[wplug] NFS Availability Issues -- more info on your network ...

[Brandon Poyner]

On 9/11/07, Michael Semcheski <mhsemcheski at gmail.com> wrote:

> I'm alright with increasing the complexity.  But I imagine if we do
> change things up, it will be to go to AFS, which from what I can tell
> has better availability.  (through local caching.)

If you're willing to consider increasing the complexity, I'll throw
out the suggestion of DRBD + Linux HA.  I use them for high
availability on the cheap and it works well.

DRBD is essentially a kernel module that behaves somewhat like RAID 1
over the network, mirroring block devices.  It isn't in the mainstream
Linux kernel at this time but they are working on that very issue.

Linux HA is the userland utility to keep the services running on the
correct server(s).

DRBD has never let me down, but Linux HA has once in 4 years.  The
latest version of DRBD can actually permit both servers to write to
the block device at the same time and some people use clustering
filesystems on it (almost like cheap SAN disk storage, only without
some of the best features of a SAN).

http://www.howtoforge.com/high_availability_nfs_drbd_heartbeat
http://www.linux-ha.org/
http://www.drbd.org/

-- 
Brandon
Kiva.org - Make a small loan, Make a big difference