LiNUX WSL slurm Ubuntu

Slurm installation in Ubuntu@WSL

"Slurm, Linux"

Posted by QL on January 27, 2019

Notes of installing slurm in Ubuntu @WSL

Jan 27th, 2019. Based on reference1

  1. Install munge and slurm:sudo apt install munge slurm-wlm. And excuting the command hostname and slurmd -C on each compute node will print its physical configuration (sockets, cores, real memeory size, etc.), which can be use in constructing the slurm.conf file.

  2. Using a browser (firefox, opera, etc) to open /usr/share/doc/slurmctld/slurm-wlm-configurator.easy.html and generate the configuration file. Note: I am using just one node, so I used the host name for the ControlMachine, the NodeName and the ClusterName in the generating process of the configuration file. And the unit of RealMemory seems to be MB, so use 65536 for example if the node has 64GB. However, my queue got stuch in status Draining due to Low Real Memory at my first attemp, so I did not specify RealMemory on my second attempt. Finally, sudo vi /etc/slurm-llnl/slurm.conf and copy/paste the configuration file from the browser. My configuration file is: 
      1 # slurm.conf file generated by configurator easy.html.
  2 # Put this file on all nodes of your cluster.
  3 # See the slurm.conf man page for more information.
  4 #
  5 ControlMachine=workstation #<YOUR-HOST-NAME>
  6 #ControlAddr=
  7 #
  8 #MailProg=/bin/mail
  9 MpiDefault=none
 10 #MpiParams=ports=#-#
 11 ProctrackType=proctrack/pgid
 12 ReturnToService=1
 13 SlurmctldPidFile=/var/run/slurm-llnl/slurmctld.pid
 14 #SlurmctldPort=6817
 15 SlurmdPidFile=/var/run/slurm-llnl/slurmd.pid
 16 #SlurmdPort=6818
 17 SlurmdSpoolDir=/var/lib/slurm-llnl/slurmd
 18 SlurmUser=slurm
 19 #SlurmdUser=root
 20 StateSaveLocation=/var/lib/slurm-llnl/slurmctld
 21 SwitchType=switch/none
 22 TaskPlugin=task/none
 23 #
 24 #
 25 # TIMERS
 26 #KillWait=30
 27 #MinJobAge=300
 28 #SlurmctldTimeout=120
 29 #SlurmdTimeout=300
 30 #
 31 #
 32 # SCHEDULING
 33 FastSchedule=1
 34 SchedulerType=sched/builtin
 35 #SchedulerPort=7321
 36 SelectType=select/linear
 37 #
 38 #
 39 # LOGGING AND ACCOUNTING
 40 AccountingStorageType=accounting_storage/none
 41 #AccountingStoragePass=/var/run/munge/global.socket.2
 42 ClusterName=workstation #<YOUR-HOST-NAME>
 43 #JobAcctGatherFrequency=30
 44 JobAcctGatherType=jobacct_gather/none
 45 #SlurmctldDebug=3
 46 SlurmctldLogFile=/var/log/slurm-llnl/slurmctld.log
 47 #SlurmdDebug=4
 48 SlurmdLogFile=/var/log/slurm-llnl/slurmd.log
 49 #
 50 #
 51 # COMPUTE NODES
 52 NodeName=workstation CPUs=36 Sockets=1 CoresPerSocket=18 ThreadsPerCore=2 State=UNKNOWN
 53 PartitionName=long Nodes=workstation Default=YES MaxTime=INFINITE State=UP
  3. Enable and start the manager slurmctld: sudo systemctl enable slurmctld and sudo service slurmtcld start. And enable and start the agent slurmd: sudo systemctl enable slurmd and sudo service slurmd start.
  4. Start the munge service: sudo /etc/init.d/munge start. Notes: others common commands sudo /etc/init.d/munge status, munge -n, reference2.
  5. Check the status slurm: sinfo, scontrol show node.
  6. Create a shell script and make it excutable: vi submit.sh 
    #!/bin/bash
sleep 30
env
  7. chmod +x submit.sh and submit the shell script: sbatch submit.sh. Then check the status of the cluster and the queue: sinfo and squeue. At last check the outpurt after 30s cat slurm-<JOBID>.out.
CATALOG
    FEATURED TAGS
    FRIENDS