Jupyter Launcher

The Jupyter launcher provides a simple interface to start Jupyter Notebook servers on the DTU HPC cluster.

Getting started

To start the launcher, go to Tools and select Jupyter launcher. You need to be logged in to access the launcher, as it will submit a job to the DTU HPC to start the server.

Here you'll see the interface with the basic options that are needed for the server:

• CPUs Specify the number of CPUs to request. A CPU (Central Processing Unit) is the core computational unit of a computer responsible for executing instructions. Requesting more than one CPU is only beneficial for tasks explicitly designed for parallel processing or multiprocessing.

• Memory (GB) Specify the amount of memory to request. Memory refers to the amount of RAM allocated for processing data. Requesting more memory is useful for handling large datasets or memory-intensive tasks, but unnecessary for smaller workloads. A good rule of thumb is to request memory equal to twice the size of your dataset. Please be conservative with the amount of requested memory - usually 8GB is enough.

• Wall time Specify the maximum run time for the server. This is the longest duration your server is allowed to run. Once the wall time is exceeded, the server will be terminated, and any unsaved changes or active kernels will be lost.

• Root directory Specify the starting directory for the server. You will only have access to files and folders within and below this directory in the hierarchy. By default, it starts in your home directory at DTU, but you can enter any accessible path if you need to work outside of it.

Once you've selected your options, click Launch Jupyter Server to submit the job to the cluster. This process typically takes around 30 seconds, provided the queue isn't too busy. If it takes longer than one minute, try refreshing the page and submitting the request again. If the issue persists, please contact us for assistance.

When the server is ready, a button labeled Open Jupyter Server will appear:

Click this button to open the Jupyter interface in a new tab. If you need a server with different configurations, change your configurations and then press the Re-launch Jupyter Server button. Note, that only one instance of the Jupyter server can run at a time and the old server will be terminated by re-launching.

Finishing your session

When you are finished with your Jupyter server, the session can be killed in one of two ways:

1) In the Jupyter Notebook file directory interface, select File -> Shut down

2) On the QIM platform, select HPC cluster after pressing your profile image in the top right corner. Under the header 'Your jobs' you can select "Kill job job-number".

Detailed configuration

To access more options regarding the Jupyter Server, press the checkmark Show detailed configuration:

• Hostname Defines the address used for SSH connections and job submissions. Change this to connect to a specific server.

• HPC queue Determines what HPC queue to enter. If you also need to access to a GPU, select the appropriate node. Note that you may not have access to all the listed HPC queues.

• Job name Provides a name for the submitted job. Change this if you need a more descriptive or specific name.

• Conda environment Specifies the name of a pre-installed Conda environment to activate. Leave this blank to use a custom environment.

• Check environment Selects a pre-installed environment to see a list of installed packages. These will be shown below the "Launch Jupyter Server" button.

• Source path Defines a path to source a custom environment. For example, the default qim3d environment accesses the source path: "/dtu/3d-imaging-center/QIM/conda/miniconda3/bin/activate". This field is ignored if a Conda environment is selected in the "Conda environment" dropdown.

• Custom environment Specifies the name of the custom Conda environment to activate. This activation will occur after the source path is sourced. For example, entering "qim3d" here will run 'conda activate qim3d' in the terminal.

• Reset SSH tunnels Defines if SSH tunnel should be reset. By default, the SSH tunnel between the cluster node and the platform is reset each time a new server is launched. If you experience connection issues, try unchecking this box to see if it resolves the problem.

• Jupyter server type Defines if the created session activates a Jupyter Notebook or JupyterLab.

• Show logs Enables visibility of logs from the Jupyter Server. The logs will appear after a job is submitted.

Logs

The logs provide detailed information that is updated throughout the session. Here, you'll find additional information about the job script, requested resources, and the node where the server is running. If you're experiencing connection issues, check the logs for useful debugging information that might help resolve the problem.