Monday, February 9, 2009

Submit and Manage Vanilla Condor Job to Swarm

To submit Vanilla condor job(s) to Swarm, first you have to install swarm and configure the service to process vanilla condor jobs. the source code is available from
svn co https://ogce.svn.sourceforge.net/svnroot/ogce/ogce-services-incubator

For the general information to install swarm, please refer,
http://decemberstorm.blogspot.com/2008/11/swarm-sever-installation-guide.html

Now, configure your swarm for the Vanilla condor pool.
Step 1. modify swarm/TGResource.properties, to process only vanilla condor pool. This is my properties:
matchmaking=FirstAvailableResource
taskQueueMaxSize = 100
taskQueueScanningInterval = 3000
condorCluster=true
teragridHPC=false
eucalyptus=false
condorCluster_numberOfToken=10
condorRefreshInterval = 2000

Step 2. start swarm server.
Upload swarm/build/jobsub.aar to your axis2 installation.

Step 3. Client example of vanilla condor job is stored at,
swarm/src/core/org/ogce/jobsub/clients/SubmitVanillaJob.java

There are three example methods:
submitJobWithStandardOutput(): job with standard output wihout input files
submitJobWithOutputTransfer(): job with standard output and output files
submitJobWithInputOutputTransfer(): job with standard output, output files, and local input files

Step 4. Compile example file
[swarm]ant clean
[swarm]ant

Step 5. Run example file
[swarm]./run.sh submit_vanilla_job

Step 6. Check the status
[swarm]cd ClientKit
[ClientKit]./swarm status http://serverIP:8080/axis2/services/Swarm your_ticket_ID

Step 7. Get the location of output file
[ClientKit]./swarm outputURL http://serverIP:8080/axis2/services/Swarm your_ticket_ID
This will return the URL of the output file.

No comments: