We have uncovered a critical data loss issue with the following CDC patches.
1) 6.5.1 interim fix 1 for Oracle
2) 6.5.1 interim fix 1 for Oracle (Trigger)
3) 6.5.0 interim fixes 31 and 32 for Oracle
4) 6.5.0 interim fixes 12 and 13 for DB2 LUW
Please see APAR JR40911 https://www-304.ibm.com/support/entdocview.wss?rs=0&uid=swg1JR40911
The data loss is unpredictable and occurs at the communications layer when the comms queue becomes full. This behavior will not be obvious to the user.
There is a field work-around. Set the SOURCE system parameter global_statistics_blocking_threshold_milliseconds to a very high value, such as 6854775808,
followed by a restart of the CDC instance
The permanent fix for this issue is expected to be available on Oct. 31, 2011