Manually Managing Nodes and Disks

HPE Ezmeral Runtime Enterprise versions 5.2 and later automate many of the processes described in this article. See:

This information is presented for educational, maintenance, and debugging by users with advanced knowledge of HPE Ezmeral Data Fabric.

NOTE
In this article, the term tenant refers to Data Fabric tenants (formerly '"MapR tenants") and not to Kubernetes tenants unless explicitly noted otherwise on a case-by-case basis.

The bootstrap utility validates the nodes and disks for deploying Data Fabric clusters and tenants, and applies labels and annotations on the nodes for scheduling pods.

Labeling Nodes

You may optionally modify how detected nodes are used by changing the value of these labels and annotations before deploying a Data Fabric clusters and/or tenants by executing the following command:

kubectl describe node <nodename>

The bootstrap utility automatically uses all nodes in the Kubernetes cluster that do not have a mapr.com/usernode label unless you update the label for a node before using the bootstrap utility, as follows:

  • To update the label to not install HPE Ezmeral Data Fabric on a node, execute the command:

     kubectl label node --overwrite <node_name> "mapr.com/usenode=false"
  • To update the label to install HPE Ezmeral Data Fabric on a node (default option), execute this command:

     kubectl label node --overwrite <node_name> "mapr.com/usenode=true"

See Node Labels for additional information about this and other labels.

Labeling Disks

You must manually apply disk labels to every node on which you are installing HPE Ezmeral Data Fabric before running the bootstrap utility. The bootstrap script does provide a fake node labeller that can perform the labeling for you, but this feature may not label the disks as desired.