Starting from February 8, 2021, AnalyticDB for PostgreSQL supports multiple coordinator nodes, which means a single instance can expose two connection endpoints: the primary endpoint and the cluster endpoint. Choosing the right endpoint determines whether your application gets maximum compatibility or maximum connection throughput.
How it works
When you create an instance or change coordinator node specifications, AnalyticDB for PostgreSQL automatically creates a cluster endpoint if you select more than 8 compute units (CUs) of coordinator node resources. The primary endpoint is always available.
Choose an endpoint
Use the primary endpoint when compatibility is the priority:
All requests go to the primary coordinator node, so behavior is identical to a single-coordinator-node instance.
Your application relies on session state, temporary tables, or transaction behaviors that require a consistent connection to one node.
Use the cluster endpoint when throughput is the priority:
Connections and I/O capabilities scale linearly with the number of coordinator nodes, as long as compute nodes are not a bottleneck.
Your workload can tolerate load balancing across nodes.
The cluster endpoint has functional limitations compared to the primary endpoint. Review the cluster endpoint limitations before switching existing applications.
What's next
Manage coordinator nodes — Add or remove coordinator nodes to scale your instance.
Connect to an instance — Get the endpoint address from the console.
If you have questions, join the DingTalk group 11700737 for technical support.