Managing multiple clusters for watsonx.governance

You can distribute governance activities across multiple servers and sync between the primary governance server and remote servers. Use multiple servers to isolate production assets on one server and control access to them for greater security.

Who needs to complete this task?

You must be a instance administrator to do the tasks related to designating multiple clusters. Instance admins can configure watsonx.governance™ clusters that they administer.

When do you need to complete this task?
Complete this task when your governance assets are mature enough to merit distribution of their governance activities across multiple servers for segregated monitoring.

An administrator for watsonx.governance who collects the metadata for model activity might want to put physical models on one server and govern them on another for security or accessibility purposes. For example, if an administrator wants to track but not to manage the model's general lifecycle, such as evaluating or deploying them.

In this example, the administrator can set up a primary governance cluster used for viewing model information. They then set up one or multiple remote clusters where models are uploaded, evaluated and deployed. The watsonx.governance remote clusters are configured to send data to the primary governance cluster automatically whenever there is model activity. The metadata for the models on primary governance cluster are synced when it receives the updated data. The physical models live only on remote clusters for their lifecycle.

Primary governance and remote clusters

You can work with three types of clusters in watsonx.governance:

Stand-alone cluster
Not configured to connect to any clusters. All clusters are by default stand-alone clusters.
Primary governance cluster
Accepts incoming requests regarding data sync from remote clusters. Each watsonx.governance should have one primary governance cluster.
Remote cluster
Sends outgoing requests to primary governance clusters to sync data.

Setting up multiple clusters

To establish multiple clusters, you start with a stand-alone cluster that you designate as the primary governance cluster. You can then configure one or more remote clusters that will monitor and relay governance activity data about models (for example model deployments) back to the primary governance cluster.
Note: Ensure that the primary governance cluster and all remote clusters are at the same version and patch level of IBM® Software Hub.

To set up the clusters, do these tasks:

  1. Designating a primary governance cluster for watsonx.governance
  2. Configuring a remote cluster using one of these options: