I'm wondering if a syncrepl behaviour I'm seeing is correct. My consumer is 4.4.42 and my provider 4.4.40. I'm using refreshOnly replication and when I first create the database entry on the consumer the whole process stops after 5-6 records.
I have tested the syncuser and its searches against the provider return
9000 results. So it's not a limit or auth issue. I have matched the
schemas on both servers successfully. I believe the issue is that the provider is sending the records in an incorrect order.
I get a record with an entryCSN in 20151001 and then a record with an entryCSN before that. So, if I understand it, operating correctly, the consumer complains the record is too old and ignores it, stopping replication. The contextCSN of the provider is in the 2016* range, the consumer's is stuck in 2015* range and the log spams "too old". I'm using the -c rid=XXX option to run slapd and deleting the DB between tests.
Shouldn't the provider send the entries in a entryCSN-ascending order during the initial sync? Thanks for any insight you can provide.