Managed Relational Databases

What "managed" gives you

The provider handles installation, patching, backups, point-in-time recovery, monitoring, failover and storage growth. You choose the engine, size, network placement and parameters, and you still own schema design, queries and access control.

High availability and scaling

• Multi-AZ / zone-redundant HA: a synchronous standby in another zone takes over automatically (typically in under a minute or two). Protects against zone failure, not against bad writes.
• Read replicas: asynchronous copies for read scaling and reporting; they can lag behind.
• Backups: automated daily snapshots plus transaction logs enable point-in-time restore (to any second within the retention window).

replica_lag_s = 2.5
events = [(0.0, "write: order 7 status=paid (primary)"), (1.0, "read on replica"), (3.0, "read on replica")]
status_on_replica = "pending"
for t, e in events:
    if e.startswith("read"):
        current = "paid" if t >= replica_lag_s else status_on_replica
        print(f"t={t:3.1f}s {e:36} -> order 7 is '{current}'")
    else:
        print(f"t={t:3.1f}s {e}")
print("read-your-own-writes: send reads right after a write to the primary")

t=0.0s write: order 7 status=paid (primary)
t=1.0s read on replica                      -> order 7 is 'pending'
t=3.0s read on replica                      -> order 7 is 'paid'
read-your-own-writes: send reads right after a write to the primary

Connect from private subnets only, use IAM/Entra ID authentication or managed secrets, and size connection pools: databases have connection limits that serverless and autoscaled apps can exhaust.