Not My Day.

[swestrup]

So, I'm in the middle of rebuilding the server and I start getting strange errors. Since there are known bugs in the VT8237r chip that's in the server, I immediately suspect that its having trouble with talking to two 1TB drives at once, or one of the data channels is bad, or something.

Nope, in the end it turns out that one of the three brand new drives was failing and has now completely failed. Even the diagnostic software from Western Digital says (in so many words) "Its toast -- ship it back".

Luckily, it wasn't the drive that I did 90% of the imaging to. I still have the majority of the image data. Alas it DID contain the image of the largest raid array in assembled form. I have all the individual partition images (plus whole drive images -- I get paranoid when doing this stuff, and I hope you can see why) but it would be a royal pain to rebuild the array image by hand, and I dislike remounting the failed drives except as a last resort. Luckily, I've been told that Trinity Rescue Kit is a LiveCD that contains software intended to automate building assembled raid images from partition images, so I'm going to try it out. If worst comes to worst, I've just lost a few days work.

But still:  GRUMPH!

Comments

[sps.livejournal.com]

Gacko in spades!

[hendrikboom.livejournal.com]

Just curious -- that brand-new drive that failed -- did you do a complete write-read test on it before using it? Or did it fail after such a test?

[swestrup]

I only did a full read test on the new drives as a full write-read test on a 1TB disk would have taken an extra day or so per drive.

In any case, the media didn't fail. The diagnostic I got back from the WD drive test indicated a servo-mechanism failure. The drive no longer seeks properly.

[hendrikboom.livejournal.com]

Yeah. A full readwrite test does take a long time. I remember taking days to test my new backup drives a few years ago. One of them failed during testing, and I got it replaced under warranty. A simple exchange of new for old in the store. The new one took another few days to test.

-- hendrik