New Database¶

The postgres component bootstraps every cluster from a Barman backup by default. For a brand-new app, nothing exists at s3://postgresql/${APP}/ yet, so that would fail with "no target backup found". The components.postgres/cnpg: init label switches the cluster to a plain initdb for its first deploy.

1. Deploy with the init label¶

apiVersion: kustomize.toolkit.fluxcd.io/v1
kind: Kustomization
metadata:
  name: &app myapp
  labels:
    components.postgres/cnpg: init # (1)!
spec:
  components:
    - ../../../../components/postgres
  healthCheckExprs:
    - apiVersion: postgresql.cnpg.io/v1
      kind: Cluster
      failed: status.conditions.filter(e, e.type == 'Ready').all(e, e.status == 'False')
      current: status.conditions.filter(e, e.type == 'Ready').all(e, e.status == 'True')
  interval: 1h
  path: ./kubernetes/apps/<namespace>/myapp/app
  postBuild:
    substitute:
      APP: *app
  prune: true
  sourceRef:
    kind: GitRepository
    name: flux-system
    namespace: flux-system
  targetNamespace: <namespace>

1. Only for the first deploy. Remove it once a backup exists (step 3).

A patch in cluster-apps matches the label. It strips spec.bootstrap.recovery and spec.externalClusters from the Cluster, and replaces bootstrap with an initdb that creates a database and owner role named ${POSTGRES_USERNAME:=${APP}}. CNPG generates the role's password into the ${APP}-postgres-app Secret as usual.

2. Get a first backup¶

Wait for the nightly ScheduledBackup, or force one:

just k8s db-backup <namespace> myapp
kubectl -n <namespace> get backup

3. Remove the label¶

Once a backup has completed, remove components.postgres/cnpg: init and merge. From then on, a cluster rebuild follows the default recovery path.