[Mauiusers] torque PAM module locks node before epilogue script runs.

Garrick Staples garrick at clusterresources.com
Wed Mar 28 16:16:50 MDT 2007


On Fri, Mar 23, 2007 at 01:30:38PM -0400, Anthony R Fletcher alleged:
> I started to look at the torque PAM module. It works just fine whilst
> the job is running on the selected nodes.
> 
> However we need to clean up the nodes afterwards. We currently use an
> epilogue script which runs on the job's master node as root. This
> becomes the user and logs on to each of the slave nodes killing
> old processes and deleting old files.
> 
> With the PAM module in place, the epilogue no longer works because the
> jobs has finished and the 'user' can no longer log on to the slaves.
> Root on the master node cannot log on to the slaves.
> 
> Is there an alternative solution? Can the epilogue script be made to run
> on all the nodes after the job finishes and not just the master node.
> Can the PAM module be configured to allow the user to log on whilst the
> epilogue script runs?

Have you tried the parallel user epilogue?



More information about the mauiusers mailing list