By default Oracle does not write the external row identifier (primary key or unique key, if there is on) to the transaction logs in case of updates (inserts and deletes are always fully logged). During replication a row identifier is required to know how to process the data downstream.
Traditionally HVR requires supplemental logging to be enabled in order to fully capture the contents for an update, and we still document this as the best practice. However nowadays you can also set up HVR to capture the database’s ROWID as the surrogate key in an extra column, and use that to process the data downstream. Depending on your scenario this may work for you if there is no option to add supplemental logging to the tables on the source that don’t already have supplemental logging enabled.
Oops! Looks like you're using an outdated browser...
It appears you are still using Internet Explorer. Support for IE has been discontinued by Microsoft, and many features of our site may not display or work properly.
Please upgrade (it's free!) to Edge, Chrome, Safari, or any modern browser for the best user experience.