[Optional] Enabling multi region active deployment setup

Multi region active deployment helps you when you have geographically distributed large number of setups. The Global Load Balancer in that setup helps your conversation session to continue even your regional cluster is not available.

For multi region active deployment setup, you need to have:
  • Two clusters with identical configuration of watsonx™ Orchestrate.
  • Identical CPD version, installation parameters, IFM models, AI-Gateway models, and their configuration in both the clusters.
  • 5.3 hot fix-1 is applied (Multi region active deployment capability is introduced in this hot fix)
  • The same multi region active deployment seed must be applied in both the clusters.
  • Each service instance of watsonx Orchestrate represents a watsonx Orchestrate tenant. You can create as many instances or tenants as needed, but each instance must be created in both the clusters with the same Instance-Name.
If you are upgrading from a previous version to 5.3, existing instances that were created in previous version do not have multi region active deployment capability. The instances that are created after you enabled multi region active deployment have multi region active deployment capability. The watsonx Orchestrate CR patch does not retroactively introduce this capability in existing instances.

If the instances were created before the seed was set in both clusters, the Assistants that are created in that instance cannot have the same UUIDs. Create the tenants after you set the seed for multi region active deployment to work for the Assistant.

If you want to use multi region active deployment AI Gateway, you need to enable multi region active deployment before you register models in AI Gateway. This applies to adding models through ADK and API. API-based URL format with:
  • Instance-ID: https://{cpd_cluster_host}/orchestrate/{namespace}/instances/{instanceid}/
  • Instance-Name: https://{cpd_cluster_host}/orchestrate/{namespace}/instances/{instanceName}/
This base URL with Instance-ID is displayed in the watsonx Orchestrate profile page in UI. The watsonx Orchestrate API documentation shows API URLs with Instance-ID. Swap Instance-ID with Instance-Name when your multi region active deployment with load balancer.

Enable multi region active deployment capability

  1. Apply the same seed along with the enabled flag in both the clusters. [watsonx Orchestrate CR patch]. To generate the seed, use uuidgen UNIX command, command to patch watsonx Orchestrate CR with seed and flag:
    oc patch wo wo \
  --namespace="${PROJECT_CPD_INST_OPERANDS}" \
  --type=merge \
  --patch="{
    \"spec\": {
      \"activeActive\": {
        \"enabled\": true,
        \"activeActiveSeed\": \"${ACTIVE_ACTIVE_SEED}\"
      }
    }
  }"
  2. Create instances with the same Instance-Name in both the clusters.
  3. If you are using AI Gateway, register the same models with the same configuration in both the instances across the multi region active deployment clusters by either ADK or API.