A PolarDB-X instance is a distributed cluster that consists of multiple nodes. Each node provides features such as SQL routing, data merging, and data aggregation. You can change the node specifications for an instance. During peak hours, you can upgrade the specifications to handle high numbers of queries per second (QPS). During off-peak hours, you can downgrade the specifications to reduce the waste of resources.
Additional considerations
- When you downgrade the specifications of an instance, transient connections may occur due to disconnections between your application and the instance. Make sure that your application can automatically reconnect to your instance.
- You can host your service on Elastic Compute Service (ECS) instances and connect the ECS instances to your PolarDB-X instance. After the specifications of the PolarDB-X instance are upgraded, additional nodes are added to the instance. If persistent connections are established between the ECS instances and the PolarDB-X instance, the newly added nodes cannot receive network traffic immediately after the instance specifications are upgraded. To resolve this issue, we recommend that you restart the ECS instances in phased mode. Restart the ECS instances that run the smallest workloads and analyze the performance of the service. After you confirm that the service runs as expected, you can restart all other ECS instances. You can use the Druid connection pool to connect to PolarDB-X instances. For more information, see Configure a frontend connection pool to connect an application to PolarDB-X.