It is possible to trigger an event in the backup slurmctld after it takes control (see "man strigger" :-c, --backup_slurmctld_resumed_operation Trigger an event when the backup slurmctld resumes operation after failure."), but there is not currently a mechanism to execute a program before it assumes control. If such logic is to be added, it would be quite simple and go into src/slurmctld/backup.c around line 200.
Quoting Tyanko Aleksiev <[email protected]>: > Hello, > > I started exploring Slurm few weeks ago as we're planing to use it on one > of > our clusters. I have the following DRBD+Heartbeat issue now: > > I've configured the JobCheckpointDir to be synced by DRBD between the > Controller and the > BackupController hosts together with Heartbeat for the HA resources > management. > > Executing: > # scontrol takeover > on the BackupController works out without problem except that in the case > when only the slurmctld daemon (running on the Controller) is dead but the > host > itself remains up&running. In such cases Heartbeat WON'T execute the > takeover > of the DRBD resources and the JobCheckpointDir WON'T be synced on the > BackupController. > > So, is there a way to trigger an action or execute a command before the > BackupController host > takes over the control? > > Thanks a lot! > Tyanko Aleksiev >
