Make snapshots routine before you need them for a restore.

OpenBao Operator runs backups as transient Jobs that authenticate separately from the main workload, stream Raft snapshots directly to object storage, and record schedule and failure state on the cluster.

Diagram

Backup execution path

A schedule or manual trigger launches a stateless Job. The Job authenticates to OpenBao, streams the Raft snapshot directly, and uploads it to object storage without sending snapshot bytes through the controller.

Decision matrix

Choose the backup auth path

Choose the backup auth path.

Path	Use it when	Operator behavior	Watch for
JWT auth	You can enable selfInit.oidc or already run the JWT auth method on the cluster.	The operator uses a projected ServiceAccount token and can auto-configure the backup auth role when OIDC bootstrap is enabled.	Keep the JWT audience aligned between the controller env vars and the OpenBao role.
Static token	JWT auth is not available yet and you need a compatibility path.	The backup Job reads a long-lived token from a Secret in the cluster namespace.	This is a legacy path. Treat the token as a sensitive credential and rotate it deliberately.

Prerequisites

• Provision a bucket or container in a supported provider:
  • S3 or S3-compatible storage such as MinIO or Ceph
  • Google Cloud Storage
  • Azure Blob Storage
• Grant the backup identity write access to that storage location.
• Allow egress to the storage endpoint. This is required for the Hardened profile.
• Decide whether the backup and restore Jobs will use a Secret, explicit workload identity metadata, or provider-default credentials.

Separate identity surfaces

The main OpenBao Pods and backup Jobs use different ServiceAccounts. Cloud KMS unseal identity on the main workload does not automatically apply to backup or restore Jobs. Check CloudUnsealIdentityReady for the main Pods and BackupConfigurationReady for the generated backup Job identity path.

Configure backup auth and storage

JWT auth (Recommended)

Use JWT auth when you want automatic token rotation and the cleanest separation between the cluster workload and backup jobs.

Automated setup

When spec.selfInit.oidc.enabled is true, the operator automatically configures:

1. the JWT auth method (auth/jwt-operator)
2. OIDC discovery
3. the backup policy (openbao-operator-backup)
4. the backup role (openbao-operator-backup)

No manual OpenBao auth configuration is required.

Configure

Enable OIDC bootstrap for automatic backup auth

yaml

spec:
selfInit:
  enabled: true
  oidc:
    enabled: true

JWT audience

The backup Job uses the audience from OPENBAO_JWT_AUDIENCE (default: openbao-internal). Set the same value in the OpenBao role bound_audiences and pass the env var to the operator through controller.extraEnv and provisioner.extraEnv in Helm.

S3

Configure

Configure S3 or S3-compatible storage

yaml

apiVersion: openbao.org/v1alpha1
kind: OpenBaoCluster
metadata:
name: backup-cluster
spec:
backup:
  schedule: "0 3 * * *"
  target:
    provider: s3
    endpoint: "https://s3.amazonaws.com"
    bucket: "openbao-backups"
    region: "us-east-1"
    pathPrefix: "clusters/backup-cluster"
    usePathStyle: false
    # Optional explicit web identity path:
    # roleArn: "arn:aws:iam::123456789012:role/openbao-backup"
    # Optional provider metadata for the generated ServiceAccount:
    # workloadIdentity:
    #   serviceAccountAnnotations:
    #     eks.amazonaws.com/role-arn: "arn:aws:iam::123456789012:role/openbao-backup"
    credentialsSecretRef:
      name: s3-credentials

S3 credentials

Create a Secret with these keys when you are not using provider-default identity:

• accessKeyId
• secretAccessKey
• sessionToken (optional)
• region (optional)
• caCert (optional)

You can also omit credentialsSecretRef and rely on:

• roleArn for the operator-managed web identity flow
• ambient workload identity or default credentials
• workloadIdentity.serviceAccountAnnotations when your platform integration is driven by ServiceAccount metadata

