there's nothing wrong. It's just the program doesn't need that much of memory.

#slurm #sbatch

thank you Brian Andrus when i saw your reply had already done that i have made those changes: taskPlugin = task/Affinity to taskPlugin = task/cgroup enabled SelectTypeParameters=CR_CPU_Memory SelectTy

#slurm #sbatch

1M is the default value if you don't specify anything for the node in slurm.conf As Brian mentioned, you can take that 'slurmd -C' output (NodeName= line) and put it into the slurm.conf to flesh out t

#slurm #sbatch

Vitalina, You should put the output from 'slurmd -C' (on the node) as the information for the node configuration in slurm.conf I usually take the RealMemory and round it down a little to ensure there

#slurm #sbatch

when i run scontrol show nodes, it shows real memory on nodes iqual to 1M. In compute node when i run slurmd -C i got Real memory iqual to 36194 But Slurm is assuming the available real memory =1M . i

#slurm #sbatch

you can check slurmctld.log and slurmd.log for more information (in slurm.conf set SlurmctldDebug and SlurmdDebug to debug or even debug2 if necessary).

#slurm #sbatch

Hello i did it bu the problem persist jobs a running using only 1MB Mem or less. It is still assumig real memory = 1 but running Slurmd -C on compute nodes we have 36000 Real Memory available

#slurm #sbatch

“I don’t know” is a typo, just ignore it. not sure why it is there

#slurm #sbatch

``` #RealMemory=4096 DefMemPerNode=4096 ``` "RealMemory" is set for a node, so you should append it after "NodeName". I don't know "DefMemPerNode" can be set for the entire cluster or for a partition.

#slurm #sbatch

Hi wang, please find bellow all the slurm.conf file ClusterName=linux ControlMachine=strge #ControlAddr= #BackupController= #BackupAddr= # SlurmUser=slurm #SlurmdUser=root SlurmctldPort=6817 SlurmdPor

#slurm #sbatch

Please post the relevant lines in slurm.conf (especially , the node and partition definitions)

#slurm #sbatch

We are running openhpc 1.3.9 on centOS 7.9, so we noticed that the process/jobs launched spend the same time, even if we add more compute nodes. We noticed throught the scontrol show nodes that the re

#slurm #sbatch

Oh, that was it! I used "$wwsh file sync" to pass the /etc/passwd, /etc/group and /etc/group to the compute nodes and then sbatch worked :) Thank you very much!

#sbatch #slurm

That is because your compute node does not have the same user IDs as the node that submitted it. Brian Andrus

#sbatch #slurm

Hello, I am having problems with $ sbatch on the cluster I am working; As shown in the image below, $srun runs just fine (above the yellow line), while the script jobtest.sh will not run with $sbatch;

#sbatch #slurm

Hi Chris, I logged directly on to the node before I ran it. The performance drop is consistent in all the ways I have run it Best regards, Per

#sbatch #slurm

Hi Per, Have you tried running on the provisioned node, but outside Slurm to rule out anything there (cgroup / affinity or other Slurm config - been there, seen that, see my post here a few weeks back

#sbatch #slurm

Hi, Now I tried to run HPL and it looks worse on the master node I get 1741 GFlops on the provisioned node I get 145 GFlops if I make a normal stand alone installation on a node I get 1752 GFlops Best

#sbatch #slurm

Hi Chris, Thanks for the answer - good point I will try HPL and see if it behaves the same way. The program is compiled for Xeon Phi, so it can't run on i7-8700K in the cluster (but it runs around 250

#sbatch #slurm

Hi, Can you confirm the numbers are as follows? - Xeon Phi 7210 = 250s - Xeon Phi 7250 = 88s - i7-8700K = 220s I would naively assume the i7 and 7250 results to be the other way around (ie 7250 should

#sbatch #slurm