The one thing to understand first
Google Cloud SQL for PostgreSQL is Google’s member of the Postgres-on-a-disk family: real community PostgreSQL running on a Compute Engine VM with a Persistent Disk underneath, managed by Google’s control plane. It is the direct counterpart to AWS RDS — the same idea, the same trade-offs, different plumbing.
Storage on Persistent Disk
Your database files sit on a Google Persistent Disk, a network-attached block device that is itself replicated for durability beneath the surface. Cloud SQL can automatically grow the disk as it fills. As with all of family one, there is a single primary that accepts writes.
High availability: regional instances
A Cloud SQL HA (“regional”) instance runs a primary in one zone and a standby in another zone of the same region. Crucially, durability rides on the storage layer: writes are synchronously replicated to a Persistent Disk in the standby zone, so when the primary’s zone fails, the standby starts up against an already-current copy of the data. Failover typically completes in around a minute. As with RDS Multi-AZ, the standby is for failover, not for serving reads.