> On Apr 18, 2023, at 01:20, Robert Sjöblom <robert.sjoblom@xxxxxxxxxx> wrote: > Another idea we've had would be to use CTID to fetch the last row (update/insert) in each table on both sides and compare row content, is this feasible? Is it safe to rely on CTIDs across logical replication? No. CTIDs aren't sent over as part of logical replication, and there's no way to compare them between a publisher and a subscriber.
