Postgres-operator: Clone from gcs using wal-g

[jaroslav-muller]

Hi,
I created a cluster that backups using wal-g to gcs. However when I try to clone that cluster (to point in time) from gcs it fails.
The source of failure is that envdir /home/postgres/etc/wal-e.d/env-clone-jarin-db-cluster doesn't contain variables that are needed (and are present in /home/postgres/etc/wal-e.d/env.
I can actually make the clone work by manually copying two variables from env to env-clone-jarin-db-cluster - GOOGLE_APPLICATION_CREDENTIALS and USE_WALG_RESTORE.

I use custom postgres-pod-config.yaml to set the variables:

apiVersion: v1
kind: ConfigMap
metadata:
  name: postgres-pod-config
  namespace: default
data:
  USE_WALG_BACKUP: true
  USE_WALG_RESTORE: true
  WALG_GS_PREFIX: gs://jarin-test-backup/spilo/$(SCOPE)
  CLONE_WALG_GS_PREFIX: gs://jarin-test-backup/spilo/$(CLONE_SCOPE)
  WALE_BACKUP_THRESHOLD_PERCENTAGE: "100"
  GOOGLE_APPLICATION_CREDENTIALS: /gcs/backup-google-credentials.json

Is there a way to set env variables in the clone envdir? If not I'd consider it a bug.

[FxKu]

Are you trying to clone the cluster to another namespace, maybe? Then I think your config map has to exist in the other namespaces, too, regarding this line.

[jaroslav-muller]

Nope, same namespace. Also, the config map (and environment variables) is propagated to the container successfully. All the variables are configured correctly in the envdir /home/postgres/etc/wal-e.d/env.
The problem is that cloning uses its own separate envdir /home/postgres/etc/wal-e.d/env-clone-jarin-db-cluster and this is where the variables are missing.
I suppose that the envdir isolation is deliberate (at least to separate CLONE_WALG_GS_PREFIX from WALG_GS_PREFIX), however it is isolated too much. For gcp restore to work, we need GOOGLE_APPLICATION_CREDENTIALS and USE_WALG_RESTORE as well.

[FxKu]

ah ok. As of now I don't see if the operator can do something about this separate envdir. @CyberDem0n or @erthalion ?

[jaroslav-muller]

Seems that I found the answer. Any env variable from postgres-pod-config confmap that starts with CLONE_ is copied to the clone envdir (after being stripped of the CLONE_ prefix). So adding CLONE_WALG_GS_PREFIX and CLONE_GOOGLE_APPLICATION_CREDENTIALS to the postgres-pod-config solved my problem.
Having it mentioned in the documentation would help.

[FxKu]

Whaaat?! I did not know about this, sorry. That's def. not very intuitive. Will try to update the docs in this regard. Thanks for keep trying and posting the solution here.