Connect CDC on z/OS provides the ability to set up multiple captures (for example, one or two Db2). Each capture is associated with a transient storage that is set to store capture data that cannot fit into memory. Each capture can have multiple publishers (targets/connections) associated with them. However, data stored in the transient storage will not be cleared until all publishers associated with the capture acknowledge to the Connect CDC server that the data can be cleared. This can create a problem where the transient storage of the capture gets filled up and causes the Connect CDC server to deadlock.
Consider this example: Publisher A is set up to read data from Table A, and Publisher B is set up to read data from Table B; however, if Table A updates enough records to fill up the transient storage, even if Publisher A consumes the data and acknowledges to the Connect CDC server that the data is consumed and therefore can be cleared, the server will not clear the data until Publisher B also sends an acknowledgment message.
To avoid this problem, Precisely recommends setting up one publisher per capture. If there is a need for multiple publishers for a given capture, we recommend that either the data for both publishers be consumed simultaneously or that the data for both publishers gets consumed one after another at around the same time.