On Wed, Apr 15, 2015 at 02:40:37PM +0200, Ulrich Windl wrote:
julien soula jsoula@univ-lille2.fr schrieb am 15.04.2015 um 13:29 in
Nachricht 20150415112905.GB3225@nickel.univ-lille2.fr:
hello,
I wanted to synchronize 2 branches of a master DB (slapd-2.4.38). So I created 2 olcSyncrepl on the slave :
olcSyncrepl: {0}rid=201 provider=ldap://master searchbase="cn=branch1,suffixDB" scope=sub olcSyncrepl: {1}rid=202 provider=ldap://master searchbase="cn=branch2,suffixDB" scope=sub
Unfortunatly, it doesn't work. A change on branch2 on the master produces often a "CSN too old" on the slave.
After investigating, it seems that the pb comes from the fact there is one contextCSN by DB. So if the sync task on branch1 is the first to process, it updates the contextCSN and therefore the sync task on branch2 thinks that change is not newer. Am I right ?
If the message you are talking about is like this: slapd[3965]: do_syncrep2: rid=001 CSN too old, ignoring 20150409131449.846699Z#000000#001#000000 (olcDatabase={1}hdb,cn=config)
yes
_and_ the CSN received is the CSN sent,...
yes
... then just ignore the message. That is the same server said before: slapd[3965]: slap_queue_csn: queing 0x7f5f30afcf20 20150409131449.846699Z#000000#001#000000
So is there a proper way to achieve what I want ?
Did you check the databases before and after sync? Are there differences, or are you just worried about the messages?
In fact, I first noticed the no-change of the slave then I took a look to the log and saw this message (the only suspect I saw).
Is this config supposed to work ?
sincerly,