[torqueusers] epilogue usage
garrick at clusterresources.com
Thu Jun 29 10:55:31 MDT 2006
On Thu, Jun 29, 2006 at 10:56:37AM +0200, Guillaume Alleon alleged:
> Hi there
> Is it considered a safe usage to use the prologue return value (> 0)
> to requeue a job if some condition is not met.
> Would the other jobs in the queue be considered for starting - or will
> this be in an endless loop until the prologue returns 0.
The job is put back into the Q state for the scheduler to run again.
The next scheduler iteration, the scheduler will likely rerun the job
again, possibly on the same nodes, excluding other jobs from running.
pbs_sched will probably do this forever. Maui will probably eventually
defer the job. Moab is probably infinitely flexible to do whatever you
This mechanism is pretty good for some transient error, but I wouldn't
start enforcing policy with it.
More information about the torqueusers