Le 22/05/2012 18:47, Quanah Gibson-Mount a écrit :
--On Tuesday, May 22, 2012 10:06 AM +0200 Jehan Procaccia
> Le 21/05/2012 18:33, Quanah Gibson-Mount a écrit :
> --On Monday, May 21, 2012 3:15 PM +0200 jehan procaccia
> <jehan.procaccia(a)it-sudparis.eu> wrote:
> ok __dd.xxx files are memory mapped files,
> then which is the file(s) that contains the database ? I suspect
> id2entry.bdb, but why on the master the size in KB is:
> 195412 id2entry.bdb
> and on the slave
> 32 id2entry.bdb
> It doesn't look to me like your replica is replicating at all. Have you
> done a slapcat of the two and verified they are identical?
> Ok, then I did a slapcat on a replica and to my big surprise, before I
> did the slapcat, bdb files on the replica were very small, then as soon
> as I did the slapcat (while slapd is off), now they are similar in size
> to the master !?
> Do we need to run regularly slapcat on replicas in order to "really"
> replicate ?
No. It sounds like you don't have a checkpoint directive set for your
database. You should fix that.
You are right, now that I added
checkpoint 200 240
in the slapd.conf, bdb files are bigger !
in fact my old repicas in 2.3 had that directive in global configuration
whereas apparently in 2.4 that directive must be in a database
definition, that's probably why I missed that one during the migration