Connecting to Penobscot with Open OnDemand


ACG's Open OnDemand (OOD) server is a web portal which provides convenient access to UMS HPC resources via a web browser. Most users find getting started with HPC using Open OnDemand is easier than using SSH and VNC sessions. Tasks such as managing files, submitting compute jobs and running interactive apps (MATLAB, RStudio, etc) can all be accomplished in a point-and-click interface within one or more browser tabs. ACG's OOD portal provides access to compute resources in the Penobscot cluster. 

Off-Campus Access

The ACG systems are visible from within the UMS networking system, but they are firewalled from outside traffic. The University of Maine System has a VPN service that is available to anyone with a email account. Once connected to this VPN, you can connect to the ACG login nodes. The URL for the VPN is:

and you can login to it with your credentials (what you use for your email and other services).

Connecting and Logging In

To access ACG's OnDemand portal, open your web browser and point it to the address below.

Log in with your UMS IT username and password, the same credentials you use to log in to your email account. If you need help with access, email If you haven't used ACG HPC resources before, you'll need to register. Once logged in, your ACG HPC account username will be shown in the top right of the navigation bar. 

Uploading and Downloading Files

Open OnDemand includes a file manager app with an easy, point-and-click user interface for common actions like moving, copying, renaming, uploading, downloading and deleting files. It also includes a simple text editor. Select Files | Home Directory from the top navigation menu to launch the Files app.

Your Home Directory for the Files app is the root directory of your ACG HPC account storage, i.e., where you are when you first log in to Katahdin or Penobscot/Login1 via SSH. 

Viewing Active Jobs

To view the active jobs on the Penobscot cluster, select Jobs | Active Jobs from the top navigation menu. Active jobs are those that are queued, running or recently completed. Initially the app will only display your jobs, but you can change that to all jobs using the button at the top right.

Starting a Desktop Session

It's easy to start a graphical desktop session on a compute node. Select Interactive Apps | Penobscot Cluster Desktop from the top navigation menu. Desktop sessions run inside slurm jobs, and jobs on the Penobscot cluster are currently limited to two weeks / 336 hours (this will be increased in the future). If you will be performing light work like editing text files and submitting slurm jobs, select 1-4 cores on a node in the Haswell partition. If you will be doing compute-intensive work within the desktop session, you can ask for more resources, up to 96 cores on an EPYC node.

Once you press the Launch button to request a desktop session, the system will start a job on a compute node, which typically takes about 30 seconds. When the job is ready the card for the session will turn green and you can then connect to it with the Launch Penobscot Cluster Desktop button. Note that if you lose your connection to a desktop session, such as if you take your computer out of wifi range, your desktop job will continue to run for the allotted time and you can reconnect to it in the same way. 

The session will appear as a regular Linux MATE desktop within a browser tab, and will persist until you shut it down (System | Shut Down from within the desktop), cancel it (the Cancel button in My Interactive Sessions) or it reaches the job time limit.

Programs you run from the desktop are on a compute node and use the resources requested when you started the desktop session job. When you open a Terminal, it will be running on the compute node. If you're using the desktop session to submit jobs, in most cases you will want to ssh to the login node (login1) and call sbatch from there, to avoid confusing slurm by submitting jobs from within a job. 

Using Interactive Apps: RStudio, MATLAB, Jupyter Notebook

Open OnDemand provides specific interactive apps for working with RStudio, MATLAB and Jupyter. Launching one of these apps is similar to starting a desktop session, select Interactive Apps | RStudio or MATLAB/Jupyter and start a job with appropriate resources (time, cores, memory). The session will persist until you cancel it from Open OnDemand or quit the app (e.g., File | Quit Session in RStudio).