Hi all
Background:
I’m new to Slurm. I Installed Slurm 16.05 (Munge auth, MariaDB managed by
slurmdbd). Users maintained by LDAP, no LDAP user imported into Sturm DB so far.
bash$ sacctmgr show user
User Def Acct Admin
--------- ---------- ---------
root test Administ+
Ldap users can run jobs successfully, sacct –u <user> shows accounting
information for past jobs. Now I want to restrict each users to only a certain
number of jobs running in a partition at a time. Therefore I created a QoS and
attached it to the partition (partition Qos):
bash$ sacctmgr create qos def_p_qos
# Add qos to partition definition in slurm.conf
PartitionName=all Nodes=node[02-04] Default=YES DefaultTime=01:00:00
MaxTime=96:00:00 QOS=def_p_qos State=UP
bash$ scontrol reconfigure
bash$ saccmgr modify qos set MaxJobsPerUser=5 where name=def_p_qos
With only this, the MaxJobsPerUser limit is not applied. Accounting and
Resource Limits Guide (http://slurm.schedmd.com/accounting.html) says:
„To enable any limit enforcement you must at least have
AccountingStorageEnforce=limits in your slurm.conf,
otherwise, even if you have limits set, they will not be enforced.“
Consequently I set
AccountingStorageEnforce=limits
In slurm.conf. Now job submission is prohibited: srun: error: Unable to
allocate resources: Invalid account or account/partition combination specified.
After adding the user to Slurm DB, job submission works, and MaxJobsPerUser
limits is being enforced. Success.
Question:
Is there a way to limit the max number of jobs per user on partition level
without adding all LDAP users to Slurm DB? From Accounting and Resource Limits
Guide:
„Without AccountingStorageEnforce being set (the default
behavior) jobs will be executed based upon policies configured in Slurm on each
cluster.“
What does that mean? Where do I define those policies? Why this extra step of
adding users to Slurm DB, how would that work without using a DB but flat files?
Many thanks,
Nico
