Reported by a customer on 6.0.0:
- replication topology with two DS-only, and two DS+RS instances.
- at restart, the DS+RS instances' error logs have thousands of errors like this (the DS-only instances have such errors too, but only a few) :
- the errors are for lots of different CSNs (and different DNs).
- many (or all?) of the CSNs have been purged from the changelogs.
I have been able to reproduce this only once so far (no consistently reproducible test case):
- once the error happens, it happens at each restart after that, on the same CSN.
My steps were:
- DS 6.0.0. Three instances: DS1 (DS-only), DS+RS2, DS+RS3.
- set up "dc=example,dc=com", and a separate backend "ou=sub,dc=example,dc=com".
- I set a global server-id for each instance (81, 82, 83).
- configure replication for the suffix and sub-suffix.
- add entries on all instances.
- set replication-purge-delay to several hours on DS+RS2 and DS+RS3.
- add more entries.
- lowered purge delay again on DS+RS2 only (this time to 1 minute).
- old changes got purged.
- restarted DS+RS2, and the "not found in pending list" error was reported on DS+RS3.
- after that, the error is reported on DS+RS3 at each restart of DS+RS3:
- in my case, the "not found in pending list" CSN originates from DS+RS2, and is the "newest-csn":