You can use the Q Capture program's ability to ignore specified transactions to improve performance in a pure, two-server bidirectional configuration.
About this task
To avoid the recapture of transactions, by default the Q Apply program inserts P2PNORECAPTURE signals into the IBMQREP_SIGNAL table for each transaction that it applies. The signals are inserted at the Q Capture server that is shared by the Q Apply program. When Q Capture reads the signals in the log, it ignores these transactions.
When there are many bidirectional Q subscriptions, the number of signal inserts can affect replication performance. To avoid this, you can specify that the programs use the IBMQREP_IGNTRAN table to avoid recapture. This method tells the Q Capture program to automatically ignore any changes that come from the Q Apply program. You also start the Q Apply program with the insert_bidi_signal=N startup parameter.
Use the following guidelines to determine which method to use to avoid recapture:
Configuration | Recapture avoidance method |
---|---|
Multiple bidirectional configurations between two servers, or a combination of bidirectional, unidirectional, and peer-to-peer configurations | You must accept the default method The default method of signal inserts ensures that all changes are propagated correctly between servers. If you start with a pure, two-server bidirectional topology but plan to later add branching unidirectional or peer-to-peer configurations, you should also accept the default method of recapture avoidance. |
Pure, two-server bidirectional configuration | Performance can be improved by using the
IBMQREP_IGNTRAN table to avoid recapture If you use the IBMQREP_IGNTRAN table method, do not later add branching unidirectional or peer-to-peer configurations to the bidirectional configuration. |
Procedure
To use the IBMQREP_IGNTRAN table to avoid recapture in bidirectional replication: