The Connector/NET supports connection pooling. This is enabled by default, but can be turned off via connection string options. See Section 20.2.4.3, “Creating a Connection String” for further information.

Connection pooling works by keeping the native connection to the server live when the client disposes of a MySqlConnection. Subsequently, if a new MySqlConnection object is opened, it will be created from the connection pool, rather than creating a new native connection. This improves performance.

To work as designed, it is best to let the connection pooling system manage all connections. You should not create a globally accessible instance of MySqlConnection and then manually open and close it. This interferes with the way the pooling works and can lead to unpredictable results or even exceptions.

One approach that simplifies things is to avoid manually creating a MySqlConnection object. Instead use the overloaded methods that take a connection string as an argument. Using this approach, Connector/NET will automatically create, open, close and destroy connections, using the connection pooling system for best performance.

Typed Datasets and the MembershipProvider and RoleProvider classes use this approach. Most classes that have methods that take a MySqlConnection as an argument, also have methods that take a connection string as an argument. This includes MySqlDataAdapter.

Instead of manually creating MySqlCommand objects, you can use the static methods of the MySqlHelper class. These take a connection string as an argument, and they fully support connection pooling.