techblog 

We are trying to migrate our database from 10.1.0.3 running on Solaris to 10.2.0.2 running on Linux. To do this I thought using transportable tablespaces was a great idea to speed things up. However there is a major bug with using transportable tablespaces, dropping a user that has been transported might end up generating this:

SQL> drop user transported cascade;

drop user transported cascade

*
ERROR at line 1:
ORA-00600: internal error code, arguments: [ktecgetsh-inc], [2], [],
[], [], [], [], []

This actually terminates the instance, and if you try and restart the instance it will be terminated again due to the following which appears in the alert log.

Non-fatal internal error happenned while SMON was doing temporary segment drop.
SMON encountered 6 out of maximum 100 non-fatal internal errors.

Instance terminated by PMON, pid = 24149

Every time the instance is restarted this error will ocurr and the instance will be terminated. Oracle are aware of this and it is bug: 5186784, allegedly fixed in 10.2.0.3. There is a workaround to get your database back up and running and it is to set an event:

SQL> alter system set event="10061 trace name context forever, level 10" scope=spfile;

Once you have this event set, the instance will not crash and you can drop the user, you may also have to drop the tablespace. Once this is done you can safely remove the event with:

SQL> alter system reset event scope=spfile sid='*';