Nibble:- Sometimes it's the Backup Media which Fails

If you haven't encountered an unrecoverable error in your backup system then you probably aren't doing enough backups.

I've lost count of the number of times this had happened to me over the last 30 years with media as diverse as hard disk, tape, optical, floppy, EPROM, punched cards and paper tape. Yes I did say 30 years - and in the early days of the microprocessor when the 8080 was still a hot product - then many of us used paper tape.

But it doesn't matter how long you've been doing backups it still comes as a surprise when it's the backup system itself which develops a fault. I was reminded of this last week when it happened to me. The strange thing is it took me hours and several iterations to figure out what was going on, even though I had had an almost identical problem about 3 years before.

It was late - so when I set off a new complete system backup on a rotated hard drive I didn't bother to hang around to see how it would end. (That's unusual. Mostly I never leave the office without a backup.)

Next morning it looked like it had completed - but the report message was different. Looking at the directory on the backup drive - there certainly was a new file there and it was pretty big - although maybe smaller than I had expected. I concluded that when I set off the backup the previous night - I had been tired and might have accidentally set up a parameter differently. So I set it off again and got back to work.

This time the report was the same, but when I looked at the backup drive directly the backup size was zero bytes. Had I accidentally set up an incremental backup? It looked weird - but still like an operator error (mea culpa). So I decided to just back up a small subdirectory which would only take a few minutes and this time watch the screen to see what was going on.

That was worth it - because I got a transient message saying "disk full" - even though this wasn't mentioned in the saved report.

That explains it - I thought. So I went to have a closer look. Maybe the backup disk was full - and that was the original problem. So I deleted a couple of older backup rotations from the disk from about 13 months ago and restarted yet another backup - this time smaller.

But that failed too.

It took me a few more minutes to realize that whereas I could read quite clearly from the backup disk - I got problems whenever I tried to write to it. Swapping out for another drive isolated the fault. It wasn't in my PC - but in the media. Then I remembered about the flaky disk, one of a batch of 4 which had caused me so many problems about 3 years before. It was definitely a media error. Why hadn't I spotted it quicker? This was just the same thing at the other end of the bathtub curve. Silly me!

Anyway - I stuck a label on the offending drive and put it in a cupboard for future destruction (with a sledgehammer or log splitter in case you're interested in the messy details) and went back to doing another complete (this time successful) backup on another disk.

Which brings me to another point. You actually need at least 3 sets of easily retrievable media which you rotate between for your backups. That's in addition to anything you have as your off site backup. Most people think you just need 2 - but that's a recipe for disaster.

I was lucky this time because I already had a good backup on another disk from the day before, and a backup on the web. But what if the fault had been in my PC and not in the backup drive? You can easily trash one set of media (be it tape or disk) by an electrical or other system fault. Then when you try to isolate the problem by switching to another set of media you will trash a second set - before you realize that the common element is the base system. If you've only got 2 sets of backup media which are conveniently locatable (but not in the same building) then you're going to have a nervous phase you go through before you get back into a safe situation again.

Oh yes and make sure that in your spare sets of backups you've also got all the cables you need, and the installation disks for the media and the software just in case the backup software goes bust just when you need to do a restore. And if you use hardware encryption make sure you've got spares of those systems too.

My own backup strategy has two parts:- the convenient and the disaster recovery.

In convenient mode - if the worst happens I can just get a new working notebook which has already been set up with all my applications from a fireproof safe in a different building and then restore from whichever backup is most easily accessible. The convenient strategy hit a snag last summer due to wasps nests appearing right on top of one of the backups. But this year I've been wacking the new nests while they're still small enough not to retaliate in force.

In disaster mode - I have what I call my credit card recovery situation. In which I use my credit card to buy a new system from the nearest store - and have to be sure that my stash of backups can reinstall critical applications and data in less than 8 hours. It's good to test your own backup regime from time to time - but real life is always more perplexing and takes you longer to deal with than the planned tests.

Serial Attached SCSI - Delivering Flexibility to the Data Center - article by LSI Logic and Maxtor

If you think you already know SAS because you know SATA and traditional SCSI then think again. Sometimes disruptive technologies wear an unassuming disguise. In fiction, Clark Kent, Frodo Baggins and Buffy Summers at first seem harmless, but we see them change into Superman, the Ring Bearer and the Slayer.

SAS too comes cloaked in plain garb - with a physical layer which looks a lot like SATA. But like the Incredible Hulk there are muscles rippling under that shirt - and you would be wrong to dismiss SAS so lightly. There's a lot more inside this interface than it says on the box as this informative article reveals.

...read the article, ...LSI Logic profile, ...Maxtor profile, Serial Attached SCSI