GCS

Configure

Configure Google Cloud Storage

yaml

apiVersion: openbao.org/v1alpha1
kind: OpenBaoCluster
metadata:
name: backup-cluster
spec:
backup:
  schedule: "0 3 * * *"
  target:
    provider: gcs
    bucket: "openbao-backups"
    pathPrefix: "clusters/backup-cluster"
    gcs:
      project: "my-gcp-project"
    # Optional Workload Identity metadata for the generated ServiceAccount:
    # workloadIdentity:
    #   serviceAccountAnnotations:
    #     iam.gke.io/gcp-service-account: "backup@my-project.iam.gserviceaccount.com"
    credentialsSecretRef:
      name: gcs-credentials

Apply

Create the GCS credentials Secret

bash

kubectl create secret generic gcs-credentials \
--from-file=credentials.json=/path/to/service-account-key.json

Omit credentialsSecretRef when you intentionally rely on Application Default Credentials or Workload Identity instead of a static service-account key.

Azure

Configure

Configure Azure Blob Storage

yaml

apiVersion: openbao.org/v1alpha1
kind: OpenBaoCluster
metadata:
name: backup-cluster
spec:
backup:
  schedule: "0 3 * * *"
  target:
    provider: azure
    bucket: "openbao-backups"
    pathPrefix: "clusters/backup-cluster"
    azure:
      storageAccount: "mystorageaccount"
      container: "openbao-backups"
    # Optional workload identity metadata:
    # workloadIdentity:
    #   serviceAccountAnnotations:
    #     azure.workload.identity/client-id: "00000000-0000-0000-0000-000000000000"
    #   podLabels:
    #     azure.workload.identity/use: "true"
    credentialsSecretRef:
      name: azure-credentials

Azure credentials

Create a Secret with one of the following:

• accountKey
• connectionString

For managed identity or Azure Workload Identity, omit credentialsSecretRef. If your cluster integration requires Kubernetes metadata, use:

• target.workloadIdentity.serviceAccountAnnotations
• target.workloadIdentity.podLabels

Static token (Legacy)

Use this path only when JWT auth is not available. The backup Job reads a long-lived OpenBao token from a Secret.

Same-namespace requirement

All referenced Secrets must exist in the same namespace as the OpenBaoCluster. Cross-namespace references are not allowed.

Apply

Create the backup token Secret

bash

kubectl create secret generic backup-token \
--from-literal=token=hvs.yourtoken...

S3

Configure

Configure S3 backup with a static token

yaml

apiVersion: openbao.org/v1alpha1
kind: OpenBaoCluster
metadata:
name: backup-cluster
spec:
backup:
  schedule: "0 3 * * *"
  tokenSecretRef:
    name: backup-token
  target:
    provider: s3
    endpoint: "https://s3.amazonaws.com"
    bucket: "openbao-backups"
    region: "us-east-1"
    credentialsSecretRef:
      name: s3-credentials

GCS

Configure

Configure GCS backup with a static token

yaml

apiVersion: openbao.org/v1alpha1
kind: OpenBaoCluster
metadata:
name: backup-cluster
spec:
backup:
  schedule: "0 3 * * *"
  tokenSecretRef:
    name: backup-token
  target:
    provider: gcs
    bucket: "openbao-backups"
    gcs:
      project: "my-gcp-project"
    credentialsSecretRef:
      name: gcs-credentials

Azure

Configure

Configure Azure backup with a static token

yaml

apiVersion: openbao.org/v1alpha1
kind: OpenBaoCluster
metadata:
name: backup-cluster
spec:
backup:
  schedule: "0 3 * * *"
  tokenSecretRef:
    name: backup-token
  target:
    provider: azure
    bucket: "openbao-backups"
    azure:
      storageAccount: "mystorageaccount"
    credentialsSecretRef:
      name: azure-credentials

Advanced backup settings

Provider-specific options

S3

Reference table

S3-specific options

S3-specific options.

Option	Default	What it changes
region	us-east-1	Sets the AWS region or any placeholder value needed by an S3-compatible implementation.
usePathStyle	false	Switch to path-style addressing for MinIO and some S3-compatible endpoints.
roleArn	none	Enables the explicit AWS web identity path managed by the operator.
pathPrefix	cluster-scoped default	Controls the object prefix used for backup keys so clusters stay separated inside a shared bucket.

