Accessing and Navigating the Cluster

1.4. Accessing and Navigating the Cluster#

There are two methods for accessing the Kempner Institute HPC cluster:

To utilize the resources, one may need to install OpenAuth two-factor authentication. A VPN connection is required to use Open OnDemand.

1.4.1. Access Requirements#

1.4.1.1. Installing OpenAuth 2FA#

FASRC uses its own two-factor authentication tool, OpenAuth, which is different than Harvard Key.

To use OpenAuth, follow these steps:

  1. Request OpenAuth: Click here to request OpenAuth instructions and installation links from FASRC. Use the following credentials:

  2. Install 2FA Tool:

    • Open the Email from FASRC: After submitting a request for OpenAuth, FASRC will send a message to the email address on record. (Since FASRC uses email verification for authentication, users must have an email address on record with FASRC.)

      The email will contain:

      • Installation instructions

      • Download links (valid for 24 hours)

      • A personalized token

    • Select and Install 2FA Tool: Choose among these options:

      • Smart phone app: uses a QR code for use in Google Authenticator (Android or IOS)

      • Duo Mobile: tokens are separate from Harvard Key tokens

      • Java Desktop App:

        • Java runtime is required. Installation instructions are here

        • Double-click the Jauth.jar file to start the Java applet. Alternatively, use one of the installation scripts for Windows (.bat) and Linux (.sh).

Credit for OpenAuth instructions: FASRC OpenAuth site.

1.4.1.2. Setting Up a VPN Connection#

To set up a VPN connection, follow these steps:

Skip Steps 1 and 2 if you have the Cisco AnyConnect VPN Client already installed on your local machine.

  1. Download the VPN Client: Download the VPN client from the https://vpn.rc.fas.harvard.edu. Use username + @ + fasrc (e.g., jharvard@fasrc) for USERNAME field.

  2. Install the VPN Client: Install the VPN client on your local machine.

  3. Connect to the VPN: Open the VPN client and enter vpn.rc.fas.harvard.edu and click on the connect button.

  4. Enter your credentials:

    • username: username + @ + fasrc (e.g., jharvard@fasrc)

    • password: your FASRC password (reset your password here)

    • OpenAuth 2FA code: your OpenAuth 2FA code (See Installing OpenAuth 2FA)

Note

For detailed instructions and additional information, please read more about the VPN at VPN setup guide on FASRC Documentation.

Warning

Please avoid logging in directly to specific login nodes, as this can disrupt the load balancing on the nodes and may exert undue pressure on a particular login node (e.g., holylogin01 through holylogin04). An exception exists for VSCode, which requires rotation between login nodes. Please refer to the VSCode Section for more information.

1.4.2. SSH Access#

You can access the cluster using an SSH client. Use the following command to connect to the cluster:

ssh <username>@login.rc.fas.harvard.edu

Replace <username> with your username on the cluster. This command will prompt you to enter your password and an OpenAuth 2FA code. After entering your password and a 2FA code, you will be connected to one of the login nodes of the cluster.

Note

Please do not use the login node for computationally intensive tasks, as these operations demand considerable computational resources and can severely impact the system’s performance for all users. The login node is intended for file management, job submission, and other lightweight tasks only.

1.4.3. Open OnDemand#

Open OnDemand is a web-based interface that allows users to access the cluster through a web browser. You can use Open onDemand to start a Remote Desktop session, launch Jupyter notebooks / JupyterLab, RStudio, or other available interactive applications and submit jobs to the cluster.

To access the cluster using Open OnDemand, follow these steps:

  • Connect to the VPN

  • Open a web browser and navigate to the following URL: http://vdi.rc.fas.harvard.edu/.

  • Enter your username and password to log in to the Open OnDemand interface.

Refer to the dedicated Open OnDemand section to learn more.

Contents