Cloudera Docs

Troubleshooting CDSW to CML migration

You need to know how to retry migration if your migration from Cloudera Data Science Workbench (CDSW) 1.10.0 to Cloudera Machine Learning (CML) fails. Also, you learn how to find logs for debugging migration and other issues.

Failed Migration

Problem: In the event your migration fails, the Migration failed indicator appears.

Solution: In Workspaces, click Options , and select Retry CDSW Migration:

Viewing logs

Problem: In the event your migration fails, and the retry also fails, you need to get information about the failure.

Solution: In Workspaces, click Options , and select View Event Logs.

Troubleshooting failure after reboot in Single Node mode

After reboot of the host in a single node Embedded Container Service (ECS) cluster, the cdpcli pod in the cdp namespace might end up in an error state. The cdpcli pod was part of a Kubernetes job for CDP Private Cloud installation. You can delete the pod and reboot.

If the cdp-embedded-db pod fails to start after the reboot, perform the following steps:
  1. Check the logs of the pod to determine the cause.

    The volume might be failing to attach to the host.

  2. Find the name of the Private Cloud (pvc) used by the cdp-embedded-db-0 pod.
  3. On the host, check the volume.meta file under /<longhorn-storage>/replicas/pvc-xxxxxx-xxxxxx/ .
  4. Verify that the Size field matches the volume size requested by the pod.
  5. Correct the field to possibly resolve the issue.

Problem when migrating parcels to a new CML workspace

If parcels are located in a custom directory specified by parcel_repo_home in Cloudera Manager, instead of the default location (/opt/cloudera/parcels), then the CDSW to CML migration script will not find them.

Workaround:
  1. Check that ECS parcels are located at /opt/cloudera/parcels and not at a custom location.
  2. If not, then copy the existing parcels to the default location: 
    cp -rp <custom-ecs-parcel-path>/ECS* /opt/cloudera/parcels/