- Main page: Job Control
If you are having trouble with job submission, job scripts, or even parallel code debugging, an important troubleshooting tool for you is the test queue. One to three nodes at each site are set aside as a rapid-access test resource. To use them your job must have a run-time equal to or less than an hour, and it must request the "test" resource explicitly. Here is how you can do this on the command line while submitting the job
$ qsub -l test=true,h_rt=00:01:00 submission-script.sh
The alternative would be to specify the request for the test resources in the submission script itself like so:
#$ -l h_rt=00:01:00,test=true
Jobs with time limits less than one hour which do not explicitly supply
test=true will not go into
test.q. Jobs requesting the
test resource but more than an hour of run time will not run. The number of parallel slots must also not exceed the size of the test queue at that particular site: you can use qsum to find the size of
test.q at the cluster you are using.
You are allowed to run test code from the command line on the head node. This method is suitable for test programs that run for only a short period of time (e.g. less than 10 minutes) and require a small amount of memory. More complex test jobs must be submitted to Grid Engine as production jobs.
You can also request test nodes through Grid Engine for interactive jobs.