|
At 22:31 14/11/2000 +0000, you wrote:
Hi,
This because, obviously, if the server that holds my NFS disk crashes or
goes down, the whole LVS will be down !
In fact, I thought about NFS rather than GFS , just because i do not have
enought time to test it !
This is true, you will have a single point of failure at the NFS server,
but then you also have a single point of failure at the director (unless
you have failover). It can require some very careful planning to ensure
that you have a useable environment that doesn't have any single failure
points.
What I would like to know is what are you using, experiment LVS users, to
share hard disks between real servers ? NFS ?
Having said that, I currently run an LVS compute cluster. This has two IRIX
origin 200 NFS file servers which share out user file stores (50% each) and
shared software which is exported by both machines. These have a private
presence inside an NAT LVS cluster which is connected with a 100Mbit switch
(to ensure good machine to machine performance). This works wonderfully and
as the shared software is on both servers if one goes down you only lose
50% of peoples filestores, not access to all the software. IRIX servers are
very stable NFS servers, we have very few problems even though they
maintain approximately 1300 user file stores which are exported to
potentially hundreds of machines.
well , this is not the proper idea of this mailling list (I know), but since
you were talking about NFS ...
If you are using a NAT cluster so of the NFS security issues become less
important as you can limit you NFS servers to only export inside the
private NAT cluster. This means that people have to break your cluster
before they can get to the NFS servers.
This is how my cluster is set up - as you wanted to know.
Steve.
----------------------------------------------------------------------------
Going to church doesn't make you a Christian any more than going to a garage
makes you a mechanic.
|
