backup method changed in the output of kubectl get backup

[alexeyshishkin-edb]

Hi, Team !

I have my CNPG 1.26.1 cluster with Barman Native configured.

tee $cluster_name.yaml << EOF
apiVersion: postgresql.cnpg.io/v1
kind: Cluster
metadata:
  name: $cluster_name
  namespace: $oname-system
spec:
  instances: 3
  imageName: $cnpg_image_name
  storage:
    size: 1Gi
  backup:
    retentionPolicy: "1d"
    barmanObjectStore:
      destinationPath: s3://cluster-backups/
      endpointURL: http://minio-service:9000
      s3Credentials:
        accessKeyId:
          name: minio-creds
          key: ACCESS_KEY_ID
        secretAccessKey:
          name: minio-creds
          key: ACCESS_SECRET_KEY
      wal:
        compression: gzip
EOF

tee barman-backup.yaml << EOF
apiVersion: postgresql.cnpg.io/v1
kind: Backup
metadata:
  name: barman-backup
  namespace: $oname-system
spec:
  cluster:
    name: $cluster_name
  target: prefer-standby
EOF

I have two backups:

• barman-backup is done by kubectl apply -f barman-backup.yaml
• cluster-barman-cnpg-20251016101137 is done by kubectl cnpg -n $oname-system backup $cluster_name -m barmanObjectStore

[alexey@wsl2 ~]$ kubectl get -n $oname-system backup
NAME                                 AGE   CLUSTER               METHOD              PHASE       ERROR
barman-backup                        57s   cluster-barman-cnpg   barmanObjectStore   completed
cluster-barman-cnpg-20251016101137   6s    cluster-barman-cnpg   barmanObjectStore   completed

Now I'm migrating to Barman Plugin 0.7.0:

tee $cluster_name.yaml << EOF
apiVersion: postgresql.cnpg.io/v1
kind: Cluster
metadata:
  name: $cluster_name
  namespace: $oname-system
spec:
  instances: 3
  imagePullPolicy: Always
  imageName: $cnpg_image_name
  storage:
    size: 1Gi
  plugins:
  - name: barman-cloud.cloudnative-pg.io
    isWALArchiver: true
    parameters:
      barmanObjectName: minio-store
EOF

tee barman-backup.yaml << EOF
apiVersion: postgresql.cnpg.io/v1
kind: Backup
metadata:
  name: barman-backup
  namespace: $oname-system
spec:
  cluster:
    name: $cluster_name
  method: plugin
  pluginConfiguration:
    name: barman-cloud.cloudnative-pg.io
  target: prefer-standby
EOF

After launching kubectl apply -f barman-backup.yaml I have the following:

[alexey@wsl2 ~]$ kubectl get -n $oname-system backup
NAME                                 AGE   CLUSTER               METHOD              PHASE       ERROR
barman-backup                        14m   cluster-barman-cnpg   plugin              completed
cluster-barman-cnpg-20251016101137   13m   cluster-barman-cnpg   barmanObjectStore   completed

So, somehow the METHOD for my older barman-backup backup has been changed from barmanObjectStore to plugin:

[alexey@wsl2 ~]$ kubectl get -n $oname-system backup
NAME                                 AGE   CLUSTER               METHOD              PHASE       ERROR
barman-backup                        57s   cluster-barman-cnpg   barmanObjectStore   completed
cluster-barman-cnpg-20251016101137   6s    cluster-barman-cnpg   barmanObjectStore   completed

Is it by design ? If not, can it be fixed ?

[dosubot]

This change is expected if you re-apply a Backup manifest with the same name as an existing Backup resource—even if the backup operation is already completed. When you apply a manifest with a different .spec.method (for example, switching from barmanObjectStore to plugin), Kubernetes updates the resource, and the METHOD field in kubectl get backup reflects the new spec, not the method originally used for the backup. This is standard Kubernetes behavior for custom resources: the latest applied manifest overwrites the resource fields, even for completed backups.

To avoid this, do not re-apply or update manifests for completed Backup resources. Each backup should have a unique name, and completed backups should be treated as immutable records. If you need to keep the historical METHOD value, avoid updating the resource after creation. This behavior is not explicitly documented as a feature or bug, but it is a consequence of how Kubernetes and the CNPG operator handle resource updates and immutability rules for the Backup spec (docs).

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}    

[alexeyshishkin-edb]

I think the problem can we solved if we had the Backup.spec.immediate field just like there's ScheduledBackup.spec.immediate