So you boot up your computer one day, and as the SCSI card is polling your drives you receive the error "START UNIT REQUEST FAILED" and that hard drive is no longer available.

What does it mean?
Unlike IDE drives, SCSI drives don't immediately power up when the system is turned on. What happens is that the drive control is passed from the system BIOS to the SCSI bios. The SCSI bios then sends a "Start Unit Request". The start unit request tells the SCSI drive to turn on an spin up. If you listen closely, you can hear when it does this as your HDD will begin to whine and do some self-diagnostics. When this is done, the drive then sends "Start Unit Request Completed" to the SCSI bios, passing along some other information with it (Drive type, speed, interface, etc). If the SCSI bios doesn't receive the completed command after a certain period of time, it tells you via the message "START UNIT REQUEST FAILED".

Is that a bad thing?
Usually. It happens a lot on hard drives that are in a data center environment or otherwise in a system that sees a lot of uptime. Even though SCSI drives are designed to higher heat and use tolerances then IDE, the circuits inside, for the most part, are not. If a drive gets hot (as SCSI drives tend to do) it usually will operate quite fine. The problem is when the machine is shutdown and allowed to cool. After so many cycles of this, the tiny inter-connects of the circuit boards can fail due to the stress involved and all of a sudden, you have that error. In a datacenter environment, the temperature differences can be quite higher then in a normal home, as datacenters usually run as cool as possible. When the machines heating up the datacenter are taken away (as in, a lot of them get shutdown) but the cooling remains the same the circuits can go from 150F to 60F quite fast, shocking the boards and rendering the drive useless. But this can also happen overtime anywhere.

So I'm screwed then, there's no fix?
Not necessarily. Like I said, it's usually the circuitry that gts damaged, not the platters inside that hold the data. You can try a few things to get it working but the odds of it working are probably around 5% for all but the data recovery option.

• Set the drive to auto-start - most SCSI HDDs have a jumper setting that allows the drive to auto-start, or auto-start after a delay. Auto-starting the drive means that as soon as it receives power it begins it's start-up procedure, it doesn't wait for the SCSI BIOS to tell it to. This might allow your drive to run long enough for you to copy data off of it and into a safe place.
• Disable "Send Start Unit Request" - On all but the lowest SCSI cards, you can tell it to not send the start request unit. That, combined with the above, may be just enough to get your drive online long enough to make some backups
• Major Surgery - Because it's mainly the circuitry, you may be able to fix your drive by removing the boards on another SCSI drive that is virtually an exact copy of the one you have now, right down to the BIOS revision. This is a long-shot, but it has been known to work in the past.
• Quick Fixes - There are a lot of quick fixes out there that have been known to get bad drives going again long enough to suck the data off of them. The big three are Freezing the drive for 5 minutes or so, Baking the drive in a 200 degree oven for 5 minutes and tapping on it with a small hammer. These are usually fixes used for stuck platters, but may work in this scenario as well.
• Professional Data Recovery Service - Your best hope of getting data off of your drive is via a data recovery service. It's very expensive to do so, and is probably not the best alternative for you unless your drive contains the secret of life. (And if it did contain the secret of life, then you made a backup of it anyway, right?)

Can I prevent this from happening?
Yes. The biggest reason these drives fail is due to the massive amount of heat they make the inadequate removal of that heat in most cases. So by far the cheapest and best insurance you can get is some hard drive cooling kits. The beefier they are, the better. They will help whisk away the heat your drives are producing to allow them to operate at peak efficiency. And not only that, having a cooler drive will reduce the thermal shock that occurs from shutting your system down, further increasing the life of your drive and ensuring many years of trouble-free use. Since I started using drive coolers some time ago, I have yet to have one fail on me due to a heat issue. Spending 10$-30$ on a hard drive that runs for 300$+ is worth it in my book!