Re: [Opaltoolkit-users] problem launching opal jobs

[Malcolm T. <mt...@wu...>]

Luca,

On Thursday 25 June 2009 12:24:21 pm Luca Clementi wrote:

> The problem you are experiencing is due to the fact that you don't have
> available the libdrmaa.so on your LD_LIBRARY_PATH. This library should
> be shipped with your scheduler.
>
> Please check that
> echo $LD_LIBRARY_PATH
> points to the base directory where you have your libdrmaa.so. If not you
> have to add the path to you LD_LIBRARY_PATH and restart tomcat.

It wasn't clear to me at what point this had to be set.  Is it during the 'ant 
install' phase, or when I start Tomcat (I did this as 2 different users).  
Since I didn't see any complaints about missing symbols, I had assumed 
everything is OK.

I stopped tomcat, made sure that LD_LIBRARY_PATH has libdrmma in it's path, 
and restarted tomcat:

-bash-3.1$ ls $LD_LIBRARY_PATH
libattr.a    libdrmaa.so.0      libsite.a       libtorque.so.0.0.0
libcmds.a    libdrmaa.so.0.0.0  libtorque.a     libtorque.so.2
libdrmaa.a   liblog.a           libtorque.la    libtorque.so.2.0.0
libdrmaa.la  libnet.a           libtorque.so    pbs_sched.a
libdrmaa.so  libpbs.a           libtorque.so.0

-bash-3.1$ /usr/local/src/jakarta-tomcat-5.0.30/bin/startup.sh 

I'm still seeing the same error when I try to submit a job from the client.

> I'm not an expert of torque but taking a look at the documentation it
> seems that you have to use routing queue to specify the default queue
> for the opaluser.
>
> Please have a look at
> http://www.clusterresources.com/torquedocs21/4.1queueconfig.shtml
> You should route all the job to
> opal_queue first and then to all the other queues. Then opal_queue
> should have the acl_users enabled only for opaluser.

This sounds like it might work.  I currently have a couple of queues set up, 
and I wanted to make sure that the existing way people submit jobs continue 
to work.  

Malcolm


-- 
Malcolm Tobias
314.362.1594