1. HOME
  2. Access
  3. Datahub in the Browser

Draft

This documentation is in draft and under active review. Figures for the Fall 2026 GPU reservation system are not yet published, and some pages describe behavior that has not been verified against a primary source. Confirm any figure with datahub@ucsd.edu before relying on it.

Datahub in the Browser

datahub.ucsd.edu provides browser-based access to course environments. It requires a browser, UCSD campus credentials, and Duo; no local installation or command-line experience is needed.

Signing In

  1. Open datahub.ucsd.edu.
  2. Sign in through UCSD single sign-on, with campus credentials.
  3. Approve the Duo prompt.

Datahub uses the standard campus sign-on, the same one used by other UCSD services. Nothing Datahub-specific is entered at sign-in. An account that cannot get past the campus sign-in page has a credential problem, not a Datahub problem, and the ITS Service Desk handles it.

The Active Directory username is required at a different prompt, ssh to the login node, which does not accept the full address. That prompt is described in Connecting over SSH.

Selecting a Course & Environment

After sign-in, Datahub lists the courses available to the account and, within a course, the environments configured for it.

Course List

Each Datahub course appears separately and holds its own files. Students and instructors enrolled in more than one course select among them in this list. Membership in more than one workspace is described in Belonging to Several Workspaces.

Environment Menu

The options in the environment menu are set by the course instructor. The menu commonly holds one CPU option, and a GPU option in courses that use GPUs. Environments not present in the menu are added by the instructor. The settings a course workspace controls are described in What a Workspace Is and What It Controls.

Startup Time

An environment takes one to two minutes to start, and longer while the cluster is busy. The live state of the cluster is shown on The Status Page.

The Browser Session

A browser session starts with 2 CPU cores and 4GB of RAM. This is the course spawn configuration, set per course. The defaults applied by the command-line launch scripts are a separate figure that describes a different object. Command-line launches are described in Working from the Command Line.

A session provides JupyterLab, with notebooks, a file browser, a text editor, and a terminal. Files persist between sessions. Installing software within an environment, and the conditions under which an installation persists, are described in Customizing an Environment.

Concurrent Datahub Sessions

A member may have one Datahub session running at a time. The limit applies to Datahub itself, and it counts sessions, not environments or courses. Working in a different course environment requires stopping the running session first, as described in Stopping a Session.

Shell, VS Code, and Batch Jobs

The one-session limit does not extend to work launched from a shell. SSH sessions, VS Code containers, and batch jobs run alongside a Datahub session, and alongside each other, in any number. Launching a container from dsmlp-login while a browser session is open is ordinary use and does not conflict with the session.

Aggregate Resource Limits

Total CPU, memory, and GPU across everything a member has running must fit within the Kubernetes limits set on the member's namespace and, where GPUs are involved, within the reservation system's limits. A launch is refused when the total would exceed them, whatever mix of sessions makes up that total. Running several jobs together is described in Running Several Jobs at Once.

Stopping a Session

Warning

Signing out does not stop a session, and neither does closing the browser tab. The container continues to run and holds its CPU, its memory, and its GPU where one is attached.

To stop a session, use File → Hub Control Panel → Stop My Server. The interface may take a short time to reflect the change.

Stopping an Unreachable Session

Where the browser session cannot be reached at all, as with a stale profile or a page that will not load, the manual-resetter under the services dropdown at datahub.ucsd.edu stops any running servers and resets the profile, leaving files untouched. The procedure is in "Spawn Failed".

Status & Quota Pages

Two pages at datahub.ucsd.edu report storage use and cluster capacity.

Page Location Contents
Disk usage datahub.ucsd.edu/hub/spawn → Services tab → disk-quota-service Storage in use against the account quota
Cluster status datahub.ucsd.edu/hub/status Cluster nodes, the GPU models on them, and free GPU counts

Warning

A full quota prevents a session from starting and produces no error message.

The storage pools and their quotas are described in Workspace and Personal Quotas, and the cluster status page in The Status Page.

Sign-In Failures

Most failures to reach a session resolve to one of three conditions:

  • The course does not appear in the course list.
  • The disk quota is full.
  • A Datahub session is already running.

Each condition is covered in Sign-In & Session Problems.

Report a problem with this page