I have hit a replication problem that I am scratching my head over.
We are in the midst of migrating from OpenLDAP from 2.3 to 2.4. In
our test environment we have a 2.3.43 master with 2.4.18 and 2.3.35
slaves. Replication to the 2.4 slave stopped this morning. When I
look at the log I see:
Oct 22 09:38:28 ldap-uat1 slapd[15032]: do_syncrep2: cookie=csn=20091022031949Z#000000#00#000000,rid=000
Oct 22 09:38:28 ldap-uat1 slapd[15032]: do_syncrep2: rid=000 CSN too old, ignoring 20091022031949.000000Z#000000#000#000000
Oct 22 09:38:28 ldap-uat1 slapd[15032]: do_syncrep2: cookie=csn=20091022031949Z#000001#00#000000,rid=000
Oct 22 09:38:28 ldap-uat1 slapd[15032]: do_syncrep2: rid=000 CSN too old, ignoring 20091022031949.000000Z#000001#000#000000
Oct 22 09:38:28 ldap-uat1 slapd[15032]: do_syncrep2: cookie=csn=20091022031949Z#000002#00#000000,rid=000
Oct 22 09:38:28 ldap-uat1 slapd[15032]: do_syncrep2: rid=000 CSN too old, ignoring 20091022031949.000000Z#000002#000#000000
Oct 22 09:38:28 ldap-uat1 slapd[15032]: do_syncrep2: cookie=csn=20091022031949Z#000003#00#000000,rid=000
Oct 22 09:38:28 ldap-uat1 slapd[15032]: do_syncrep2: rid=000 CSN too old, ignoring 20091022031949.000000Z#000003#000#000000
Oct 22 09:38:28 ldap-uat1 slapd[15032]: do_syncrep2: cookie=csn=20091022031949Z#000004#00#000000,rid=000
Oct 22 09:38:28 ldap-uat1 slapd[15032]: do_syncrep2: rid=000 CSN too old, ignoring 20091022031949.000000Z#000004#000#000000
Oct 22 09:38:28 ldap-uat1 slapd[15032]: do_syncrep2: cookie=csn=20091022031949Z#000005#00#000000,rid=000
Oct 22 09:38:28 ldap-uat1 slapd[15032]: slap_queue_csn: queing 0x7f09c283b075 20091022031949Z#000005#00#000000
Oct 22 09:38:28 ldap-uat1 slapd[15032]: syncrepl_message_to_op: rid=000 mods check (homePostalAddress: value #0 invalid per syntax)
Oct 22 09:38:28 ldap-uat1 slapd[15032]: slap_graduate_commit_csn: removing 0x7f09c1621438 20091022031949Z#000005#00#000000
So, I slapcat'ed the master's database and looked for that contextCSN.
The entry that appears to be the problem has a homePostalAddress of:
homePostalAddress: c\o Bain & Company, Two Copley Place, Boston, Massachusetts
We had hit this problem before when loading a 2.4 server using slapcat
output from a 2.3 server and wrote a simple filter to correctly quote
the backslash.
This is a problem for us because we expect to have a mixed 2.3/2.4 during
the transistion to 2.4 on our production servers. The really nasty bit is
that the only way I know for sure to fix this problem is to reload the
slave. Is there an alterntive way?
Bill
--
Bill MacAllister, System Software Programmer
Unix Systems Group, Stanford University