Lots of storage

[ciphergoth]

If I buy one of these: and five of these:

and stick it all together using Linux software RAID, will that work as a relatively cheap (£370) 2TB RAID-5 storage array?

Will it be very noisy, and if so, should I be looking at getting one of these too so I can stick it in another room?



As you might imagine, I'm giving the whole question of backups a lot of thought at the moment...

Comments

[skx.livejournal.com]

It should work as a nice raid system, But if it were me I'd be expecting only 1Tb total.

(md0 = RAID0 on disks 1+2, md1 = RAID0 on disks 3+4, disk 5 as a hot spare.)

In terms of noise I'd assume noisy, just because more than one disk in the same place always seems to be noisy - but I guess it depends what they're plugged into. I've got a couple of desktop machines here and the noise they produce easily masks/exceeds the noise from the external USB drives attached to them.

Having an NSLU2 would be a neat thing to have - but I've not yet used one. I think probably the idea of backups is good but you might want to think about where you're going to place the drives. If they're sat upon the top of a PC in an obvious location then there's nothing stopping them from getting lifted along with the main box ...

[skx.livejournal.com]

For raid0 read raid1.

D'oh.

[ciphergoth.livejournal.com]

I was thinking of RAID-5 rather than RAID-0 - since this is for home rather than business use, I'm not sure it's worth halving the cost efficiency in order to have RAID-0 and hot spares.

Yes, the danger the drive array will be stolen is one thing that makes me think it might be worth placing it elsewhere. However, another option might simply be to bolt it firmly to a sizeable piece of furniture..

[olethros.livejournal.com]

Curious to know why you'd suggest raid 1 and a hot spare over raid5?

[skx.livejournal.com]

The bolting the enclosure might be a good plan, but I'd assume it would still be easy to remove the actual drives from there..

If you keep an eye upon the status, via mdadm, then RAID-5 should be just fine. I'm always a little paranoid when it comes to my data!

[skx.livejournal.com]

It strikes me that when you have a lot of drives in the same physical location a single problem is liable to affect multiple drives at roughly the same time.

For example vibration, heat, flood damage(!), etc, are both going to be shared across neighbour drives - even if those at the "far side" of the enclosure are OK.

To compensate for that I'd arranage the drives (physically) in a pattern like this:

[md0-1][md1-1][spare][md0-2][md0-3]

(The idea is that adjacent pairs of disks come from different RAID sets - and with this arrangement any two contiguous drives can die without a problem.)

For RAID-5 if you have two drives die at the same time you're toast, regardless of location.

[babysimon]

It might be quite noisy. Drives vary in noise quite a bit - when I upgraded the mythtv from 200GB to 500GB it got a lot quieter since I actually looked for a quiet drive, even though the drive was from the same manufacturer.

Why do you need RAID? As I'm sure you know well, RAID is not a backup technology; all it buys you is uptime. Do you need the uptime that badly at home? Out of curiosity, what will you use all 2TB for?

[nikolasco.livejournal.com]

I'd consider putting the RAID 5 layer under a RAID 0, MD-append, or LVM layer for the possiblity of expanding the array; otherwise it's problematic (the tool for expanding RAID 5 is not well tested and it's much more complex by necessity). I'm also not sure I see the point of RAID 5 without a hot (or at least a cold) spare on-hand.

I don't see what the NSLU2 buys you since the array doesn't appear to have a USB connection, just SATA. If you stick something between them, I'd go with a mini-ATX system with gigabit. I couldn't find smaller form-factor that had enough ports, but I also didn't look very hard.

I'd love to recommend Nexenta+ZFS but I've heard of problems with over 1TB of data (scrub/checksum-checking process hanging), even though that was a year ago it still makes me fret.

[nikolasco.livejournal.com]

Not just the same physical location, but also the drivers are from same manufacturer and bought at the same time from the same place... It jus seems likely that once one gives up due to foo, they all will.

At home, I've always tried to have enough slots to have completely empty ones between drives.

[ciphergoth.livejournal.com]

You'd need some sort of SATA/USB adaptor to connect it to the NSLU2.

[mskala.livejournal.com]

I run RAID1 across two drives. It's actually on top of the partitioning - I have both drives partitioned the same, and each pair of corresponding partitions is RAIDed. One pair is RAID0; I use that for large temporary files from video editing.

Although people say "it's not for backup", that's really a big part of what I use it for. I'm more worried about losing data to drive failure than my own mistakes, and having it mirrored across the two drives means I lose almost nothing (instead of "back to the last backup") when a drive fails. So far I've lost one, and was up and running again just a few hours later with no real data loss. I couldn't really make backups at that frequency.

[topbit.livejournal.com]

My home dev-server has a couple of drives softraid-ed and then with LVM on top.

Because it's mostly about web-scripts being edited, I've also got a cron job running 'rsnapshot' (for rsync diffs).

[skx.livejournal.com]

rsnapshot rocks for remote backups.

I've got a server sitting on my floor which takes complete backups of a couple of remote machines and it has saved me more than once.

(Too often I've had backups which exluded too much, so I lost files, nowadays I remotely backup "/" ignoring only /proc, /sys, and /dev. I haven't lost a file I cared about for a couple of years now :)

[aidan-skinner.livejournal.com]

I'd grab the linksys and stick it somewhere out of the way just because.

