For those in the business, have you ever seen or heard of a case where three hard drives fail on the same server on the same day?

This server I'm talking about has 42 SCSI drivers in 6 arrays, running Raid5. When the first drive failed, a tech was called on site to swap it out. While the tech was rebuilding the Raid, a second drive on the same Raid also died. They had no choice but to recreate the raid, losing everything. After they restored the system from backup, another drive in another raid failed. That's 3 drives in 12 hours.

How's that for a Friday the 13th story?

 

I haven't seen it happen, but whenever I have a drive fail it does cross my mind. The odds have got to be slim. The only question I would have would be about the environment - I mean, this system isn't in a room where the ambient temp. sits at 90 all the time is it? That's rhetorical; I realize noone would do that but....

 

[QUOTE]Originally posted by Strike
[B]I haven't seen it happen, but whenever I have a drive fail it does cross my mind.

 

I figured. Now you've got me worried

 

I think it only happens on Friday the 13th.

Most people consider backup as mindless routines. I suggest that we not only backup as often as possible, but to go through the logs after each backup to verify its success. If possible, practice restoring files, whole directories w/ subs, ever OS (if you are lucky enough to have test systems).

 

Are you absolutely certain the drives went bad, or were you relying on software to tell you they went bad? I ask beacuse I've had several Compaq Proliant systems tell me that a drive went down. In one case it was the controller card and in another the SCSI cable came loose.

 

Wait... doesn't RAID 5 use an algorithm where there are 7 disks, and if up to three of the disks fails, you can reconstruct their content from the remaining four?

 

Raid 5 is defined by a set of N drives with similar volume where N-1 drives contain data, and one parity drive. The effective storeage volume is (N-1)x drive storage volume. If any one of the drives are bad, you just replace it and rebuild the raid. But if more than one drive fails, you are out of luck.

Raid 5 can be implemented by hardware and software. Higher end systems usually use SCSI raid controller where the processing is done on board. On windows servers since NT 4.0, you can implement raid 5 using the OS.

 

Yes, I have. if you have 42 drives in 6 array. you should consider raid 10 or raid 15.

 

[QUOTE]Originally posted by jefftse
[B]Yes, I have.