HP XC System Software Administration Guide Version 2.1

SLURM allows you to collect and analyze job accou
nting information. Section 11.4.3 describes
how to configure job accounting information o n
the HP XC system.
Section 16.2 p rov ides SLURM troubleshooting informatio n.
11.2 Configuring SLURM
The HP XC provides glo bal and local director
ies for SLUR M files:
•The/hptc_cluster/slurm directory is the sharable locatio n fo r SL URM files
that need to be shared among the nodes. SLURM state files, job logging files, and the
slurm.conf configur ation file reside there.
• The l ocati on for S L URM files that should rem ain local to a given node is /var/slurm;
the files in this directory are not be shared between nodes. SLURM daem on logs and
daemon state inform ation are stored there.
All SLUR M configuration options are s
et and s tored in the /hptc_clus-
ter/slurm/etc/slurm.conf file; the
slurm.conf
(5) manpage provides information
on available options. The slurm.con
f file also contains useful commentary on the purpose
of each setting.
The following SLURM configuration settings are preset statically on HP XC systems:
StateSaveLocation=/hptc_cluster/slurm/state
SlurmdSpoolDir=/var/slurm/state
SlurmctldLogFile=/var/slurm/log/slurmctld.log
SlurmdLogFile=/var/slurm/log/slurmd.log
SlurmctldPidFile=/var/slurm/run/slurmctld.pid
SlurmdPidFile=/var/slurm/run/slurmd.pid
AuthType=auth/munge
JobCompType=jobcomp/filetxt
JobCompLoc=/hptc_cluster/slurm/job/slurm.job.log
JobAcctType=jobcomp/log
JobAcctLoc=/hptc_cluster/slurm/job/jobacct.log
JobCredentialPrivateKey=/opt/hptc/slurm/etc/keys/.slurm.key
JobCredentialPublicCertificate=/opt/hptc/slurm/etc/keys/slurm.cert
ReturnToService=1
Table 11-1 displays the SLURM configu rat
ion settings that are set (an d, if necessary, adjusted)
during the e x ecution of the cluster_conf
ig utility.
Table 11-1: SLURM Configuration Settings
Setting
Default Value*
ControlMachine
Lowest-numbered resource_management node
BackupController
Second-lowest resource_management node (if available)
SlurmUser ’slurm’
NodeName
all compute nodes, plus ’Procs=2’
PartitionName
’lsf RootOnly=YES Shared=FORCE Nodes=com-
pute_nodes
* Default values can be adjusted during i nstallation
HP recom mends that you adjust the NodeName manually after executing the
cluster_config utili ty to ensu re the correct processor count Procs for the compute
nodes, and to add the RealMemory setting with the expected amount of mem ory (in MB) for
all nodes. Here is an example of a completed NodeName entry:
NodeName=n[1-128] Procs=2 RealMemory=3456
11-2 SLURM Administration