On Wed, 2008-04-16 at 08:34 +0000, marian.eichholz@freenet.ag wrote:

Full_Name: Marian Eichholz Version: HEAD-20080411 OS: Linux URL: ftp://ftp.openldap.org/incoming/ Submission from: (NULL) (194.97.7.65)

When I made a consistency check with our two syncrepl consumers, I found several irritating differences.

After the initial refresh cycle, the DNs were complete. But when the consumers got all new objects after, they now have objects that were deleted on the provider in the meantime. Unfortunately I cannot tell when that happened to a specific DN. The set of undeleted objects is not identically on the two consumers. So I guess that it is not a problem with the data itself.

So the consumers are corrupted now.

Both consumers suffer from hanging replication sessions. This is why we switched from 2.4.8 to HEAD (11.4.2008, see ITS 5463). They were stopped and restarted repeatedly, of course. The daemons have never been killed. So the DBD backend SHOULD be OK.

Any idea how to find such events in the LOCAL4-Log? I did not apply "any" as loglevel for predictable performance issues and so liked to keep the entries to a reasonable volume.

Thank You very much for any idea.

Yours sincerely - Marian Eichholz (freenet AG)

Hi Marion. I saw both those problems with 2.4.8. But they both seem to have been fixed in the RE24 cvs tag as of 14/04/08. I have an arrangement of syncrepl servers based on that release that has been running fine for a few days now with about 10 connections per second on the consumers. Martin.