Back up an ML workspace

Backing up an ML workspace preserves all files, models, applications, and other assets in the workspace, although files in external NFS-backed workspaces are not backed up by CML automatically.

  1. In the Workspaces UI, find the workspace to back up. The workspace must be in the Installation completed state, otherwise backup is disabled.
  2. Enter the workspace, and manually stop all workloads (sessions, jobs, applications, and models).
    For external NFS backed workspaces, manually back up the configured external NFS data to another location. This manual backup of the NFS data will be used when this particular backup is restored in future. Ignore this step if the workspace is configured with internal NFS, as internal NFS data is backed up and restored automatically by CML.
  3. In the Actions menu for that workspace, select Backup Workspace.
  4. In the Backup Workspace modal, enter a Backup Name to identify the workspace, and enter an appropriate timeout value.
  5. Click Backup to start the process.
The workspace shuts down, and the backup process begins. The workspace state changes to reflect the ongoing backup progress. If necessary, click Cancel to cancel the backup process. The backup process can take some time to complete, depending on the amount of data to copy.

Monitoring event logs

You can monitor the progress of the backup process by checking the event logs. In the Actions menu for the workspace, click View Event Logs, and then on the Events & Logs tab, click View Event Logs again for the latest backup event.

When the backup process completes, the workspace enters the Installation completed state again.

If there were issues during backup, appropriate error messages will be displayed in the event logs. However the workspace will recover from failure and will be reverted back to the original state when backup was triggered.