Unfortunately the USB HDDs I have tried (Hitachi and Transcend) cannot be monitored with smartd even though the HDD model within the enclosure is SMART capable. Presumably the designers decided the USB firmware would not pass SMART commands from bus to device. Is Firewire better in this respect?

I think you are confusing this with the MTBF (Mean Time Between Failures). The MTBF is in no way a guarantee, but a term of reference how long a drive runs on average. That means, even with MTBF of 1,000,000 hours your drive can die one day after you have bought it.

FSCK has now been running for over 72 hours. I know it's a relatively big drive but I don't think it's a good sign. I fear I might not be able to recover anything. Is there anything else I can do after fsck finishes (and there's nothing in lost+found)? I mean without giving the drive to some specialist recovery company?

Thanks for the enlightenment. That's what I get for trying to sound smart I suppose.

Use testdisk or foremost to recover files.

The best speed is the lowest speed. I've never had any coasters burning at lowest speed. I have had coasters at higher speeds.

Everyone is giving you recommendations of this brand or that brand, and those recommendations conflict.

No surprise. The answer isn't that simple. To the best of my ability to tell, and I have researched it extensively and I continue to revisit the topic periodically, the right answer is that all manufacturers have at times put out series of drives that were prone to failure. All manufacturers have also put out series of drives that proved to be reliable over time.

The best way to approach it, therefore, is to do research on specific drives that you are considering purchasing. For instance, in the fairly recent past I was considering a particular Samsung SATA 2 TB drive because of price and availability. However, a bit of googling showed that there seemed to be a lot of failures of that drive. So I passed. I then investigated some Seagate drives and there were two series commonly seen in the field. One series seemed to be reliable, the other wasn't. I decided I couldn't easily tell which was which when I was in the store shopping, so I passed. I finally bought a WD green series drive because it apparently was reliable, based on my reading, although it wouldn't be acceptable for a system drive due to the way it continually loaded the heads in that role. Since I wanted it for a data drive, I didn't worry about that too much.

Similarly, a year ago I had a SCSI drive fail. This was a system drive and rather critical to me. I replaced it with a Hitachi SCSI drive, which failed within 72 hours. The warranty replacement for that drive from the vendor was the same series. It failed after 92 days. I purchased a Fujitsu SCSI drive as a replacement this time, and it is still working. The 92-day old failed Hitachi was RMA'd to Hitachi, and their replacement was a different series. I installed it as a data drive and it is still working.

The point is that you need to do your homework on a per-series basis. If you buy a brand new drive of a new series, you will be the one experimenting to find if it is reliable or not. Past performance by a manufacturer is no guarantee of future performance - for either good or ill.

Thanks. Very true. Still waiting for fsck to finish.

It might never finish .

I've gone to external drives, I got to the point that I have so much data that anything else just isn't practical. We bought our first digital camera in 1999 and now have over 60G of pictures, then add in mp3's've purchased, and documents I've created, I have close to 100G of data. I buy the drive then an enclosure, that way I know exactly what drive is inside.
I have a pair that i use and alternate between them and both are kept in a safe.

External drives here too. Though I don't use USB or Firewire. My motherboard's sATA controller supports hot swapping, so I bought a few enclosures that sit in a 5 1/2" bay and have a removable drive tray. They were only about £4 a piece so well worth it just to keep the cables away. Obviously, I bought one of these per backup drive, but only fitted one in the machine and just use the trays from the others.

Thanks for everyone's input. Fsck has been running for a week now. It's definitely not frozen as I can see the blocks/inodes changing regularly. Once it's finished I'm going to run the tools that h_texmex_h recommended.

I'm getting more and more frustrated as I had some work-related documents on that drive and my computer notes that took me quite a long time to create. Well, the only person to blame is myself.

As per your recommendations I've decided to buy two SATA disks (+ enclosures) as well as a pack of archival grade DVDs.

rsync is going to backup data to discs on a daily basis, while I'm going to burn a DVD when there's enough data for one (every couple of months)

I've found some reviews/comparisons of eSATA vs USB3.0 and it seems that USB3.0 is the way to go.

USB 3.0 support is in the kernel, but last time I checked it was still marked experimental. However, I'm sure it'll probably work anyway.

The other problem is that it needs a sufficiently quick drive to utilise its speed. I've seen a review film where the speed of transferring a file was initially very good but after a few seconds dropped due to insufficient number of RPMs.

WOW - fsck has finished its job (over a week) and recovered most of my data (199GB) - now it's a question of copying and sorting it. It's not as bad as I thought!!!!

I'm waiting for the new disks to arrive, but there's a question of the old drive. Can I trust it again?

After retrieving data, I say run 'smartctl -t long /dev/sda'. Then wait and check the results. Check the attributes too after it finishes.