[rescue] Hard disk lifespans in Sun gear

[Mr Ian Primus]

Here's an unusual occurrence for me. Here at $WORK I
have a Sun Ultra 2 (running Solaris 2.6) that belongs
to a customer. They have the machine on maintenance
with us. I got a call the other day - "The Sun won't
boot. We had a power failure here yesterday, and when
we tried to bring it back up, it gives errors".

I fully expected to just need to run a manual fsck,
but when I got over there, I saw metadata errors. Not
good. This machine has two internal disks. But only
one shows up in probe-scsi-all.

I bring the machine back here. It has two disks. A
ST39173WC, and an ST318404LC. Obviously, these are two
different capacities, a 9 and an 18. I learn that the
last time this happened a couple years ago, a failed 9
was replaced with an 18 (all that was on hand at the
time), which was then only half-utilized. The 18 has
failed now.

I replace it with a proper ST39173WC from the shelf,
and run the replacement drive through the analyze
routines in format just be sure, then rebuild the
mirror. I won't go into the details of the hoops one
must go through to replace a failed disk in a bootable
mirrored drive, but I got the system back up and
running properly. I leave it to run overnight.

This morning, when I come in, I find that metatool now
shows one drive as critical. So does metadb. I shut
the system down and get back into the firmware prompt.
Now probe-scsi-all only shows one disk. The
replacement drive has failed.

What is going on here? I haven't seen this level of
disk failure before. In my experience, these Seagate
drives are very reliable. I know that the slot isn't
bad, because during the process of replacing the
drive, the positions got reversed. Another odd thing
is that they both failed the same way. They both don't
spin at all. Normally, when I do see these drives
fail, they spin, but make clunking noises. I am
willing to chalk it up to coincidence, but I would be
interested to hear other people's history with Sun
disks.

-Ian