[Date Prev][Date Next]
openldap replication issue
- To: openldap-software@OpenLDAP.org
- Subject: openldap replication issue
- From: Moe <firstname.lastname@example.org>
- Date: Fri, 4 Aug 2006 00:32:36 -0700 (PDT)
- Domainkey-signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.com; h=Message-ID:Received:Date:From:Subject:To:MIME-Version:Content-Type:Content-Transfer-Encoding; b=TJ5I+ldOOgzW4lTp2DJoJj+JDkCOszbef/O7/RmrDsKvX5pZuPMqkvsWZZC+NgS+hjrTMxrqCBuPbdZ+OU26G67vhOcOrvR7i5/1uYw5le834RPptDScUkAOPp6CP5qdgeFuE0VXC6j8v2V6b2CD9+rdyTHMrfzE1Qnoa12x4Gc= ;
I have a master slave replication configuration using slurpd.
openldap version is : 2.2.26 and database is berkeley db 4.3.28.
Entries from master to slave gets replicated fine for a while before i start seeing the following behaviour:
I create an entry in the master but it does not show on the slave side when i do ldapsearch. Same things happens as i create more entries (Master and slave r out of sync now /slave has less entries ).
Sometimes if i keep creating new entries, some old entries that were not showing on the slave site appear when i do ldapsearch but the newly created ones don't appear.
If i shutdown slapd on the slave site and start it, some of the entries that were not getting displayed by ldapsearch on the slave site suddenly appears, but not all of them.
After i restart slapd on the slave site, new created entries appear on the slave and slave starts working normal. At this point i still have master and slave out of sync (some permenant missing entries that were created before slapd on the slave was restarted).
I have enough cache size set in slapd.conf and im using checkpoint parameter too. I have it set every 5 minutes.
My guess is that the dbwriter of the slave is crashing after a while due to some kind of problems on the slave box. And when i restart slapd on slave things start acting normal again.
Any suggestions on what might be happening ?
Also when the master and slave are out of sync (slave is missing some entries) what would explain the behaviour where adding a new entry to the master result in poping one of the old missing entries on the slave?.