Logs for ML Workspaces

You can access logs to troubleshoot issues with the CML service and your workloads on ML workspaces.

Access to logs

When you create a CDP environment, you specify an S3 bucket (on AWS) or an Azure Storage container (on Azure) in that environment for storing logs. If you have access to the log storage, you can use the AWS or Azure console to access certain CML and Spark logs directly. You can get the details of the specific bucket or container from the environment's Summary page.

ML Workspace access to logs

CML workspace users also have access to these logs depending on their authorization level:

  • Site Administrators

    Site administrators can download the same logs directly from their workspace Site Admin panel (Admin > Support). For more details, see Downloading Diagnostic Bundles for a Workspace.

  • Data Scientists

    While data scientists don't have access to the full set of workspace logs, they do have access to engine logs for their own workloads (sessions/jobs/experiments). While in an interactive session or on a job/experiment's Overview page, click Download Logs at any time to review the full set of logs for that workload's engine. In the case of Spark workloads, Spark executor and event logs are also downloaded as part of this bundle.