I've been using Backup Manager for a while, it's got nice integration with a number of upload programs for automatic offsite storage.

[ciphergoth.livejournal.com]

Hmm

* Noise: I'm now thinking the easiest fix is to move it into the spare room. I have most of a PC built and this could complete it.

Having done that, of course, I'd feel bad for having two PCs on 24/7, so I would I think move all the 24/7 functions onto the file server (DNS/DHCP and such, recording TV), buy one of these for about £60 more, then re-purpose Trent as a dual boot desktop and games machine.

* RAID: Hard drives fail often enough that it's worth having a measure in place especially for that.

* 2TB: well, 0.5TB drives are the sweet spot at the moment, and 5 is the minimum for RAID to be economical, plus if I've exhausted the 400GB of storage I have at the moment and I'm planning on doing a lot more backing up, then a small improvement will just be a pain - I want *lots* more storage. Plus I'd like to be able to rip DVDs.

[ciphergoth.livejournal.com]

I was thinking of using backuppc for the backups to the RAID array and brackup for the offsite stuff, but that's a discussion for another time...

[calum]

I've been thinking something similar, but will probably just get a Buffalo Terastation when I get the money together. Cheap, effective, hackable (runs Linux), and designed for running at home.

[pavlos.livejournal.com]

My only suggestion is do include the NAS server, because attaching the array temporarily to your laptop will be too inconveient.

[meta.ath0.com (from livejournal.com)]

For the stuff you really care about, you might want to consider backing up to Amazon S3 at a cost of around 15 ¢ per GB, e.g. via JetS3t.

[ciphergoth.livejournal.com]

My thoughts exactly. I was looking at brackup (as linked above), which can also write to S3 and which gpg-encrypts each backup.

[ciphergoth.livejournal.com]

Hmm, I was tempted by that, and I wondered if it was hackable - cheers! If I build my own, it'll be £300 or so cheaper, and I can put the TV card in it. OTOH if I buy a Terastation it will work...

[ciphergoth.livejournal.com]

I'm now thinking I'll just build a storage PC, which is these components plus the spare PC in the cupboard, and turn Trent into games machine.

[calum]

Twonkymedia runs on the Terastation.. :)

[pengshui-master.livejournal.com]

I wouldn't want to run raid5 without a hotspare, as a un expected second failure will render you data lost.

OTOH, I've seen resync trigger drive failures when a drive is marginal.

In your place I'd configure md0 [sda sdb sdc sdd] spare:sde. Which will give you 3xHd disk size.

Since SATA hotswap support seems limited on most of the SATA controllers I've seen . I've tried hotswapping SATA devices before now , and the experience was overall less than satisfactory. It sort of worked but a rebooted tended to be needed anyway.

I suspect this has better reliability than a dual raid1 setup. But I'm too hot to check my gut feeling with hard reasoning.

[alexmc.livejournal.com]

please keep us informed of how you get on. I'm interested as well. (My machine I have here is a full sized tower with four SATA drives in it right now plus the one PATA drive.

[juliet]

Definitely agreed re RAID5 & a hotspare. Especially given that it is not uncommon IME for more than one drive to fail at close to the same time (for fairly obvious reasons given that you've bought 'em all at the same time).

I'd certainly recommend either RAID5 + hotspare or mirror-RAID over stripe-RAID, because even if it is only backup, sod's law says...

[juliet]

There's also rsync.net who I've had an account with for a year or so for important personal stuff.

Other option for backup: Bacula (which is what I use at work; not entirely sure that it's worth it for a v small home setup, *but* it is very easily expandable if/when you get more machines).

[skibbley.livejournal.com]

At work I've been using Buffalo Terastation Pro II 1 and 2TB NAS boxes and have been happy with them as Gigabit connected ftp servers running SATA RAID. Not cheap though, and you probably don't need them to be as portable as we did. Not too noisy.

[skibbley.livejournal.com]

You may also wish to consider RAID6 - our bigger systems use that.

[damerell.livejournal.com]

RAID isn't backups. One fire, one burglary, one flood, one faithful recording by the RAID of an inadvertent rm -rf, and the whole thing's gone.

If I were you I'd get three big discs and enclosures for same with the money and arrange for at least one to be offsite at any given time.

[shevek.livejournal.com]

The image of the card has broken so I don't know which card you are considering.

I am likely to be disposing of two 8-port SATA Adaptec 2810SA cards soon, since I have bought a 16-port 3ware 9550SX to replace it. Also, you pretty much have to run raid6 not raid5 on disks that size - the rebuild stresses the disk, and if you have had something cause one disk to fail, the likelihood of a second disk failing caused by the stress of the rebuild is even higher. I'm reconfiguring to 16 disks, 14 in a raid6 plus two hot-spares. I used to run 7-disk raid5 plus one hot spare, and it wasn't enough under the circumstances I describe (which was mostly domestic use).

Linux software raid does win, I've been using it for years, but it is not (as I have recently discovered) a substitute for backup. I recommend LTO drives, they're very practical and realistic. Alternatively, the latest generation of DLT is also quite realistic.

The throughput of cards varies dramatically. The best is the 3ware card, but they cost around 500 quid. They do transfer 300Mb/sec off 7 disks, I have yet to try off the full 14, and I wrote a kernel patch to use the card's buffering to the fullest (assuming you have no other controllers in the system).

Um. Feel free to ask if I can help more.