Il 22/09/14 15:34, unimore ha scritto:
once in a while, a modification entry in cn=accesslog on provider is
lost by consumer (both openldap-2.4.39). Later modifies on the same
entry are cought up, but entry is now inconsistent.
If I am right, eventually a "janitor process" recognizes messed up
entries and fixes them, but in my limited scenario it could take hours.
No, I am wrong.
replication on consumer fails with:
Sep 23 13:26:25 unimore01 slapd: do_syncrep2: rid=003 delta-sync
lost sync on (reqStart=20140923112625.000001Z,cn=accesslog), switching
Later on, consumer asks all entries to provider. That's how inconsistent
entries are eventually fixed.
No need to invoke a fictional janitor process.