Choose the operating posture before you tune anything else.

spec.profile is the top-level decision that shapes bootstrap, unseal, TLS, image verification, and failure tolerance. Pick the posture you plan to keep operating, then let the rest of the cluster baseline follow from that choice.

Decision matrix

Choose the profile deliberately

Choose the profile deliberately.

Profile	Use it when	What it assumes	Avoid it when
Hardened	The cluster is intended to become a real production service.	External unseal, self-initialization, verified TLS, and supply-chain guardrails are part of the normal path.	You cannot meet the external trust, identity, or networking requirements yet.
Development	You need a fast local or evaluation path and accept weaker security defaults temporarily.	Bootstrap material may live in Kubernetes Secrets, TLS can be operator-managed, and verification can be relaxed.	You are trying to define the baseline that will survive into production.

Diagram

How the profile shapes the baseline

The profile is not cosmetic. It determines whether the cluster can rely on operator-generated trust material and stored bootstrap credentials or whether those paths are explicitly disallowed.

What actually changes with the profile

Reference table

Profile effects

Profile effects.

Surface	Development	Hardened
Bootstrap credential handling	Manual init is allowed and can leave a root token in a Secret when self-init is disabled.	Self-initialization is the supported path and root-token persistence is not the normal operating model.
Unseal	Static unseal in a Secret is allowed for fast evaluation.	Use an external trust source such as cloud KMS or transit. Treat static unseal as a non-production exception.
TLS	Operator-managed TLS is acceptable for development and internal evaluation.	Use External or ACME; the certificate authority should not be operator-generated in production.
Image verification	Can be introduced gradually and warning-only rollouts are possible.	Verification is expected, warning-only behavior is not the production posture, and official-image defaults still verify when trust material is omitted.
Networking and jobs	You can tolerate more permissive local defaults while standing up the cluster.	Backup and other lifecycle paths should assume explicit egress and identity wiring before go-live.

Representative starting points

Hardened

Configure

Start from the supported production baseline

yaml

apiVersion: openbao.org/v1alpha1
kind: OpenBaoCluster
metadata:
name: prod-cluster
spec:
profile: Hardened
replicas: 3
version: "2.4.4"
selfInit:
  enabled: true
tls:
  enabled: true
  mode: External
unseal:
  type: awskms
  awskms:
    region: us-east-1
    kmsKeyID: alias/openbao-unseal
imageVerification:
  enabled: true
  failurePolicy: Block
operatorImageVerification:
  enabled: true
  failurePolicy: Block

Hardened is the supported production path. It assumes an external trust source for unseal, non-operator-managed TLS, and a self-initializing bootstrap flow.

Development

Configure

Use the lightest safe evaluation baseline

yaml

apiVersion: openbao.org/v1alpha1
kind: OpenBaoCluster
metadata:
name: dev-cluster
spec:
profile: Development
version: "2.4.4"
replicas: 1
tls:
  enabled: true
unseal:
  type: static

Development is for local testing, proof-of-concept work, and environments where you explicitly accept stored bootstrap material and weaker trust boundaries.

Do not let development defaults drift into production

The dangerous part of the Development profile is not that it exists; it is that it feels easy to keep. If the cluster matters, switch to Hardened before other systems begin to depend on it.

Choose the unseal root of trust

Decision matrix

Unseal options by posture

Unseal options by posture.

Path	Use it when	Why it fits or does not fit
Cloud KMS	You run in AWS, GCP, Azure, or another managed platform with a usable external key service.	This is usually the cleanest Hardened path because the root of trust stays outside Kubernetes.
Transit	You already run a central OpenBao cluster or equivalent external trust service.	This works well for multi-cluster or hybrid environments where central trust management is intentional.
PKCS#11 or KMIP	You need HSM-backed or enterprise key-management integration.	Valid for production, but usually more specialized and operationally heavier than cloud KMS or transit.
Static Secret	You need a local development path and understand the blast radius.	This is convenient but keeps decryption material inside the same cluster state you are trying to protect.

AWS KMS

Configure

Use AWS KMS for Hardened unseal

yaml

spec:
profile: Hardened
serviceAccount:
  annotations:
    eks.amazonaws.com/role-arn: "arn:aws:iam::123456789012:role/openbao-awskms"
unseal:
  type: awskms
  awskms:
    kmsKeyID: "arn:aws:kms:us-east-1:123456789012:key/..."
    region: "us-east-1"

GCP KMS

Configure

Use GCP Cloud KMS for Hardened unseal

yaml

spec:
profile: Hardened
serviceAccount:
  annotations:
    iam.gke.io/gcp-service-account: "openbao@my-project.iam.gserviceaccount.com"
unseal:
  type: gcpckms
  gcpCloudKMS:
    project: "my-project"
    region: "us-central1"
    keyRing: "openbao-ring"
    cryptoKey: "openbao-key"

Azure Key Vault

Configure

Use Azure Key Vault for Hardened unseal

yaml

spec:
profile: Hardened
serviceAccount:
  annotations:
    azure.workload.identity/client-id: "87654321-4321-4321-4321-210987654321"
podMetadata:
  labels:
    azure.workload.identity/use: "true"
unseal:
  type: azurekeyvault
  azureKeyVault:
    vaultName: "my-vault"
    keyName: "openbao-key"

Transit

Configure

Use a central OpenBao transit key for unseal

yaml

spec:
profile: Hardened
unseal:
  type: transit
  credentialsSecretRef:
    name: transit-unseal-creds
  transit:
    address: "https://central-openbao.example.com"
    keyName: "tenant-1-key"
    mountPath: "transit"

The referenced Secret should hold the transit token and any optional CA or client-certificate material required by that upstream cluster.

Optional runtime hardening

Configure

Enable AppArmor when the nodes support it

yaml

spec:
workloadHardening:
  appArmorEnabled: true

AppArmor is opt-in because support depends on the underlying node OS and cluster runtime. Pair this with the broader workload baseline in Pod and runtime security.

Continue cluster baseline

Self-initializationConfigure the bootstrap requests and operator OIDC flow that follow from the profile choice.External accessChoose the TLS and exposure pattern that matches the baseline you just picked.Workload protectionsReview the runtime and supply-chain controls expected behind the Hardened posture.