2

I'm looking to buy a NAS unit for file storage on my network. It doesn't need to be massively fast, but high stability for a low price point is the ideal.

I have a maximum budget of about AUD$1000, including the disks. I can get 1TB disks for about AUD$100 each, though if the disks can included with the machine, all the better. Also, as is obvious by the pricing, being available in Australia is also a good idea.

I need the system to hold at least two disks, and be RAID capable (Probably RAID 1 for two disks, or RAID 5 for more). The storage capacity needs to be at least 500GB per disk, preferably 1TB. It also needs to be NFS capable, because that's how the machine will be accessed from the rest of the network.

We've tried building setting up our own NFS machines before, and they've always fallen over. So we're going with an all-in-one box this time.

Does anyone have any recommendations for machines that would fit my requirements?

3
  • Please expand on what you mean when you say 'fallen over', are you referring to hardware failures, or what? What component of the systems are failing?
    – Zoredache
    Aug 25, 2009 at 21:37
  • According to the techy (I'm really not), the NFS shares keep going stale...
    – Margaret
    Aug 25, 2009 at 23:07
  • @Margaret, an iSCSI solution with a cluster file system like OCFS2 is preferrable
    – HTTP500
    Aug 26, 2009 at 2:00

3 Answers 3

1

if your stale NFS handle problems are the same as the ones i've been having intermittently over the last few months, then there was a patch to fix that by Cristoph Hellwig released about two weeks ago. it's not yet in 2.6.30.5 but it should be in 2.6.30.6 (it's already in the 2.6.31 release candidate).

if you (or your sysadmin) is comfortable with patching and compiling the kernel, there IS a patch available for 2.6.30.5. i have applied it and recompiled and it works.

from what i've read, the problem was introduced around 2.6.29.x, and it is specific to sharing XFS file systems over NFS. the bug is actually in the XFS code rather than the NFS code.

you can identify the problem by searching your kernel log for lines like:

"reconnect_path: npd != pd"

you can read about the problem and the fix at http://bugzilla.kernel.org/show_bug.cgi?id=13375

(oddly enough, i never got this problem on my home network where i use NFS for very high bandwidth recording, playing, and transcoding of MythTV shows....but i got it quite often at work where NFS is used fairly lightly to export virtual host directories from the file server to the webservers. both running similar kernel versions and with XFS as the filesystem)

0
5

You could buy an average beige box and stock it with a number of drives and install OpenFiler or FreeNAS, set them up to do software RAID, and that way you'd save by not spending for hardware RAID and could still get a decent network card for it. OS is free, administrate it with a web interface, and since it's off-the-shelf you can easily maintain it.

Just make sure you have a good backup solution in place since; RAID isn't a backup.

4
  • The problem with this is, as I said, we've been going the 'beige box' route previously, and they always seem to fall over. I don't know whether we don't fully understand the system or what, but we've had major stability issues with the last couple we've had.
    – Margaret
    Aug 25, 2009 at 21:27
  • 1
    What do you mean by "fallen over"? If you had a bad memory stick or a bad motherboard or some other component failure, that is a possibility no matter what. You can make a router out of a PC or buy a router; I've had Cisco routers go bad and I've had Netgears at home go bad. I've had PCs run for five or six years, others dead on arrival. Depends on what you had fail. In a beige box you can at least replace parts instead of losing the whole unit, and OpenFiler/FreeNAS are turn-key interfaces to configure it. Aug 25, 2009 at 21:41
  • According to the techy (which I'm really not), the NFS shares keep going stale...
    – Margaret
    Aug 25, 2009 at 23:10
  • You mean you have a server on which 2 systems are accessing the same file, and machine A modifies or removes a file that system B is using? That's outlined here: cyberciti.biz/tips/… It may be that it's not necessarily something a different NFS server will solve in that case...it's an NFS thing. Aug 25, 2009 at 23:25
2

I don't know whether it's the best possible option, but I've got a D-Link DNS-323 that I installed Debian onto, and it's running as the NFS server for my house. I haven't benchmarked it, but it's got enough grunt to serve music and movies, as well as do backups onto.

5
  • Cool, did you use this?: cyrius.com/debian/orion/d-link/dns-323 Have you tried turning it into an iSCSI target/server? You might have given me an idea for a home project, thanks!
    – HTTP500
    Aug 25, 2009 at 21:54
  • I have exactly that device, but I haven't bothered with Debian, just optware. Some of the built-in software is junk, like the Xbox compatible uPNP media sharing. I haven't tried NFS, but the throughput is basically crap. A newer Debian kernel might fix that though.
    – jldugger
    Aug 25, 2009 at 22:24
  • 1
    There's also apparently continuing work on the firmware to add things like BT and maybe even NFS: forums.dlink.com/index.php?board=251.0. It seems 1.07 has been released since I last looked, and 1.08 is in beta.
    – jldugger
    Aug 25, 2009 at 22:30
  • @jldugger Right you are, looks like their 1.08 firmware is adding an NFS server. forums.dlink.com/index.php?topic=5485.0 39. Add NFS Server. OP should try this firmware in a lab
    – HTTP500
    Aug 25, 2009 at 22:37
  • @Jason: I was the person who did that port. @jldugger: Yes, a newer kernel makes a big difference to throughput.
    – womble
    Aug 25, 2009 at 23:17

Not the answer you're looking for? Browse other questions tagged .