[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: Syncrepl question

On Friday 29 August 2008 17:59:39 Ivan Ordonez wrote:
> Hi all,
> We have a small size domain with about 500 users and computers.  We are
> using Samba with Openldap integration to authenticate user login and
> file sharing.  Our setup is consist of 3 servers running Gentoo Linux -
> 1PDC and 2BDCs.  As for replication, we are still using "slurpd".  Any
> changes or modification is done through the PDC which replicates the
> changes to BDC1, then from BDC1, it then goes down to BDC2 - it's like a
> chain.

Why? Why not have PDC replicate to BDC1 and BDC2?

You didn't provide your OpenLDAP version (which is quite important in 
discussing the options).

> We want to start using "syncrepl" soon as a way to replicate our
> database but I'm not sure were to start.  We want to setup all of our
> machine to sync with each other everyday,

Only once a day?? Syncrepl in RefreshAndPersist will be immediate (as or more 
immediate than slurpd, with fewer problems regarding resyncing etc.).

> and not worry which machine is
> use to make changes, modification, etc....  

Changes done via what? Samba has referral chasing, so you don't need multi-
master to have a BDC be able to write to OpenLDAP.

> I'm not sure which syncrepl
> function to use to achieve what we want to do.  Is "N-Way Multi Master
> replication" the correct choice to do this?

It's not a requirement, and it may be excessive complexity if you just want 
replication to work.

> We are using "BDB" database
> on each servers, and would like to achieve this with minimal downtime if
> possible.  What is the best way to do this?

Configure syncrepl on the provider and the two consumers, assuming you have a 
recent enough OpenLDAP version, and restart ...

Configuration of syncrepl does require at least:
1)syncprov overlay configured on provider
2)limits and access statements on provider to ensure DNs used by consumers can 
get all entries/attributes
3)entryUUID and entryCSN attributes indexed on all servers
4)syncrepl statements in consumer configuration

(so, your restart may be a stop;slapindex;start for the indexes, but your 
database is small)