Re: [Opaltoolkit-users] DRMAA failure on SGE (Starcluster)

[Charles G. <ce...@uw...>]

Hi Luca, Hi Nadya,


> On Jan 23, 2015, at 12:15 PM, Luca Clementi <luc...@gm...> wrote:
> 
> On Fri, Jan 23, 2015 at 9:18 AM, nadya williams <na...@sd...> wrote:
>> HI Charles,
>> 
>> can you submit and execute jobs successfully outside of opal using age submission?
>> If not, first need to make sure that your age configuration is correct.
>> 
>> If yes, check  how is the opal-jobs/ created.
>> This directory needs to be NFS-mounted on all the nodes from the fronted (or opal server)
>> node of the cluster. For example,
>>        /opt/tomcat/webapps/opal-jobs -> /share/opal/opal-jobs
>> And /share/opal/opal-jobs is NFS mounted.
>> 
>> Check your in opal.properties:
>>        tomcat.url  - FQDN of your  cluster fronted
>>        drmaa.queue - set to your default SGE queue
>>        drmaa.pe - set to parallel environment of your SGe configuration
>> 
>> Are you running on ec2?
>> The starcluster is not the same “cluster" as we have in rocks. It is a group of VMs,
>> and in rocks we have a computing cluster. I don’t know how the SGE configuration and
>> inter-node communication is handled.
>> My guess would be if the SGE is working from the command line and SGE obs are running
>> correctly, then the opal configuration should just follow your SGE specifics and the above 4  variables
>> should take care of it.
> 
> Hey Charles,
> you might want to take a look in the SGE message log. I don't know
> where starcluster places it though.

Thanks for your input. I was able to submit jobs to the cluster on the command line, but was never able to get DRMAA to work.

I turned on Condor on the cluster and switched the Opal job manager setting to Condor, and that seems to be working.

Charles