[slurm-users] Requirement to use QOS?

Loris Bennett loris.bennett at fu-berlin.de
Mon Jan 22 02:02:45 MST 2018


Loris Bennett <loris.bennett at fu-berlin.de> writes:

> Hi,
>
> Some while ago I defined several QOS thus:
>
>      Name   Priority     MaxWall MaxJobs MaxSubmit 
> ---------- ---------- ----------- ------- --------- 
>     normal          0                               
>      short     100000    03:00:00      10        20 
>     medium      10000  3-00:00:00     100       200 
>       long       1000 14-00:00:00                                 
>
> This worked OK with Slurm 17.02.7 and previous versions.  The default
> QOS for all users was 'long' and all users were able to use both 'short'
> and 'medium'.
>
> Recently, still using version 17.02.7, I added a restriction on the
> number of cores per user thus:
>
>      Name   Priority     MaxWall MaxJobs MaxSubmit     MaxTRESPU 
> ---------- ---------- ----------- ------- --------- ------------- 
>     normal          0                                             
>      short     100000    03:00:00      10        20       cpu=20 
>     medium      10000  3-00:00:00     100       200       cpu=200 
>       long       1000 14-00:00:00                                 
>
> However, now users are unable to use 'short' and 'medium' as the
> following error is produces by 'sbatch':
>
>   sbatch: error: Batch job submission failed: Invalid qos specification
>
> Deleting the MaxTRESPerUser settings does not fix the problem.
>
> My assumption is that that some change in this or a previous version was
> made in the way QOS can be accessed by associations, but this change has
> only now manifested itself after the QOS have been updated.
>
> Can anyone confirm this?

In a similar vein, the documentation for 'sacctmgr' says

  Note: the QOS that can be used at a given account in the hierarchy are
  inherited by the children of that account.

However, if I do the following:

  $ sacctmgr modify account name=root set qos+=medium,short

the result is

  Modified account associations...
    C = soroban    A = root
    C = soroban    A = anemometry         U = alice 
    C = soroban    A = anemometry         U = bob 
    C = soroban    A = barometry          U = carol   
    C = soroban    A = barometry          U = dave    
    C = soroban    A = calorimetry        U = ethel 
    ...

To me this looks as if the QOS are in fact being explicitly added to
each association.  In this case, will a new association added within
this hierarchy automatically be associated with the QOS available to the
other associations?

Cheers,

Loris

-- 
Dr. Loris Bennett (Mr.)
ZEDAT, Freie Universität Berlin         Email loris.bennett at fu-berlin.de



More information about the slurm-users mailing list