I am a consultant DBA filling in for the DBA who has gone onto new adventures.
My client has PULL merge replication in place with 18 subscring servers on serveral publications.
The retention period has been kept very low at 3 days due to the volumne of data that is getting replicated.
I have come across an issue which is causing my head to hurt, a subscriber creates a record, it gets replicated as an INSERT as we there is an AUDIT table in place which captures the UTC datetime it occurs, eveybody happy at this stage. Just less than 3 days later an UPDATE is carried out to the record, row replicates, still eveybody happy.
Then four days later (last update: 08/03/2019 02:03:55 new update: 12/03/2019 08:01:33) this is when we get a conflict, in the conflict viewer it states that there was a update at the PUBLISHER and that is the winner, which was the update that was applied at the 08 March @ 02:03:55 and that the looser is the latest update that was carried out by the subscriber on the 12March @ 08:01:33
I have double-checked all the audit tables across all the servers and they all have same entries, you can see the different update date on the audit record when replication kicks in on each server, but I cannot see a record on the publisher that is causing the conflict to happen, all I can think of is that the retention period is less than the time-difference between the updates, therefore the publisher will always win with an update that is outside of the retention period, or am I being thick?