Skip to content

Welcome to the CQLS HPC Documentation Website

The Center for Quantitative Life Sciences (CQLS) has established a new high-performance computing (HPC) infrastructure (hpc.cqls.oregonstate.edu, aka the 'Wildwood' HPC Cluster) that is connected to, but distinct from, the 'old' computing infrastructure (shell.cqls.oregonstate.edu).

Connecting to the Wildwood HPC cluster

Note

Instead of using your old cqls password, you will be using your ONID password for validating your account.

The default port 22 is sufficient to connect to the Wildwood HPC. An alternatively defined port is uneccesary on the command-line.

If you are hard-wired on campus and attempting to connect, or logged in to the campus VPN, you can directly connect to the Wildwood HPC cluster through ssh:

ssh hpc.cqls.oregonstate.edu

Warning

If you attempt to connect directly from an off-campus location, your connection will fail!

For more information, see the Connecting page.

Accessing the newly installed software

More information can be found in the Configuration section.

To get up and running straight away, run this tool:

hpcman user update-dotfiles -c all
exec $SHELL

This will modify your dotfiles so that the latest installations are preferred over the legacy software. New software is installed in /local/cqls/software/$ARCH where $ARCH depends on the cpu architecture of the machine you are logged into. You can find more information in the Software section.

Running jobs on the cluster

We are using Slurm for submitting jobs on the infrastructure. If you are familiar with SGE_Batch or SGE_Array, or more recently, hqsub (more information here), you can continue using hqsub for submitting jobs on the cluster. See the documentation, or run hqsub --help from the command-line to get a run-down of how to use the software.

hqsub is built to use SGE or Slurm, and will use Slurm by default on the Wildwood HPC.

Interactive jobs

Using SGE, one could check out an interactive session on a compute node using the qrsh command. The replacement for this command in Slurm is the salloc command. See here for more information.

Finding compute resources

You can use the hqavail command to see what servers you have available to submit jobs to. You can use the hqsub -q $QUEUENAME option to send jobs to a specific queue.

Monitoring jobs

Use the hqstat command to see what jobs are running at the current moment. Run scontrol show job $JOBID to get more information about your jobs that are currently running.

Terminating jobs

Use the scancel JOBID command to terminate a single job. To kill all of your pending or running jobs use scancel --me -t pending or scancel --me -t running, respectively.

Account request

See our old account request form to request an account. An updated form is in the works.

Support request

See our old support request form to request support or software installs. An updated form is in the works.