have to add to *_ATTRS / *_EXPRS?

Suggestions (and pointers to recipes) are greatly appreciated!


Steffen Grunewald * MPI Grav.Phys.(AEI) * Am Mühlenberg 1, D-14476 Potsdam
Cluster Admin * --------------------------------- * http://www.aei.mpg.de/
* e-mail: steffen.grunewald(*)aei.mpg.de * +49-331-567-{fon:7274,fax:7298}