15.5 Handling Power-Related Events
Context
Power actions are considered NODEMODIFYURL events and are not recorded by default, but you can configure Moab to include power-related events in the logs. Also, if a node fails to turn on (or off), it's best to associate a reservation on the failed node so that Moab won't keep trying to perform the power action over and over.
-
Modify the RECORDEVENTLIST parameter.
RECORDEVENTLIST +NODEMODIFY
Power-related events are logged to the Moab log file.
To put a reservation on a node that fails to perform a power action
-
Configure the NODEFAILURERSVPROFILE attribute of RMCFG and create an RSVPROFILE with a high duration.
RMCFG[torque] NODEFAILURERSVPROFILE=failure RSVPROFILE[failure] DURATION=3600
Nodes that fail to power on or off have a 1-hour reservation placed on them.
Related Topics