[Linux-cluster] Redundant (multipath)/Replicating NFS

Gordan Bobic gordan at bobich.net
Fri Feb 15 20:47:37 UTC 2008


isplist at logicore.net wrote:
> I recall that there is a multipath tool for RHEL which allows one to have 
> redundant SCSI/FC paths in case one goes away. 
> 
> Are there any tools which would accomplish the same with NFS?

If you have 2 NFS server you need replicated and failed over, you can 
use DRBD to replicate a virtual block device. You can have this in 
primary/primary mode with GFS on top, and you can then export the FS 
from both machines via NFS. If you use UDP NFS, you can have a floating 
IP address for the server that you fail over in case of a node failing. 
There may be a way to load balance NFS, too, but I've yet to see that 
actually work. If you are just failing over, you could run DRBD in 
standard mode without GFS. DRBD has heartbeat plugs.

> Is there any way of replicating NFS data in real time rather than as a regular 
> backup, such as RAID1 for example, and be able to fail over to the second path 
> or NFS IP should the first fail.

DRBD is effectively RAID1 over TCP/IP.

Gordan




More information about the Linux-cluster mailing list