User Tools

Site Tools


maintenances:baobab-202111

Action pré-maintenance

Maintenance

Update slurm1

Désactiver les services slurm:

(baobab)-[root@slurm1 ~]# systemctl disable slurmdbd
(baobab)-[root@slurm1 ~]# systemctl disable slurmctld

Stopper les services slurm:

(baobab)-[root@slurm1 ~]# systemctl stop slurmctld 
(baobab)-[root@slurm1 ~]# systemctl stop slurmdbd

Update Yum:

(baobab)-[root@slurm1 ~]# yum versionlock delete *slurm*
(baobab)-[root@slurm1 ~]# yum update 

Si tout est OK, rallumer slurmdbd pour lancer la conversion de la base (surveiller les logs):

(baobab)-[root@slurm1 ~]# tail -f /var/log/slurm/slurmdbd.log
(baobab)-[root@slurm1 ~]# systemctl start slurmdbd

Une fois la conversion fini, éteindre slurmctld et lancer salt-check slurm pour vérifier, si tout est OK ⇒ salt-apply slurm

(baobab)-[root@slurm1 ~]# salt-call --state-output=changes state.sls slurm test=true
(baobab)-[root@slurm1 ~]# salt-call --state-output=changes state.sls slurm

Vérifier les services slurm et la conf:

(baobab)-[root@slurm1 ~]# systemctl status slurmctld slurmdbd
(baobab)-[root@slurm1 ~]# less /etc/slurm/slurm.conf

Valeur a vérifier (sans ça le service ne démarre pas):

  • TaskPlugins=task/affinity
  • AccountingStoreFlags=job_comment

Enfin lancer le salt-apply:

(baobab)-[root@slurm1 ~]# salt-call --state-output=changes state.apply
maintenances/baobab-202111.txt · Last modified: 2021/11/25 12:08 by Adrien Albert