[prev in list] [next in list] [prev in thread] [next in thread] 

List:       ssic-linux-users
Subject:    Re: [SSI-users] Distributed Root FS?
From:       "Brian J. Watson" <Brian.J.Watson () hp ! com>
Date:       2004-05-25 19:31:35
Message-ID: 40B39F17.4 () hp ! com
[Download RAW message or body]

Adrian Fletcher wrote:
> Hi,
> 
> Is it possible to have an OpenSSI cluster that can run its root fs and  
> home fs on a distributed filesystem?
> 
> I am basically wondering if this is possible:
> 
> I have a LAN with identical machines on it, and I would like to run 
> them  as an OpenSSI cluster. However, I would like that the machine be 
> HA for  everything, including the filesystem, such that if any node dies 
> due to  hardware failure or whatever, the others keep running without 
> any  interruption (except the processes that were running on that node 
> of  course). The distributed file system should be able to run on the 
> nodes in  the cluster with all of them contributing disk space and each 
> have a  private swap partition also - without a metadata server or lock 
> manager or  something outside the cluster as a cental point of failure. 
> The DFS should  also be able to cope with any node failing, and thus 
> keep 2 (or 3?) copies  of everything across all the disks and make 
> another copy as soon as one  fails.
> 
> I know this is an extremely complicated taks, but perhaps someone has 
> done  it or at least got some experimental attempt that I could hack at?

It is very complicated, and I don't know of anything exactly like what 
you describe. Something similar, however, might be using DRBD to mirror 
each filesystem between two nodes, so that the failure of any one won't 
disrupt access to the filesystem that it was serving. For example, you 
could have nodes 1 and 2 mirroring the root, nodes 2 and 3 mirroring 
/home, and nodes 3 and 1 mirroring /var.

This approach uses CFS, which only allows one node to be the server for 
a given filesystem. If you want the scalability of a distributed 
filesystem, then Lustre might be what you're interested in. It stripes 
data across multiple file servers. I think it also can be configured to 
do HA mirroring. The only drawback currently is that your Lustre file 
server nodes must be outside the OpenSSI cluster that uses it. Of 
course, you might be able to configure the Lustre file servers to be in 
an OpenSSI cluster of their own, for easier management:

OpenSSI cluster A
-----------------
Lustre_server_1 .. Lustre_server_2 .. Lustre_server_3 ..
HA-CFS/DRBD root for Cluster A mirrored between nodes 1 and 2
The remainder of the nodes' internal disk space used for Lustre 
filesystem(s), to be mounted by Cluster B

OpenSSI cluster B (Lustre client)
---------------------------------
Node_1 .. Node_2 .. Node_3 ..
Either HA-CFS/DRBD root or Lustre root
Other filesystems may also be Lustre

Hope this helps,

Brian



-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. 
Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click
_______________________________________________
Ssic-linux-users mailing list
Ssic-linux-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/ssic-linux-users
[prev in list] [next in list] [prev in thread] [next in thread]