Configure

Set S3 provider-specific options

yaml

spec:
backup:
  target:
    provider: s3
    region: "eu-west-1"
    usePathStyle: true
    roleArn: "arn:aws:iam::123456789012:role/backup-role"
    pathPrefix: "clusters/prod-a"

GCS

Reference table

GCS-specific options

GCS-specific options.

Option	What it changes
project	Pins the GCP project when credentials or ADC do not already provide it.
endpoint	Overrides the storage endpoint for emulators such as fake-gcs-server.

Configure

Set GCS provider-specific options

yaml

spec:
backup:
  target:
    provider: gcs
    endpoint: "http://fake-gcs-server:4443"
    gcs:
      project: "my-gcp-project"

Azure

Reference table

Azure-specific options

Azure-specific options.

Option	What it changes
storageAccount	Selects the Azure storage account. This is required when provider: azure is used.
container	Overrides the container name when it should differ from bucket.
endpoint	Overrides the blob endpoint for testing tools such as Azurite.

Configure

Set Azure provider-specific options

yaml

spec:
backup:
  target:
    provider: azure
    endpoint: "http://127.0.0.1:10000"
    azure:
      storageAccount: "mystorageaccount"
      container: "backups"

Workload identity metadata

Use target.workloadIdentity when your cloud identity integration depends on ServiceAccount annotations or pod labels on the generated backup and restore workloads.

Configure

Attach identity metadata to backup and restore workloads

yaml

spec:
backup:
  target:
    workloadIdentity:
      serviceAccountAnnotations:
        iam.gke.io/gcp-service-account: "backup@my-project.iam.gserviceaccount.com"
        azure.workload.identity/client-id: "00000000-0000-0000-0000-000000000000"
      podLabels:
        azure.workload.identity/use: "true"

serviceAccountAnnotations are applied to the generated backup and restore ServiceAccounts. podLabels are applied to backup and restore Job pods without replacing operator-managed labels.

Emulator support

GCS and Azure support custom endpoints for local testing with fake-gcs-server and Azurite. When those endpoints use self-signed certificates, include the CA certificate in the credentials Secret.

Retention policy

Retention cleanup runs after a successful backup and works across S3, GCS, and Azure.

Configure

Keep a limited number of recent snapshots

yaml

spec:
backup:
  retention:
    maxCount: 7
    maxAge: "168h"

Performance tuning

Reference table

Multipart upload tuning

Multipart upload tuning.

Parameter	Default	When to change it
partSize	10MB	Increase it for high-bandwidth links and large datasets when larger chunks reduce upload overhead.
concurrency	3	Increase for throughput, or reduce it when memory pressure or object-store throttling becomes the limiting factor.

Configure

Tune upload chunking and parallelism

yaml

spec:
backup:
  target:
    partSize: 20971520
    concurrency: 5

Pre-upgrade snapshots

Take a snapshot immediately before the rolling update or blue-green cutover begins.

Configure

Require a snapshot before upgrades start

yaml

spec:
upgrade:
  preUpgradeSnapshot: true
backup:
  target: { ... }

Upgrade safety

preUpgradeSnapshot: true only works when spec.backup.target is already configured. Confirm backup status before you start the upgrade rather than assuming the pre-upgrade snapshot can be taken on demand.

Verify and operate

Inspect

Inspect backup status on the cluster

bash

kubectl get openbaocluster my-cluster \
-o jsonpath='{.status.backup}'

Check lastSuccessfulBackup, nextScheduledBackup, and failure counters before you rely on the policy as a recovery control.

Apply

Trigger a manual backup from the generated CronJob

bash

kubectl create job --from=cronjob/my-cluster-backup manual-backup-1

Use a manual run to prove the full path: identity, cluster auth, storage reachability, and object naming before the first production upgrade.

Official OpenBao background

• OpenBao Backups
• Operator Raft Command
• JWT/OIDC Auth Method

Next operating steps

Restore from backupUse the restore guide when you need to consume one of the snapshots this page configures.Plan upgradesBackups should be validated before you depend on pre-upgrade snapshots and cutover safety.Open the production checklistUse the checklist to confirm backups, restore readiness, and day 2 controls before calling the cluster production-ready.