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

Re: [HTCondor-users] prioritizing one group of machines over another



hi,
thanks for the replies! michael, you are right that the rank _expression_ is quite useless in my scenario (using only partitionable slots) if consumption policy is not switched on. however, the system starts to behave weird if i do. for instance, jobs seem to oscillate between the running and idle state. any idea, why that may be the case?

best,
thomas

Am 2016-01-20 um 17:50 schrieb Michael V Pelletier:
From: Thomas Hartmann <thomas.hartmann@xxxxxxxx>
Date: 01/20/2016 05:35 AM
 
> hi,
> i have two groups of machines in my setup. the machines of one group are
> faster, more reliable. is it possible to tell the scheduler to only
> submit jobs to the other group if the first group of machines cannot
> take any more jobs?

What you're looking for is the RANK _expression_:

http://research.cs.wisc.edu/htcondor/manual/latest/3_5Policy_Configuration.html#SECTION00451400000000000000

The manual gives the example of memory size of the jobs as a determining
factor, but in your case you'd want to use, for example, the machine's
KFlops attribute, so that the faster machines get top priority for jobs.

Similarly, I have a startd_cron which probes the network interface speed
and incorporates that into the rank _expression_, so that machines with
bonded or 10G interfaces get preference over single-interface gigabit.

If you're using partitionable slots this can go a bit sideways depending
on how you're configured - by default the negotiator will make a trip
through all of the partitionable slots, which means that it will include
the low-ranked machines' partitionable slots in its matching and you'll
wind up with one job on each machine in both groups. If I'm remembering
correctly, using consumption policies instead of claim partitionable
leftovers addresses this.

        -Michael Pelletier.

_______________________________________________
HTCondor-users mailing list
To unsubscribe, send a message to htcondor-users-request@xxxxxxxxxxx with a
subject: Unsubscribe
You can also unsubscribe by visiting
https://lists.cs.wisc.edu/mailman/listinfo/htcondor-users

The archives can be found at:
https://lists.cs.wisc.edu/archive/htcondor-users/

-- 
Dr. Thomas Hartmann

Centre for Cognitive Neuroscience
FB Psychologie
Universität Salzburg
Hellbrunnerstraße 34/II
5020 Salzburg

Tel: +43 662 8044 5109
Email: thomas.hartmann@xxxxxxxx

"I am a brain, Watson. The rest of me is a mere appendix. " (Arthur Conan Doyle)