[Ekhi-users] Ekhi SLURM configuration

[Ion Errea]

Thanks Iñigo!

Does of you using Ekhi, do you feel that the queue system has improved?

Bests,

Ion Errea

Fisika Aplikatua 1 saila, Gipuzkoako Ingeniaritza Eskola, and
Centro de Física de Materiales (CSIC-UPV/EHU),
University of the Basque Country (UPV/EHU)
         
Manuel de Lardizabal 5, 20018 Donostia, 
Basque Country, Spain

Tel:    +34 943 01 8417
Mail:  ion.errea en ehu.eus
Web: http://ionerrea.wordpress.com/








> On 12 Nov 2020, at 13:20, Inigo Aldazabal Mensa <inigo.aldazabalm en ehu.eus> wrote:
> 
> Hi all,
> 
> I just set up slurm accounting database, and as I expected (ahem) all
> running and queued jobs were unaffected.
> 
> With this, Slurm is now taking into account all different factors
> for the jobs' priorities, including the "fair-share" factor, i.e. your
> job submission history with a decay half life of 7 days.
> 
> This will also allow us to setup different what is called Quality of
> Services (Qos)  with different limits, weights, etc. (think of QoS as a
> kind of partitions/queues). See an example at DIPC documentation;
> 
> http://dipc.ehu.es/cc/computing_resources/systems/atlas-edr/#qos-and-partitions
> 
> Now you can think on more elaborated schemes for the priorities, queues
> etc. as this all has to be tuned, in case you find it necessary, of
> course.
> 
> If interested, you can find more details about the Slurm
> Multifactor Priority Plugin we are using at:
> 
> https://slurm.schedmd.com/priority_multifactor.html
> 
> Bests,
> 
> Iñgio
> 
> 
> On Mon, 9 Nov 2020 19:38:16 +0100
> Inigo Aldazabal Mensa <inigo.aldazabalm en ehu.eus> wrote:
> 
>> Hi all,
>> 
>> Finally other tasks preempted and made impossible for me to do the
>> slurm changes in ekhi. I'd like to do it tomorrow or on Wednesday,
>> I'll confirm you as it's done. As by my tests you should not notice
>> anything regarding your running jobs, and your scripts will still work
>> unchanged, at least by the moment. We'll later adjust priorities etc.
>> 
>> Bests,
>> 
>> Iñigo
>> 
>> On Wed, 4 Nov 2020 21:22:38 +0100 Inigo Aldazabal Mensa
>> <inigo.aldazabalm en ehu.eus> wrote:
>> 
>>> Hi all,
>>> 
>>> I finally got to try Slurm job accounting changes in a test
>>> environment and all jobs and user allocations seemed unaffected.
>>> 
>>> That being said, I'll like to be sure that you are not running any
>>> important, deadline kind of job, just in case jobs get canceled,
>>> even if they shouldn't.
>>> 
>>> So, if no one objects, my plan is to implement the changes in Slurm
>>> next Monday the 9th. Speak now or forever hold your peace :-)
>>> 
>>> Bests,
>>> 
>>> Iñigo
>>> 
>> _______________________________________________
>> Ekhi-users mailing list
>> Ekhi-users en list.ehu.eus
>> http://list.ehu.eus/mailman/listinfo/ekhi-users
> _______________________________________________
> Ekhi-users mailing list
> Ekhi-users en list.ehu.eus
> http://list.ehu.eus/mailman/listinfo/ekhi-users