Skip to content

How to use the GSC Clusters

rmcclosk edited this page Sep 18, 2014 · 10 revisions

Submitting jobs with qsub

qsub is the command used to submit jobs to the cluster. You can use qstat -u username to check on your currently running jobs (replace username with your username). mqsub is a helpful wrapper around qsub to submit batches of jobs (do mqsub --help for usage). This section is based on advice from Bruno Grande.

Requesting Memory

To request 1 GB of memory, use -l mem_free=1G -l mem_token=1G -l h_vmem=1G.

Requesting multiple CPUs

To request 12 CPUs, with 1 GB of memory each, use -pe ncpus 12 -l mem_free=1G -l mem_token=1G -l h_vmem=1G. Here, the memory requirement is 1 GB per CPU, so we have requested 12 GB in total.

Using java

To run a java program, you need to specify the heap size in the call to java with the -Xmx option. For example, to run GATK with 2 gigabytes of memory, you would do

java -Xmx2056m -jar /path/to/GenomeAnalysisTK.jar args --more-args

Note that -Xmx2056m is case-sensitive, so -Xmx2056M will not work. To submit the job to qsub, allow 1 GB more memory than you specified for the heap size (so in this case, you would request 3 GB of memory).

Environment variables

If you have defined environment variables in your ~/.bash_profile which you want to use in cluster jobs (for example, adding $HOME/bin to $PATH), prepend source $HOME/.bash_profile; to your submitted jobs.

Copying files from apollo to genesis

This section is taken from an email from Rebecca Johnston.

To copy large files and/or a large number of files from apollo to genesis, use the thost queue on the apollo node.

First, create a batch file of cp commands, where the genesis path starts with /genesis/, e.g.

less copy_commands.batch

cp /home/rjohnston/test1.txt /genesis/extscratch/rmorinprj2/
cp /home/rjohnston/test2.txt /genesis/extscratch/rmorinprj2/
cp /home/rjohnston/test3.txt /genesis/extscratch/rmorinprj2/
...

Then on the apollo node, run the following command:

mqsub --file copy_commands.batch --name copy_to_genesis --chdir logs --qsub
'-q thosts.q -P transfer'

Software

On genesis, there are several useful software packages installed in /gsc/software/linux-x86_64-centos5/.