[Date Prev][Date Next]
Re: hints for troubleshooting Replication problems
On 6/25/06, Pavlos Parissis <firstname.lastname@example.org> wrote:
On Sun, 25 Jun 2006 13:03:34 -0700
Quanah Gibson-Mount <email@example.com> wrote:
> --On Sunday, June 25, 2006 6:49 PM +0200 Pavlos Parissis
> <firstname.lastname@example.org> wrote:
> > Do you have any hints how I can troubleshoot the issue and spot the
> > problem?
> The problem is that you are using an ancient version of OpenLDAP
> which only had LDBM as the backend, and it is likely corrupting
> itself. You should really look at upgrading to a modern, supported
> release of OpenLDAP.
I agree with you that the old version of OL could be the root cause of my problems.
But, I need to make sure that is the case and not something else.
Imagine the scenario of upgrading and still have the same problem.
I believe that the only way to avoid the mentioned scenario is to find technical evidences which
support what you are saying. I suppose that you say that because you have seen similar problems
and the cause of them was the LDBM backend.
Thus, I asked about hints which could help me to find these evidences and make a proposal to my Managers
for upgrading to the latest version.
Just do a slapcat or ldapsearch objectclass=* on the master and then
on the replicas to find missing entries. You will then have to
manually fix them. Or just copy your master's database to the slaves
and start them from scratch.
As far as a proposal for upgrading, I installed slurpd and found that
openldap 2.3 + bdb could do write operations 10x faster than openldap
2.1+ ldbm and read operations 5-6x faster. That means that the same
hardware you're currently using can now service at least five times as
many clients. My management seemed very pleased with the free upgrade
that equated to $100k worth of server equipment and no additional
licensing/support costs. :)
This is also a good time for you to upgrade to solaris 10, since
solaris 8 is almost EOL.