The Connection Manager provides the following features. You can use these features to increase system availability.

This feature can only be used in Advanced Edition.

Detects kernel panics between the server running the client and the server running the instance, physical server failures, and inter-server network link downs, and notifies the client or instance. The client is notified as an error event through the SQL connection, and the instance will be notified in the form of a force collection of SQL connections with clients that are out of service.

When an application wants to connect to an instance of an attribute (Primary/Standby) configured with replication, it can do so without knowing which server the instance is running on.

With load balancing, if selection is possible, connections can be preferentially made to the server accepting the fewest connections.

Using read connection redistribution, when a standby server is promoted, Connection Manager disconnects existing connections, allowing the promoted server to focus on the workload of the newly added primary server. Only connections using Connection Manager are disconnected, and the disconnection occurs when a new query is executed.

Additionally, when the number of connections between standby servers becomes uneven, existing connections can be disconnected. This gives applications an opportunity to reconnect, thereby balancing the number of connections. For example, this can handle changes in the number of connections due to the addition or removal of standby servers, or the suspension of specific operations. Only non-update connections using Connection Manager are disconnected, and the disconnection occurs when a new query is executed.

You can exclude standby servers that are not synchronized with the primary server on a transaction-by-transaction basis from the connection destination candidates.