Topology & multi-region
Node groups, synchronous replicas, and multi-region disaster-recovery layouts for PostgreSQL clusters.
Beyond single-node and basic high availability, the advanced provisioning flow gives PostgreSQL clusters finer control over topology.
Replicas & synchronous replication
A high-availability cluster runs a primary plus one or more streaming replicas with automatic failover. By default, one replica is synchronous: a commit returns only after that replica has it, so a failover loses no committed data.
You can set the number of synchronous replicas from 0 up to the number of electable nodes minus one. Zero means all replicas are asynchronous: commits return faster, but a failover can lose the last commits that had not reached a replica.
Multi-region distribution
For PostgreSQL, choose a distribution model:
- Single cluster — all nodes in one region.
- Multi-region standby (DR) — a standby in another region for disaster recovery.
- Multi-region stretched — nodes spread across regions.
A primary-region selector controls where the writable primary lives.
Node groups
Advanced topology is expressed as node groups:
- Electable node groups participate in elections (with a priority).
- Replica pools can have their own instance type and be promotable or sync-eligible.
Each node group has a role:
| Role | Can be promoted | Acknowledges commits |
|---|---|---|
electable | Yes | Yes |
standby | Yes | No — stays out of the commit path (a remote group keeps failover without slowing commits) |
read_only | No | No |
Storage and IOPS for the cluster are set alongside topology. Storage starts at the tier's base size (5 GiB for R5) and, on AWS gp3, goes up to 16 TiB; gp3 IOPS range from 3,000 to 80,000 depending on volume size (at most 500 IOPS per GiB). See Scaling & storage for the rules and to change them later.