Re: (ITS#3512) LDAP Sync Replication stops after single server fa ilure

[hyc@symas.com]

Darren Gamble wrote:

>We have had replication break yet again, although it appears to only have
>failed on about half of our consumers this time (last time, it failed on all
>of them).  Again, cleaning out the replica's DB and resyncing the replica
>will resolve the issue.
> 
>We started one of the broken replicas up with debugging (L7) enabled.  We
>saw that it seems to connect to the provider server, and then it appears to
>find the information that's changed, as there are "bdb_modify_internal:
>delete and bdb_modify_internal: replace" messages for these entries.  But,
>the local entries on the server don't actually get changed.  If we stop and
>start the server again, we see it connect to the provider again and do a
>couple of queries, but no more.  There are no (obvious) error messages.
>
>Some of the information in the debug file is sensitive- is there a secure
>place we could send the file to, that only one or two developers could see,
>perhaps?  Is there anything in particular we could be looking for in the
>log?
>
Mainly I wanted to see that the context information in the syncrepl 
cookie was going across the network correctly, and that the replica was 
correctly writing the changes to its local database. I'm surprised that 
you can see modify requests succeeding and yet see no change in the 
local database.

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