[Issue 9858] cn=config replication refresh breaks mdb database

https://bugs.openldap.org/show_bug.cgi?id=9858

--- Comment #4 from Quanah Gibson-Mount quanah@openldap.org --- (In reply to Howard Chu from comment #3)

Looks like syncrepl caused the particular backend's config to change, forcing it to be closed and reopened. One of the changes affected index definitions, so the online indexer task was added to the runqueue. But because of the close and reopen it was running on a backend that was no longer open.

There was no configuration change at all. I can trivially trigger this behavior as I noted initially simply by making the cn=config database on a replicated node do a full REFRESH on cn=config. There have been no changes made to indexing on these systems in at least a year.