Re: [Opaltoolkit-users] problem launching opal jobs

[Luca C. <lcl...@uc...>]

Malcolm Tobias wrote:
> I'm attempting to set up our own local pdb2pqr Opal server following the
> instructions at:
> 
> http://www.nbcr.net/software/opal/docs/2.0/installation.html
> 
> After customizing the XML configuration to point to our local pdb2pqr, I 
> deployed the service:
> 
> ant deploy -DserviceName=pdb2pqr -DappConfig=./pdb2pqr_config.xml 
> 
> I then tried to use the client to launch a job:
> 
> java edu.sdsc.nbcr.opal.GenericServiceClient -l 
> http://agave:8080/opal2/services/pdb2pqr -r launchJob -a "--ff=amber 
> sample.pdb output.pqr" -f samples/sample.pdb
> 
> Reading command line arguments
> Service URL: http://agave:8080/opal2/services/pdb2pqr
> Invoking operation: launchJob
> 
> Command line arguments: --ff=amber sample.pdb output.pqr
> Making non-blocking invocation on Opal service -
> Exception in thread "main" AxisFault
>  faultCode: {http://schemas.xmlsoap.org/soap/envelope/}Server.generalException
>  faultSubcode: 
>  faultString: 
>  faultActor: 
>  faultNode: 
>  faultDetail: 
>         {http://nbcr.sdsc.edu/opal/types}opalFaultOutput:<message>Can't 
> instantiate Opal Job Manager - Job manager class 
> edu.sdsc.ncbr.opal.manager.DRMAAJobManager not found</message>

Tobias,
you have to typo, it's nbcr not ncbr.

>         
> {http://xml.apache.org/axis/}exceptionName:edu.sdsc.nbcr.opal.FaultType
>         {http://xml.apache.org/axis/}hostname:agave.wustl.edu
> 
> The job manager that I have specified in my pdb2pqr_config.xml file is:
> 
> <jobManagerFQCN>edu.sdsc.ncbr.opal.manager.DRMAAJobManager</jobManagerFQCN>
> 
> The corresponding job manager defined in my etc/opal.properties is:
> 
>    opal.jobmanager=edu.sdsc.ncbr.opal.manager.DRMAAJobManager
> 
> Any ideas on what might be wrong?
> 
> 
> A second question that I'll need to address at some point is how to customize 
> the batch script that is used for launching the jobs.  We have an existing 
> cluster (running PBS/torque) and I'd like to define a property that denotes 
> that some special nodes are available for running Opal jobs.  If I were doing 
> this manually, instead of specifying:
> 
> qsub -I -lnodes=1
> 
> I'd do this by specifying:
> 
> qsub -I -lnodes=1:opal
> 
> Is there some way to customize how Opal talking to the queuing system via 
> DRMAA?
> 

Not really. Unfortunately we use DRMAA so that we don't have to generate 
  submission scripts for each scheduler (DRMAA is an API library).

On our machine we don't use PBS but SGE and to specify a default queue 
on our cluster we have used a SGE configuration file that can be placed 
in the home of the user called .sge_request.
Below an example we use to specify that opaluser (the user running 
opal-tomcat) should use the urgent queue by default. I'm sure that PBS 
has something similar.

Please let us know if you have any problem with torque/opal, since we 
never tested that configuration.

Sincerely,
Luca


[opaluser@kryptonite ~]$ tail .sge_request

-l urgent