Re: slurpd vs single data store

[Frank Swasey <Frank.Swasey@uvm.edu>]

Today at 8:09am, David Koopman wrote:

> I am debating whether to use slurpd to replicate a
> database for redundancy or to run two openLDAP server
> machines, both sharing a mount on the NFS to the same
> data store directory.

Are you thinking they can both be a master and somehow update the NFS
data without getting in each others way?  Not a chance!

What backend are you going to use?  If ldbm, what dbm?

What is the purpose of having multiple servers?  Will your clients send
requests to both, or is the second just there in case the first one
fails?

> Are there known problems with
> doing this?  I would prefer to share the same data
> with both servers through the NFS, but I wonder if
> they will have problems locking files and crashing
> into each other. Any thoughts are much appreciated, I
> cannot seem to find any answers or discussions along
> these lines.

I have just finished setting up a pair of machines that use a cross-over
cable between them for heartbeat and slurpd replication.  The first
machine is the master and the second machine the slave.  The entire
purpose of the slave is availability.  If the master fails, the slave
takes over the address.  It doesn't change its configuration, so updates
will fail until the master is restored (I may extend this [in the
future] to allow the slave to become the master if the true master fails
frequently)

I have not seen any discussion of using NFS to allow multiple servers to
access the same database files -- I would expect there to be a LOT of
problems with that.  Especially since there is no way for the server
doing the updating to communicate to the other server that updates have
been made to the files.

Replication is easily set up and works reliably, unless the human
involved messes up... Grab Adam Williams' ldapv3.pdf file and read up on
it.

F