Bootstrapping cluster from backup

[jakovistuk]

Hi, I have a problem to spin up a new cluster from backup..

This Cluster is ok, and sends WAL files to Minio.

apiVersion: postgresql.cnpg.io/v1
kind: Cluster
metadata:
  name: my-cluster
spec:
  resources:
    limits:
      cpu: 100m
      memory: 200Mi
    requests:
      cpu: 100m
      memory: 100Mi
  instances: 2
  primaryUpdateStrategy: unsupervised
  storage:
    size: 2Gi
  backup:
    barmanObjectStore:
      destinationPath: 's3://postgres01'
      endpointURL: 'https://minio-api-lab:9000'
      endpointCA:
        name: "certname"
        key: "ca.crt"
      s3Credentials:
        accessKeyId:
          key: 'MINIO_ACCESS_KEY'
          name: 'minio-creds'
        secretAccessKey:
          key: 'MINIO_SECRET_KEY'
          name: 'minio-creds'

I try to spin up recovery cluster, with yaml, but always get ,"error":"no target backup found".

apiVersion: postgresql.cnpg.io/v1
kind: Cluster
metadata:
  name: recover-my-cluster
spec:
  resources:
    limits:
      cpu: 100m
      memory: 200Mi
    requests:
      cpu: 100m
      memory: 100Mi
  instances: 2
  primaryUpdateStrategy: unsupervised
  storage:
    size: 2Gi
  bootstrap:
    recovery:
      source: my-cluster
  externalClusters:
    - name: my-cluster
      barmanObjectStore:
        destinationPath: 's3://postgres01'
        endpointURL: 'https://minio-api-lab:9000'
        endpointCA:
          name: "certname"
          key: "ca.crt"
        s3Credentials:
          accessKeyId:
            key: 'MINIO_ACCESS_KEY'
            name: 'minio-creds'
          secretAccessKey:
            key: 'MINIO_SECRET_KEY'
            name: 'minio-creds'
      wal:
         maxParallel: 6

What I am doing wrong?

[gbartolini]

Did you actually ever take a full backup of the original cluster, by for example manually creating a backup object or by creating a ScheduledBackup object?

[lukasz18dg]

In my situation, when I had the same problem: no.
That means: WAL files are not enough to recover the database and create a functional database from scratch?

[Lennart01]

nope. You need a base backup. The WAL is only a log of operations performed on your database.
Maybe we should add a disclaimer to the docs 🤔.
Take a look here on how to create a base backup / schedule a backup.
When you are bootstrapping a new cluster the base backup is loaded and then the WAL is replayed on the backup.
Keep this in mind when choosing your backup frequency. Depending on your load this can impact the recovery time.

[lukasz18dg]

Thanks for answer!
I think, adding something in documentation may be worth, because I omitted that information I need backup and WAL files to recovery my database.

[gbartolini]

Thanks for the feedback. I hope that the new docs that we have prepared for volume snapshot in release 1.21.0 clarify this. You can get a preview here: https://github.com/cloudnative-pg/cloudnative-pg/blob/main/docs/src/backup.md

This is indeed 'basic' Postgres information, but it is important for us to understand that, although we should point to Postgres, we should provide at least some critical info.