[Ocfs2-users] initial cluster setup

[Sunil Mushran]

Matthew Boeckman wrote:
> Thanks to all the replies on this topic, I'm getting it slowly!
>
> I originally did have both systems using local disks. Now what I have is 
> node0 has a directly connected IDE drive that I have iscsi exported to 
> both node0 (via localhost) and node1 (via 10.1.1.100). Both systems now 
> see /dev/sda. I think this is what I need to have, correct?
>   
In principle, yes.
> So now my question is procedural - what's the next step? I have 
> formatted /dev/sda with ocfs2console on node1... do I have to do the 
> same thing on node0?
>   
As it is a shared disk, the volume needs to be formatted from any one node.

> Perhaps I should have started with what I'm trying to accomplish, it's 
> possible I'm in entirely the wrong place:
>
> One or more SCSI RAID systems directly attached to Linux heads that are 
> capable of sharing storage to a dozen clients via gigabit etherenet. NFS 
> is how we do it now, it's slow and I'm looking to replace it. The 
> clients each need to access the volume, but do not need to access common 
> directories (server a accesses /volume/client1, /volume/client2, server 
> b accesses/volume/client3, client4, etc).
>
> Am I on the right path, or down a rabbit hole with ocfs2?
>
>   
If your shared device is also accessible via the gige (iscsi), you 
should not expect any
better performance on ocfs2 than nfs. You will probably get better 
performace
with nfs, because of the latter's propensity to cut corners.