3.8 Multi-Host Connections

3.8.1 Configuring Server Failover for Connections Using JDBC

3.8.2 Configuring Server Failover for Connections Using X DevAPI

3.8.3 Configuring Load Balancing with Connector/J

3.8.4 Configuring Source/Replica Replication with Connector/J

3.8.5 Advanced Load-balancing and Failover Configuration

The following sections discuss a number of topics that involve multi-host connections, namely, server load-balancing, failover, and replication.

Developers should know the following things about multi-host connections that are managed through Connector/J:

• Each multi-host connection is a wrapper of the underlying physical connections.

• Each of the underlying physical connections has its own session. Sessions cannot be tracked, shared, or copied, given the MySQL architecture.

• Every switch between physical connections means a switch between sessions.

• Within a transaction boundary, there are no switches between physical connections. Beyond a transaction boundary, there is no guarantee that a switch does not occur.

  Note

  If an application reuses session-scope data (for example, variables, SSPs) beyond a transaction boundary, failures are possible, as a switch between the physical connections (which is also a switch between sessions) might occur. Therefore, the application should re-prepare the session data and also restart the last transaction in case of an exception, or it should re-prepare session data for each new transaction if it does not want to deal with exception handling.