The Capture program and the Apply program

The Capture program uses some of the replication control tables to indicate what changes have been made to the source database, and the Apply program uses these control-table values to detect what needs to be copied to the target database.

The Capture program does not capture any information until the Apply program signals it to do so, and the Apply program does not signal the Capture program to start capturing changes until you define a replication source and associated subscription sets.

The following lists describe how the Apply and Capture programs communicate in a typical replication scenario to ensure data integrity:

Capturing data from a source database

1. The Capture program reads the IBMSNAP_REGISTER table during startup to identify those registered replication sources for which it must capture changes. Having done so, it holds their registration information in memory.
2. The Capture program reads the DB2® log or journal continuously to detect change records (INSERT, UPDATE, and DELETE) for registered source tables or views. It also detects inserts to the IBMSNAP_SIGNAL table in order to pick up signal actions that have been initialized by the Apply program or a user. When the Apply program inserts a CAPSTART signal in the IBMSNAP_SIGNAL table and the Capture program detects the committed signal, the Capture program initializes the registration and starts capturing changes for the associated source.
3. Once the Capture program has started capturing changes for a registered source, the program writes one row (or two rows if you specified that updates should be saved as DELETE and INSERT statements) to the CD table for each committed change that it finds in the DB2 log or journal. The Capture program keeps uncommitted changes in memory until the changes are committed or aborted. Each registered replication source that is not an external CCD table has an associated CD table.
4. At each commit interval, the Capture program commits the data that it has written to the CD and UOW tables, and also updates the IBMSNAP_REGISTER table to flag which CD tables have new committed changes.

Applying data to a target database

1. For all newly defined subscription sets, the Apply program first signals the Capture program to start capturing changes. Then, a full refresh is performed for each member of the set (unless it is not a complete target table).
2. When any subscription set is eligible for replication, the Apply program checks the IBMSNAP_REGISTER table to determine whether there are changes that need to be replicated.
3. The Apply program copies any changes from the CD table to the target table.
4. The Apply program updates the IBMSNAP_SUBS_SET table to record how much data the Apply program copied for each subscription set.
5. The Apply program updates the IBMSNAP_PRUNE_SET table with a value that indicates the point to which it has read changes from the CD table.

Pruning the CD tables

When the Capture program prunes the CD tables, it uses the information located in the IBMSNAP_PRUNE_SET table to determine which changes were applied, and deletes those changes already replicated from the CD table.