Fixed Issues
This section lists issues fixed in this release of Cloudera Machine Learning (CML) on Private Cloud.
- DSE-32672: CML Sessions fail intermittently with exit code 34
- Earlier, CML sessions failed because the timezone of the web and database pods was off by one hour (local time zone vs UTC), causing the timestamps of the session to be inconsistent and causing the reconciler to terminate the pod. This issue has been fixed.
- DSE-32639: Installation and upgrades failing due to environment name character restriction
- In Private Cloud Data Services 1.5.2, the supported length for environment names was restricted between 5 and 28 characters for creating and upgrading workspaces, causing failures. This issue has been fixed. You no longer need to restrict the environment name length to 28 characters.
- DSE-34100: Migration readiness check fails on BuildKit-enabled cluster
- Earlier, the migration readiness checks failed on clusters having BuildKit instead of Docker. This issue has been fixed.
- DSE-31914: The model metrics ingress configuration does not support TLS
- This issue has been fixed.
- DSE-32250: Spark executor usage is not reported correctly
- Earlier, when you exported the Spark executor usage data from the tab, the exported data in the report was same as the session in which the executor was launched. This included incorrect CPU, memory, create time, and stop time details of the Spark executors. This issue has been fixed.
- DSE-33098: Unable to synchronize teams in CML without having an MLAdmin user in that team
- Earlier, CML required that each team to have an MLAdmin user to manage the team. This is no longer a hard requirement. You can create teams of users with MLUser-only roles.
- DSE-33765: CML workspace upgrdes failing due to buggy logic in fetching the Helm values
- This issue has been fixed.
- DSE-34088: Unable to view Spark UI executor logs in CML
- This issue has been fixed.
