2.4 Deploying the Milvus Vector Database

The second major component of our RAG system is the vector database. For this workshop, we are using Milvus, a powerful open-source database specifically designed for storing and searching vector embeddings at scale.

Our data pipeline will populate this database with vectors generated from the ServiceNow incident tickets. You will now deploy the Milvus application into your project namespace using the same GitOps approach as the mock API.

Applying the ArgoCD Application Manifest for the Milvus Database

You will create the ArgoCD Application resource using the Import YAML feature built into the OpenShift web console.

In the OpenShift Web Console, click the + (Import YAML) button in the top-right toolbar.

Paste the following manifest into the editor.

ArgoCD Application Manifest

apiVersion: argoproj.io/v1alpha1
kind: Application
metadata:
  name: 'userX-milvus' (1)
  namespace: openshift-gitops
spec:
  project: default
  source:
    repoURL: 'https://github.com/cnuland/hello-chris-rag-pipeline.git'
    path: services/milvus (2)
    targetRevision: main
    kustomize: {}
  destination:
    server: https://kubernetes.default.svc
    namespace: 'userX' (3)
  syncPolicy:
    automated:
      prune: true
      selfHeal: true
    syncOptions:
      - CreateNamespace=true
      - ServerSideApply=true

1	A unique name scoped to your account, preventing resource conflicts in the shared cluster.
2	The Kustomize overlay directory within the repository that defines the Milvus deployment and its etcd dependency.
3	The target namespace where ArgoCD will deploy the Milvus workloads.

Click Create.

You’ll be taken to the newly created Application, which will be in the process of syncing.

ArgoCD will immediately begin reconciling the application. Initial synchronization, including pulling the Milvus and etcd container images, typically takes 2–3 minutes.

Verifying the Deployment

Once ArgoCD has applied the manifest, confirm the pods are healthy from the OpenShift Console.

In the OpenShift Console, select the userX project from the project dropdown at the top, then navigate to Workloads → Pods in the left-hand menu.

Wait until both the vectordb-etcd-0 and vectordb-milvus-standalone-… pods report a Running status with 1/1 containers ready.

NAME                                         READY   STATUS    RESTARTS   AGE
mock-servicenow-api-5f6f8b9d6c-xxxxx        1/1     Running   0          10m
vectordb-etcd-0                              1/1     Running   0          5m
vectordb-milvus-standalone-8566db697-xxxxx   1/1     Running   0          4m

If the pods show ContainerCreating or Pending, the node is still pulling the container images. Refresh the page after 30 seconds.

With Milvus running, we now have our data source (the API) and our data destination (the vector database) ready. The next step is to import the pipeline that connects them.

Summary

Deployed Milvus (standalone mode) and its etcd dependency into your namespace via a second ArgoCD Application
Verified both the vectordb-milvus-standalone and vectordb-etcd-0 pods are running and ready
With the mock API and Milvus both live, the data source and vector store are in place — the pipeline will connect them next