> The problem was mine, no syncrepl. I usually start the client with a > fresh/recent dump from the master, but neglected to dump the extended > attrs along with the regular ones. That resulted in "slapadd" creating > new entryUUID/etc attrs on the consumer, which then would not match > the master. > > If a warning would be useful, then perhaps "slapadd" could warn on > importing an LDIF without entryUUIDs, but it's not a bug in this case. > > An alternate to reduce "operator error" might be to compare a master > UUID on the first sync and notify of a mismatch, which could also help > prevent accidentally trying to sync from the wrong master.
Please keep replies to the ITS. OK, I understand. In this case, slapadd(8) needs to be able to load a LDIF without operational attrs and create them accordingly. Still, there's room for warning in that case if slapadd(8) is done for a database configured to use syncrepl. p.
