On 08/06/99 14:44:03 you wrote:
"Soren S. Jorvang" soren@t.dk writes:
For instance, I wouldn't be surprised if parity disks had different failure patterns than data disks.
Interesting, the disks I have seen the *least* ammount of failures has been on the parity drive. Usually, any new generation of drives has some issues that need to be shaken out, so the first few months proove to have higher infant-mortality-rates for (new) drives.
In my experience, parity drives were the most likely to fail, although the ratio of failures of parity:data may have been as slim as 3:2. Most failures were *not* correlated with load, mainly because Netapp load tests all the drives before selling them, and any such failures are usually seen within the first few days. So you're left with fairly normal disks that just decide to fail whenever in the middle of the night just to annoy you. :)
The exception would be a greater chance of a failure during the intense load caused during disk reconstruction, but mainly that was due, I think, to a priori problems with the disk that went undetected. RAID scrubbing took care of that issue for the most part.
This is just my experience. I'm sure Netapp has data on all of their parts and their relative failure rates; this would also address the earlier question about NIC failures. (BTW, I agree with the general consensus that the onboard NIC seems to be the most likely to fail among various models.) They might be willing to share it if you ask them (probably under NDA).
Bruce