ntp are set up on all the masters and offset is quite small. so we can rule
out ntp as possbile cause.
On Dec 26, 2013 6:15 PM, "Daniel Jung" <mimianddaniel(a)gmail.com> wrote:
Set up 4 way MMR setup with 2.4.37, centos 6.
Reading the doc[0] indicates that adding the URL is an optional and is
used to avoid future serverID collision and using the number alone does
work without any issue.
Based on that information, i created 3 MMR setup with each provider with
unique serverids and unique rid within each sever pointing to other
providers.
I have seen examples where each provider has multiple serverIDs (all
providers), as the doc indicates, this is possible only if URL option is
specified. Since i didnt use the URL, i only configured single serverID per
the provider in the global cn=config.
I had tested cn=config with 2 MMR and i didnt see the CSN too old,
ignoring message in the log. But, when i set it up to be 3 way MMR, i am
seeing this msg with new sync where other provider sends do_synrepl2 to the
original provider and hence CSN too old, ignoring is logged.
I searched the mailing list and some of the reputable members here have
mentioned it is due to wrong MMR serverID setup. Could you point out if
serverid without URL may be the culprit to this problem? I see no other
issues other than this, replication to other consumers work fine and data
are in synced.
my setup serverID and rid setup.
serverID 1
rid 000 for replicating from serverID 2
rid 001 for replication from serverID 3
serverID 2
rid 000 for serverID 1
rid 001 for serverID 3
serverID 3
rid 000 for serverID 1
rid 001 for serverID 2
Thank you
[0]
http://www.zytrax.com/books/ldap/ch6/#serverid