Re: Slurpd vs. Syncrepl

[Samuel Tran <stran@amnh.org>]

On Mon, 2005-10-31 at 11:13 -0800, Quanah Gibson-Mount wrote:
> 
> --On Monday, October 31, 2005 6:19 PM +0200 Chen Shapira 
> <cshapira@mercury.com> wrote:
> 
> > Hi,
> >
> >
> >
> > I'm using Slurpd for openldap replication in order to have master-slave
> > high availability in our production environment.
> >
> > I saw Syncrepl mentioned on this list as another method to have a replica
> > of the directory. I've read about Syncrepl and how it works, but some
> > parts of the picture are still missing:
> >
> >
> >
> > 1. Is there any reason to change from Slurpd to Syncrepl? Syncrepl has a
> > much more complicated protocol, but in what ways is it preferable to
> > Slurpd?
> 
> Syncrepl allows a slave to catch up from a given point in time to the 
> master, without the master having to initiate anything.  If the slave is 
> too far out of date, it will even completely reload itself.  With slurpd, 
> you have to suffer a continually growing replication log while a slave is 
> offline.
> 
> 
> > 2. Are there any good reasons or situations where I should not use
> > Syncrepl?
> 
> A heavy write environment.   The current implementation of syncrepl only 
> does complete entry replacement, rather than doing change delta's to the 
> existing entry.  This will hopefully be fixed in 2.3.12 with the 
> introduction of delta-syncrepl.
> 
> 
> > 3. Does Syncrepl overcome any of Slurpd limitations? Can I have two
> > servers each replicating the other, so I can have a multi-master
> > environment with much easier failover?
> 
> See #1, yes, it does, for your first question.  On the second part, I 
> believe some people have been doing things like that.
> 

I have a small question regarding syncrepl: are there some kind of logs
that show the replication talking place, the failures, ...?

Thanks.

--
Sam