Previously there were multiple versions of Python and Jupyter Notebook offered in OnDemand apps, but that is no longer the case. Users can either use CCR's latest installation of Python and Jupyter or their own custom environment.
- Slurm Account: Your default Slurm account will be used if nothing else is entered. To see what Slurm accounts you have access to, use the 'slimits' command in a terminal window.
- Partition: Use the drop-down bar to make a selection for partition. To get more details about the partitions in the UB-HPC cluster click here.
- QOS: Quality of Service value - this will be the same as partition unless you have access to a QOS priority boost. More info can be found here.
- Number of hours: Most partitions have a max walltime of 72 hours. Debug partition max=1 hour.
- Number of nodes: Only select more than one node if you're sure your program can run on multiple nodes.
- Number of cores: Only select more than one core if you're sure your program can run on multiple CPUs.
- Node Features/Tags: All nodes have tags to distinguish various hardware types. Use the 'snodes' command in a terminal window to see available tags.
- Amount of memory per node: If not specified, Slurm will allocate the minimum amount of RAM for your job, based on the type of node selected and the number of cores requested. If your job requires more than this, specify the amount in MB.
- Exclude host: If you want to ensure your job doesn't run on a specific node, enter the node short name here. This is in the form of: cpn-rack#-u# i.e. cpn-u22-01
Once you click the Launch button, you will be redirected to the session information where you can see the status of the job. This one is queued:
If you have elected to receive an email when your job starts, you can leave the CCR OnDemand portal until the email arrives and then log back in to open your session.
This shows the session was successfully created and is starting the notebook on the node:
When a session is ready to launch, it will look like this in the Interactive Sessions section of CCR OnDemand:
Click on the blue button "Connect to Jupyter" and you will see your Jupyter Notebook window on the node you've been allocated. You are now logged into the notebook and running on a node in the CCR UB-HPC cluster. You can begin your work!
To launch a new notebook, click on the 'New' drop down menu and select "Python 3" or your own kernel, if running a custom installation.
To launch a terminal window, click on the 'New' drop down menu and select Terminal. This will open a terminal window on the node your notebook is running on.
To learn how to delete or end a running session click here.