up vote 1 down vote favorite
share [g+] share [fb]

Early this year, I installed Ubunto 10.4 on a 10 year old Dell tower to use as my home file server. It was working great, I even had a good backup system going using rsnapshot and an external drive.

Then, three months ago, I woke up one morning to find that it wasn't available on the network. Checking it, I saw that it was stuck in a boot process and said that it could not find the boot disk.

I spent a bit of time trying to diagnose the problem. Testdisk reported thousands of error messages just trying to read the drive. A last-ditch attempt to just reinstall Ubuntu ended with failures to write a decent partition table to the drive. I swear that I even heard some odd clicking noises coming from the drive (this may have been my stressed imagination). Needless to say, I assumed that this 10 year old hard drive was now a brick. I was very, very thankful for that good backup system I'd set up.

Anyway, my wife said that I would have to wait until Christmas sales to buy something to replace it, so I unplugged the whole thing and put it aside and started living off of an external drive on my laptop.

Now, yesterday, while researching what kind of hardware I should get to replace the drive, I turned on the machine in order to check how much RAM I had in it. Distracted by children, I didn't press the function key fast enough to go into the BIOS, and surprise surprise, I saw it booting up into Ubuntu 10.4. I hadn't left the install CD in the drive, so it was definitely booting from the hard drive. After logging in, I checked around the hard drive, and everything looked exactly like it had been before that initial crash. I forced a boot with fsck, which reported no errors.

So, what else could have gone wrong? I'm pretty much only at slightly advanced newbie level when it comes to hardware. I assume that there's something else in the machine that caused the problem, but what? After it had crashed, the BIOS was reporting the model number of the drive, so I assume that the connections were still good. Should I just put the server back into service and wait for it to crash again?

link|improve this question
Short answer: yes :-) – Randolph West Nov 22 '10 at 18:39
feedback

2 Answers

up vote 2 down vote

When BIOS reads the model number, it is one short (and AFAIK low-speed, as the capabilites aren't determined yet) transfer, and when you begin to transfer data with much higher speed and volumes, bad cables can of course produce this result. I'd suggest you to replace the cable and then do some high-volume tests (like copying a few gbytes).

link|improve this answer
I didn't consider that it might not be using the full cable for that. Replacing the cable is cheap, but what would cause the cable to be broken then but work now? This is a very low-traffic file server, and at the time it originally crashed there would have been almost no traffic to or from the drive. – user4815162342 Nov 22 '10 at 23:45
No, it (of course) uses the full cable, but at lower speed, and bitrate is what matters with bad cables (en.wikipedia.org/wiki/UDMA#Speed_of_defined_transfer_modes). And, as it is not broken but rather of bad quality, the errors are statistical and may appear anytime, but are more probable under heavy load. – whitequark Nov 22 '10 at 23:57
I just did a cp on a 2 GB file that worked fine. If it's copying the file from and to the same partition, will that still be a good test of the cable? – user4815162342 Nov 23 '10 at 16:38
Well, unless I'm wrong about the cable, this test is good enough. – whitequark Nov 23 '10 at 16:46
Okay, I'm thinking it's not a cable. If there were a statistical error going on, then it went from >99.99% error rate when the machine originally crashed (it was consistently failing to work) to ~0% now. That seems to point away from it being the cable. Thank you for the great idea, though. – user4815162342 Nov 23 '10 at 17:39
feedback
up vote 2 down vote

Older drives tend to be more temperature sensitive than new ones. Perhaps the drive got either too hot or cold and when you rebooted the drive found the temperature more to it's liking. I've had that happen now and then on older systems.

In either case you should probably still plan to buy that new hard drive. Ten years is excellent service and it should probably be retired while still working - or at the very least not used to store data you can't afford to lose.

link|improve this answer
Well, there is a bit of a difference in temperature between then and now around here. – user4815162342 Nov 22 '10 at 23:29
yup this temperature thing is a good catch. people put their old hard disks in freezers to get them to startup for a short while. – bronzebeard Nov 23 '10 at 8:57
While there was a temperature difference, it couldn't be more than 5 degrees (72 to 68) unless my furnace and air conditioner aren't getting into my office correctly. So, I'm not completely certain that this was the answer, so I'm sorry that I'm not going to check it off as correct. But, I think I am going to take the advice on replacing the drive. – user4815162342 Nov 23 '10 at 17:40
What matters is not the ambient temp, but the temp inside the drive. Generally speaking older drives used much more electricity and I've had some that ran so hot you couldn't hold your hand to the case without getting a burn. Still, with ambient at mid 70's it doesn't sound like temp is the culprit unless your case fans quit or something. – hotei Nov 23 '10 at 19:31
feedback

Your Answer

 
or
required, but never shown
discard

Not the answer you're looking for? Browse other questions tagged or ask your own question.