Re: (ITS#3421) Replication problem with slurpd

[hyc@symas.com]

Yes, it sounds like a problem on the master. However, the logfile lock 
protocol looks pretty solid in the OpenLDAP code. If there is a problem 
here, it may be the C library failing to flush correctly, or the 
filesystem failing to manage locks.

raphael.ouazana@linagora.com wrote:

>Finally I can reproduce the bug (not always, but often). I am not sure
>it is exactly the same bug, but it looks like.
>
>What I have :
>1 master (IP : 10.69.247.38) and 2 slaves (10.69.247.57 and
>10.69.247.39). All are 2.2.19. I can join my slapd.conf files and my
>DB_CONFIG if you want.
>
>What I do :
>10 threads add 50000 entries to the master :
>dn: uid=userX-Y,o=example,c=fr
>objectClass: top
>objectClass: person
>objectClass: organizationalPerson
>objectClass: inetOrgPerson
>sn: userX-Y
>cn: userX-Y
>uid: userX-Y
>
>with X from 1 to 10 and Y from 1 to 50000
>
>
>So what I think :
>- the problem is on the master.
>- slurpd or slapd have a problem writing replog or slurpd.replog
>- maybe it is a lock problem or a flush problem
>
>Hope it helps...
>
>Raphael Ouazana
>

-- 
  -- Howard Chu
  Chief Architect, Symas Corp.       Director, Highland Sun
  http://www.symas.com               http://highlandsun.com/hyc
  Symas: Premier OpenSource Development and Support