Longhorn interactions

• Database volumes are excluded from the default recurring-job group, via the recurringJobSelector on longhorn-single. How that works, and why removing the label instead does not, is in Postgres.

Backups go to Scaleway bu-k3s

The default backup target is s3://bu-k3s@nl-ams/, set under defaultBackupStore in flux/20_controllers/longhorn/helm-release.yaml. The weekly-backup recurring job writes there for every volume in the default group and keeps 26 backups per volume.

The credentials are the SealedSecret backup-credentials in flux/20_controllers/longhorn/backup-sealed-secret.yaml, with AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, AWS_REGION (nl-ams) and AWS_ENDPOINTS (https://s3.nl-ams.scw.cloud). Seal it the same way as the Grafana secret, into longhorn-system.

The bucket is the one the previous cluster backed up to. Its backups show up in this cluster's UI and restore from there. Deleting a backup in the UI deletes it from the bucket, whichever cluster made it.

The UI is at longhorn.gewis.nl

flux/50_apps/longhorn/ingressroute.yaml publishes longhorn-frontend behind the authentik proxy outpost, like Hubble. The UI can delete volumes and backups, so unlike Hubble it only admits the groups in proxy_access_groups — see authentik configuration.