Onboard a namespace

OpenBaoTenant authorizes the controller to manage an existing namespace and applies tenant guardrails. It does not create the namespace or an OpenBaoCluster.

Skip this page only when the operator uses the verified single-tenant mode. Review tenant boundaries when you need the authority model behind this handoff.

Choose the onboarding owner

The target namespace must already exist. spec.targetNamespace is required and admission treats it as immutable.

Onboard through self-service

1. Create the target namespace if it does not exist.

   apply

   Create the evaluation namespace

   Copy

   kubectl create namespace openbao-demo

2. Save this request as tenant.yaml.

   configure

   Declare a self-service tenant

   Copy

   apiVersion: openbao.org/v1alpha1
      kind: OpenBaoTenant
      metadata:
        name: openbao-demo
        namespace: openbao-demo
      spec:
        targetNamespace: openbao-demo

3. Apply the request.

   apply

   Onboard the namespace

   Copy

   kubectl apply -f tenant.yaml

4. Verify that provisioning completed.

   verify

   Verify the tenant handoff

   Copy

   kubectl -n openbao-demo get openbaotenant openbao-demo \
        -o jsonpath='{.status.provisioned}{"\t"}{.status.conditions[?(@.type=="Provisioned")].status}{"\n"}'
      kubectl -n openbao-demo get rolebinding \
        openbao-operator-tenant-rolebinding

   The first command must print true and True. The condition carries the current generation, reason, and message; the RoleBinding is the concrete handoff that allows cluster reconciliation to continue.

Onboard centrally

Create the request in the actual rendered operator namespace when a platform administrator manages onboarding for a team:

apiVersion: openbao.org/v1alpha1
kind: OpenBaoTenant
metadata:
  name: team-a
  namespace: openbao-operator-system
spec:
  targetNamespace: team-a-prod
  # Optional platform-managed values:
  # quota:
  # limitRange:

The trusted namespace is the rendered operator namespace, not a fixed project-wide value. If the release uses another namespace, change metadata.namespace accordingly.

Understand the applied resources

Onboarding manages:

• the namespace-scoped controller Role and openbao-operator-tenant-rolebinding;
• Secret allowlist Roles used by operator-managed workloads and Jobs;
• a ResourceQuota and LimitRange;
• restricted Pod Security enforce, audit, and warn labels by default.

It does not create workload NetworkPolicies. Those are reconciled after an OpenBaoCluster is accepted. The explicit OpenBaoTenant request also means the Provisioner does not need to discover tenant namespaces through a cluster-wide namespace label watch.

The default quota allows 50 Pods, 20 requested CPU, 64 GiB requested memory, 40 CPU limit, and 128 GiB memory limit. The default per-container LimitRange is:

Set tenancy.namespacePodSecurityLabels.mode=external when another platform controller owns the Pod Security labels. This delegates only label mutation; the operator still manages tenant RBAC, Secret allowlists, quota, and LimitRange.

Apply tenant and cluster together

GitOps can apply OpenBaoTenant and OpenBaoCluster in one synchronization. The cluster controller pauses until openbao-operator-tenant-rolebinding exists, then continues without treating the missing handoff as a permanent failure. Watching the tenant status first still gives the clearest rollout signal.

Troubleshoot onboarding

Remove or retarget a tenant

You cannot change spec.targetNamespace in place. Before deleting an OpenBaoTenant, delete every OpenBaoCluster in the target namespace; the finalizer waits until they are gone.

When no other active, authorized OpenBaoTenant claim targets the namespace, cleanup removes tenant RBAC and the operator-managed ResourceQuota and LimitRange before removing the finalizer. Another active claim preserves those shared resources while the deleting claim finalizes. Claim discovery and cleanup fail closed: an error keeps the finalizer in place.

Pod Security labels remain unchanged because the operator does not retain the namespace’s earlier label state. Review them explicitly when decommissioning or retargeting the namespace.

Continue with cluster creation.