[slurm-users] can't get fairshare to be calculated per partition

[Igor Feghali]

hi there

i'm pretty new to slurm and trying to learn my way through it's many
configurations. my config looks like:

SchedulerType=sched/backfill
PriorityType=priority/multifactor
PriorityWeightAge=10000
PriorityWeightFairshare=10000
PriorityWeightJobSize=5000
PriorityWeightPartition=10000
PriorityWeightQOS=10000

and partitions:

# Node Configurations
NodeName=nodeb[01-10]n[01-05] NodeAddr=nodeb[01-10]n[01-05] CPUs=8
RealMemory=95000 TmpDisk=1700000
PartitionName=part1 Nodes=nodeb[01-10]n[01-05] LLN=YES Default=YES
MaxTime=INFINITE State=Up

NodeName=nodeb[01-10]n[06-46] NodeAddr=nodeb[01-10]n[06-46] CPUs=8
RealMemory=95000 TmpDisk=1700000
PartitionName=part1b Nodes=nodeb[01-10]n[06-46] MaxTime=INFINITE State=Up

NodeName=nodeb[01-10]n[47-48] NodeAddr=nodeb[01-10]n[47-48] CPUs=36
RealMemory=190000 TmpDisk=3500000
PartitionName=part2 Nodes=nodeb[01-10]n[47-48] MaxTime=INFINITE State=Up

NodeName=nodeb12n[01-04] NodeAddr=nodeb12n[01-04] CPUs=8 RealMemory=380000
TmpDisk=5600000
PartitionName=part3 Nodes=nodeb12n[01-04] LLN=YES MaxTime=INFINITE State=Up
PartitionName=part3b Nodes=nodeb12n[01-04] LLN=YES MaxTime=INFINITE State=Up


i'm trying to distribute jobs in partitions evenly per account. that is,
for a given partition if I have jobs queued for accounts A and B I expect
the partition to be split approximately 50/50 between those two accounts
(no matter how the other partitions look like).
but things are not working as I expect and as I can tell fairshare is been
calculated for the entire cluster and not per partition.

sample output of `sshare -lm":

             Account       User    Partition  RawShares  NormShares
 RawUsage   NormUsage  EffectvUsage  FairShare
 GrpTRESMins                    TRESRunMin
-------------------- ---------- ------------ ---------- -----------
----------- ----------- ------------- ----------
------------------------------ -----------------------------
root                                                       1.000000
 1059109661                  1.000000   0.500000
     cpu=0,mem=0,energy=0,node=0,b
 root                      root                       1    0.043478
  0    0.000000      0.000000   1.000000
 cpu=0,mem=0,energy=0,node=0,b
 2dpama38b                                            1    0.043478
  0    0.000000      0.000000   1.000000
 cpu=0,mem=0,energy=0,node=0,b
 2pama38                                              1    0.043478
31906    0.000030      0.000030   0.999520
 cpu=0,mem=0,energy=0,node=0,b
etc.

as you can see the partition column is not been populated so I guess that's
my problem. i am having a hard time to fix it so any tip would be
appreciated.

thank you.

side note: part3 and part3b are identical because I need jobs to be split
evenly between those two partitions. if one of them has no jobs than the
other partition gets all the machines. I have no clue how to do that but
that's a subject for a future email.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.schedmd.com/pipermail/slurm-users/attachments/20191029/e41a77c8/attachment.htm>