Fixed issues in 1.5.5 SP2
Cloudera AI on premises 1.5.5 SP2 addresses issues previously identified as known issues.
Model serving
- DSE-44238: Cannot create Cloudera AI Inference service application deployment using CDP CLI when Ozone credentials are passed
-
The Cloudera AI Inference service can now be created using CDP CLI as well. This issue is now resolved.
- DSE-44141: Failed to delete deployment in executing DeleteMLServingApp
-
Cloudera AI Inference service did not remove all namespaces if the Cloudera AI Inference service was deleted after an installation failure. This issue is now resolved.
Model training
- DSE-41635: Missing cleanup on deletion of registry
-
The namespace deletion logic is now integrated into the deletion workflows. A new
kubeClientclient was introduced in theModelRegistryProvisionercomponent to enable namespace removal following a Helm release cleanup. The updated code now performs a thorough cleanup process, including the removal of Helm releases, persistent volumes, database records, and Kubernetes namespaces. - DSE-46352: Kserve fails to pull images in air-gapped environments where the Docker registry is not listed in the trusted list
-
Model endpoints no longer fail in air-gapped environments. This issue is now resolved.
User management
- DSE-48606 Team creation and synchronization in Cloudera AI on premises adds default user permission
-
Cloudera AI on premises did not store the selected user permission, which are viewer, collaborator, or administrator, in the database, when creating and synchronizing a team. Instead, it assigned a default permission, granting write access to all team members. The issue is now fixed, the database is updated correctly and Cloudera AI assigns the proper role to the users.
Usability
- DSE-46966: Increase sdx-config-template mount size from 100 MB
-
The size of the
sdx-config-templateemptyDirvolume is increased from 100 MB to 500 MB in the Helm chart, reducing the frequency of mount size issues. - DSE-47738: Cloudera AI temporary backup pods gets stuck into Pending state
-
Cloudera AI temporary backup pods got stuck in the Pending state on specific OpenShift Container Platform versions with Umbra enabled. A Seccomp profile has been added for OpenShift Container Platform setups in which the
restricted-v2profiles prevent the backup pod from starting. This issue is now resolved.
