|
For those of you following this thread, we've got a solution.
It turns out that the problem is Oracle. The Oracle agent buggers up the mirror on the blades, and one of the disks gets marked bad. We've patched the agent, and the disks have been stable through the weekend.
I can only assume that the agent is doing some sort of low level SCSI ioctl or SMART query or something that interrupts the mirror long enough for the RAID card to see it as an error.
I take my hat off to the 3rd line support techs, it's not something that I had considered.
| 
