Postgresql running in Kubernetes with CloudNativePG operator

Intended audience

sysadm staff members

staging

In the staging environment, dedicated postgresql clusters are now running in kubernetes. They are managed by the CloudNativePG operator.

Each postgresql cluster are running a version 17 of postgresql.

Each postgresql cluster manages its own database with one primary and one stand-by for high availability.

The primary is running in one node of the kubernetes cluster and the stand-by in another node. Should one node go down (for maintenance or a plain crash), the operator will ensure the switch between the primary and the stand-by occurs between the primary and the replica.

Backups are configured to happen daily (over night). The wals and the backups are stored in our minio instance in a dedicated bucket s3://backup-cnpg/archive-staging-rke2/cnpg/<db-name>.

They are configured with a retention policy of 7 days for most of them.

Those databases, when relevant, can be accessed through a dedicated pgbouncer instance with the guest user (follow howto connect to postgresql instance).

staging-next-version

It’s a similar setup as the staging one except there is no high availability nor backups. As it’s a scratch environment, there is no guarantee about the data there.

So when an upgrade occurs, there will be some downtime (if services are using the databases to upgrade).

production

All ancillary databases are still running in bare metal machines for now so nothing in kubernetes yet.