I'm trying to build a cluster but I'm stuck in the slurm partition part. I did create an account and a user, but I don't know how to make a partition to assign it to an account. I'm using Bright Computing, so it automatically installs SLURM and creates a partition by the name of defq. If someone has an example of a command that creates a partition and assigns it to an account, that would be so helpful. If I have any kind of misundertsanding, please correct me.

Thanks

CodePudding user response：

It is in the file slurm.conf, typically found in /etc/slurm. Generally you would edit this file and distribute to all nodes in the cluster. When adding a node you'll need to restart surmctld process.Documentation at slurm.schedmd.com/slurm.conf.html#lbAI

For example: PartitionName=debug Nodes=node[1-4] MaxTime=1800 AllowGroups=users State=UP Default=NO

This line would allow users who are members of the group users access to the debug partition with node1 node2 node 3 node4 available for jobs.

The nodes specified must each be or collectively defined in a node statement in slurm.conf as well. Documentation at slurm.schedmd.com/slurm.conf.html#lbAE

NodeName=node[1-4] CPUs=24 Sockets=2 CoresPerSocket=12 ThreadsPerCore=1 RealMemory=1000000 TmpDisk=40000 State=UNKNOWN

You would want the resources specified to match each/all machine(s).

Good luck, Endo