[ofa-general] Re: [OpenSM] How to find active routing algorithm ?

Al Chu chu11 at llnl.gov
Wed Jun 18 08:56:05 PDT 2008


Another idea would be to write an script that is setuid root, that can
rsh/ssh/etc. to the remote node an get the information via the console
or grepping /var/log.

You just have to install the script on every compute node, make sure
permissions are right, etc.

Al

On Wed, 2008-06-18 at 07:55 -0700, Ira Weiny wrote:
> I think the "easiest" method at this time is to use the console via an expect
> script on the SM node.  Get the routing algo out of the console output (see
> "status" command).  Then have that script "push" this information to all the
> nodes to be cached.  Compute nodes then simply look at this file.
> 
> You could set up a cron job to do this periodically although once OpenSM is run
> it should not change.  But this would separate this script from having to know
> when OpenSM is restarted...
> 
> That said, I really don't know why user jobs need to know this information.
> But right now the best way to get it is via the console.
> 
> Ira
> 
> On Wed, 18 Jun 2008 16:57:45 +0530
> "Keshetti Mahesh" <keshetti.mahesh at gmail.com> wrote:
> 
> > Thanks for your replies.
> > 
> > > If that were to be done, it would need to be done via a vendor specific
> > > SM query as there is no standard for IB routing algorithms or reporting
> > > this to other nodes. It would not work with other SMs or prior OpenSMs.
> > > Is this really needed over IB ? Isn't some remote access to the SM node
> > > sufficient for this and then use one of the methods previously described
> > > on the list ?
> > 
> > Yes, all the above methods are sufficient to find out the active
> > routing algorithm
> > if the user has the remote access to the SM node. But in our subnet, SM node
> > is also a compute node and when some user reserves that node for running jobs
> > no one else except 'root' is allowed to enter that node. There I am
> > facing the problem.
> > 
> > -Mahesh
> > _______________________________________________
> > general mailing list
> > general at lists.openfabrics.org
> > http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
> > 
> > To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
> _______________________________________________
> general mailing list
> general at lists.openfabrics.org
> http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
> 
> To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
-- 
Albert Chu
chu11 at llnl.gov
925-422-5311
Computer Scientist
High Performance Systems Division
Lawrence Livermore National Laboratory




More information about the general mailing list