It's a short list, but, if I find more later, I'll add them.
- You will most likely need to configure one of two federation interfaces - the ODBC wrapper or the JDBC wrapper.
- Why? CDC already supports a wide range of databases, so you'll probably be targeting one for which IBM federation doesn't have a named wrapper. For example, it doesn't have a "PostgreSQL wrapper."
- If you use the JDBC wrapper, use the Create Server options found in the JDBC wrapper examples on ChannelDB2.
- If the replication technology has its own tables that are written to in the same unit of work as target table nicknames, consider creating that table in the target database and then create a nickname for that table.
- For example, I believe you'll need to do this for CDC's TS_BOOKMARK table.
- If you do not do this, you will need to use 2-phase commit, which may not be available with a given target or may hurt performance.
- If you create nicknames for multiple target databases, have the replication technology process different target databases in different units of work.
- For example, create each CDC subscription so that it writes to nicknames for only for one target database within that subscription.
- Otherwise, you hit the 2-phase commit issue mentioned above.
- If you need to write to both local tables in the federated database and nicknames, put them in separate units of work.
- For example, that would mean having either local tables or nicknames in a CDC subscription, but not both.
- Otherwise, we're back to the 2-phase commit issue.
If you have questions, you can ask them in the comment section of this blog or use the message boards here on developerWorks.