I mean run multiple MySQL server instances on different machines when you have large clusters with heavy load. Usually you can co-locate the MySQL with the service role that uses it.
Derby and SQLite should not be used in production if the service has a real database alternative.
I think the problem is you tried to create two jobtrackers in the wizard. You should only pick one. We are already planning to enhance this to make sure you can't pick two by accident.
After you've set up everything in simple mode, go to MapReduce and go through the JobTracker High Availability wizard there.