Decommissioning/Deleting a Kubernetes Host

NOTE
This article uses the term host to denote both Kubernetes hosts and Data Fabric cluster nodes, except if noted.

This article describes the following:

Decommissioning a Kubernetes Worker Host

Clicking the Decommission icon (barred circle) in the Actions column of the Kubernetes Host(s) Status table in the Kubernetes Hosts Installation screen (see The Kubernetes Hosts Installation Screen) removes the persistent storage service and vacates any data volumes to other nodes so that the host can be removed. You may then delete the Kubernetes host, as described in Deleting a Kubernetes Worker Host, below.

NOTE
You can only decommission one Kubernetes Worker host at a time, and the deployment cannot have any fewer than four (4) commissioned Kubernetes Worker hosts.

Deleting a Kubernetes Worker Host

Deleting a Kubernetes host completely removes it from the deployment. To delete one or more Kubernetes Worker hosts:

  1. Remove the affected Kubernetes Worker hosts from any existing Kubernetes clusters.
  2. Access the Kubernetes Hosts Installation screen (see The Kubernetes Hosts Installation Screen).
  3. Decommission the Kubernetes Worker hosts by clicking the Decommission icon (broken line) for each Kubernetes Worker host being deleted.
  4. In the Kubernetes Host(s) Status table, either:
    • Remove a single Kubernetes Worker host by clicking the Delete icon (trash can) for the host you want to remove.
    • Remove multiple Kubernetes Worker hosts by selecting the affected hosts and then clicking the Delete button above the table.
  5. The Status of the affected Kubernetes Worker hosts shows deleting.

The selected Worker hosts are removed.

Monitor the deletion process and address any reported problem. The log is located on the affected host, in the /tmp/worker_setup_<timestamp> folder.

In the unlikely event that Kubernetes Worker deletion fails, you can delete the Kubernetes Worker host manually by executing the following command on that Worker host:

  • If the Kubernetes Worker host was installed without using the agent:

    /opt/bluedata/bundles/<common-epic.bin> -ef

  • If the agent was used when installing the Kubernetes Worker host:

    /opt/bluedata/bundles/<common-epic.bin> -ef --onworker --node-type worker --worker <worker-ip>

After deletion has completed, execute the following commands to verify that HPE Ezmeral Runtime Enterprise has been removed from that host:

  • bdconfig -sysinfo

    The system should return command not found.

  • rpm -qa | grep -E "bluedata|hpe"

    The system should return an empty response.

If you plan to re-use this host as a new Worker host for either Big Data/AI/ML or Kubernetes, then you will need to ensure that Docker storage is cleaned-up. Follow these steps:

  1. Make sure that /var/lib/docker is empty.
  2. Make sure that /etc/sysconfig/docker-storage has the value DOCKER_STORAGE_OPTIONS=

To reinstall the host immediately, you will need to begin again from Kubernetes Host: Add the Hosts. Otherwise, you must return to Kubernetes Worker Installation Overview and restart the installation process from the beginning.