Redundant NFS storage setup

[Nick_Parrott at Dell.com]

MD3000 has a 2TB limit on arrays - so you need to use LVM on top to get
a larger array. This is a limitation (hopefully not for the unit's
lifetime) that is on the 2 x PERC5's in the MD3000. MD1000 will not
suffer this issue using a PERC 5 HBA in the host.

No experience with LVM snapshots, however the MD3000 does have a
snapshot feature, again, 2Tb issue may kick in..

No supported Linux clustering as yet on the MD3000, however it's pretty
easy via the CLI (or a script) to change the LUN ownership paths on the
MD3000 to another host, so the change-over can be pretty quick..

Regards,

Nick

-----Original Message-----
From: linux-poweredge-bounces at dell.com
[mailto:linux-poweredge-bounces at dell.com] On Behalf Of Matthias Saou
Sent: 13 November 2007 12:39
To: linux-poweredge-Lists
Subject: Redundant NFS storage setup

Hi,

I'm looking for the a simple, cheap yet reliable way to have some
redundant NFS storage for a dozen servers. My idea is to go for two
PowerEdge servers running RHEL5 connected to external MD devices. I'd
like to know if anyone had experience with something similar and could
give me some advice.

- PowerEdge 1950 (possibly not "certified" for use with MDs) or 2950
- MD1000 or MD3000 with SATA 750GB or 1TB drives in RAID-10 (14 + spare)
  for low cost, maximum storage space and more than decent read speeds

What I'd like ideally is to load-balance the NFS traffic between the
devices. I've never tried it... I don't even know if it's possible. At
worst, a simple solution of having both servers be accessed through
"floating" IP addresses which would automatically migrate to the other
device in case of a failure might work (I usually use keepalived with
vrrp), and I'd just have to configure half the clients to access one
device and the other half to access the others. Thoughts?

The catch here would be writes, but luckily this setup is to be
read-only, which is why it would be RAID-10 (for performance), so
uploading to only one of the servers, then syncing regularly to the
other would be fine ;-)

I do have a few specific questions about this, though :

- Could this setup benefit from an MD3000 vs. an MD1000? I don't think
so, but again, apart from the redundant SAS controllers, I don't see
any differences between the two (which I might be missing)

- Could I see a single 14TB raw, 7TB usable block device from the RHEL5
server? With SCSI PERCs and the PowerVault devices I have now, it's
impossible to create RAID arrays greater than 2TB, so for 3TB+ I had to
create two RAID arrays and concatenate them using LVM. It would be
great if I could avoid this hack from now on ;-)

- Could I make remote LVM snapshots from one server to the other, or
use any cool LVM features I haven't yet used to synchronize both
servers? :-) Since it'll be to store big files, few files, I can resort
to using rsync regularly from one device to the other, but a more
elegant solution would be welcome.

- Has anyone ever load-balanced NFS somehow? Because it would be a pity
to have two identical storage machines, and have only one take all of
the daily load while the other one gathers dust as the spare ;-)

- Does anyone have experience with GFS? Could it be useful in this
setup to replace NFS?

Any comments are very welcome. Any insight as to how to improve any of
the above too.

Matthias

-- 
Clean custom Red Hat Linux rpm packages : http://freshrpms.net/
Fedora release 7 (Moonshine) - Linux kernel 2.6.23.1-10.fc7
Load : 0.36 0.46 0.43

_______________________________________________
Linux-PowerEdge mailing list
Linux-PowerEdge at dell.com
http://lists.us.dell.com/mailman/listinfo/linux-poweredge
Please read the FAQ at http://lists.us.dell.com/faq