Configuring Two-Phase Commit Engine

If you have multiple databases involved in your transaction, you must configure a two-phase commit engine for managing all changes to all databases involved in the transaction. A two-phase commit engine is responsible for contacting all of the databases at the end of the transaction and managing the commit or rollback of all updates to all included databases. Thus, this two-phase commit engine must have access to database links to each database included within the transaction.

To configure for a two-phase commit, your system administrator must do the following:

1. Designate an Oracle8i database as the two-phase commit engine.

2. Configure database links from the two-phase commit engine to each database that may be involved in the global transaction. This is necessary for the two-phase commit engine to communicate with each database at the end of the transaction.

3. Provide the database link name in the -dblink option of bindds for each individual database when binding that database's DataSource into the namespace.

   bindds /test/empDatabase -url jdbc:oracle:thin:@empHost:5521:ORCL 
   
   	-dstype jta -dblink 2pcToEmp.oracle.com
   
   

4. Provide the two-phase commit engine's URL address, username, and password when binding the UserTransaction into the namespace.

   bindut /test/myUT -expprop -url jdbc:oracle:thin:@2pcHost:5521:ORCL
   
   	-user SCOTT -password TIGER
   
   

Once all of this configuration is complete, your client or server differs from the single-phase commit scenario in the following aspects:

• You must provide the username/password when retrieving the database connection through DataSource.getConnection.

• All DataSource objects retrieved must have been bound with a database link from the two-phase commit engine to itself.

Except for the in-session activation rules, the client and the server code is the same. The following example shows a server object that performs an in-session activation to retrieve both the UserTransaction and DataSource objects that have been bound locally. So, all that they need to provide is the username and password information for the two-phase commit engine.

//create the initial context.
InitialContext ic = new InitialContext ();
UserTransaction ut = (UserTransaction)ic.lookup ("/test/myUT");

// lookup the local datasource and a remote database.
DataSource localDS = (DataSource)ic.lookup ("/test/localDS");
DataSource remoteDS = (DataSource)ic.lookup ("/test/NewYorkDS");
Connection localConn = localDS.getConnection ();
//provide the database username and password in the getConnection method
Connection remoteConn = remoteDS.getConnection (SCOTT, TIGER);
...
//close the connections
localConn.close();
remoteConn.close();

//end the transaction
 ut.commit();

For a client or a server where the namespace is remote, you must initialize the environment with the authentication information, namespace location, and register the OracleDriver.

---