Having built a RAIDZ array I started benchmarking with some large files copies. The copies were being done while I was SSH’ed in from home but seemed to stop in the middle of testing - I couldn’t SSH in again and sure enough, when I checked in the morning the server had hung. Giving it the benefit of the doubt I restarted the server and rebuilt the array as RAIDZ2. Again, the server hung in the middle of a file transfer. By this time (bearing in mind the hours spent trying to get OpenSolaris just to boot on the ProLiant), I’d lost all patience with OpenSolaris and put Ubuntu on the new server.

Not two hours into some more benchmarking the server hung again. I did some searching and found out there was a small possibility that a recent Linux kernel bug was the problem here (which I won’t describe to prevent Google from suggesting there’s any correlation) . The bug was definitely occurring in the storage system. Was it possible that something similar yet unrelated was causing a similar problem in the Solaris kernel? Yes it was possible, but pretty unlikely. However, I installed the only older Linux distro I had kicking around - CentOS 5.2. Sure enough, it hung in the middle of some transfers but this time the logs were showing two different smartd failures on both of the brand new disks we’d bought to fill up the vacant slots on the server. Sigh. Hardware all along…

So after all of that, chances are that the Solaris install was probably OK after all. My copy of the OpenSolaris bible finally arrived, so now I’ve got some reference material to work with I might five it another shot - third time’s a charm?

The holiday cottages have a website - I made changes to its zone file during a migration last month, in fact. It’s always great to see little bits of your daily grind pop up on television.

A while back we got hold of a QNAP box in work for throwing backups on to. It wasn’t without it’s faults, but it did the job - it exported NFS shares and had space left over for the occasional request of temporary disk space. It had 4 x 500GB disks, which at the time was pretty big. The disks were running in a RAID-5 configuration, so I was quite happy in my naiveté that when the eventual disk failure occurred we’d be able to handle it. The disk failures will no doubt still arrive, but they’ve been pre-empted by a failure of the processor fan. A fan I couldn’t find a replacement for.

Rather than buy a new NAS box, I had the great idea to buying a standard server we could replace all the parts of. We ordered an HP ProLiant DL 160 G5, a box that I noticed later on was on the Solaris HCL. Fantastic - I could put Solaris on there and and get all that ZFS-sy goodness you want in a storage box. In theory, anyway…

Solaris wouldn’t install. I tried Solaris 10U5, Solaris 10U6, OpenSolaris 2008.5, a few betas and eventually 2008.11. I even tried the (unsupported) Solaris Express, which wouldn’t even boot from CD. I asked dumb questions on #opensolaris (and to be fair, received some good suggestions). I Googled and found out about installation turds and devid patches. None of it worked - the furthest I ever got was the server hanging at the hostname report (although I got very used a small red line at the bottom left of the screen that I still don’t know the significance of).

Eventually I gave up. Stuck an 8GB USB drive into the box and installed Ubuntu Server 8.10. I don’t have ZFS, but I do have a RAID-10 volume that I’m happy with (even if Steve Cassidy isn’t - he railed against this sort of config in the February PC Pro). With any luck, BTRFS will be along soon (although I’d still love to see ZFS on Linux and have a file system common to  Linux, FreeBSD, OS X and Solaris).