RV Tec wrote:
Folks,
I had two crashes, on two different days, with the same reason: a dying
hard drive. Definitively, it is really unpleasant to get caught with my
pants down.
[there were a few potential comments here, but we'll keep this a
family-oriented mailing list! :) (ok, the real truth is I couldn't come
up with a punch line that was worthy)]
There is a way to test hard drives for possible failures or foresee
those errors?
Sure:
Just expect it.
Stuff happens. Just expect it, plan for it, know how you will deal with
it, and do so WHEN it happens.
Assume it won't happen to you, and you will end up in big trouble
eventually. Plan for it, you can make it an annoyance, rather than a
disaster.
The SMART thing isn't that smart at all. Even after the server crashed
twice due faulty harddrive, SMART keeps teeling me everything is OK.
Trust your brain, not the machine.
Nothing is going to catch all events. I had a drive fail recently where
the drive decided to toss a dead short across the power supply...nice,
brand new Seagate 500G SATA drive. Casualties:
1 Drive
Accusys RAID box (didn't seem to like a dead short)
Power supply (agreed with the RAID box, dead shorts suck)
Power supply on another P4 machine (which is pissed that Nick felt
the need to keep testing this obviously bad drive)
85+G data (which wishes Nick had noticed the incorrect jumper setting
on the new Accusys box before giving up and rebuilding the array)
Three or four days feeding backup media and uncompressing it.
The good news is I anticipated the possibility of such an event, and we
had a good backup system in place. Took me several days to fully
restore the system, but the entire system was designed to tolerate
several days of failure on THIS machine.
note: this failure had nothing to do with moving parts. This is a
"hypothetical" (ha!) failure mode I've been warning about for years,
usually with people rolling their eyes at me and thinking, "There goes
Crazy Nick again, warning about things we've never seen". However,
after 24 years in the business, I've seen enough to fully believe in
Murphy's law..."Anything that can go wrong, will", and the addition:
"Some things you didn't imagine could go wrong, will, too".
This is a SEAGATE SATA, only 1 year old. I'd expect a longer life of
those drives. Am I wrong?
You are wrong to expect any particular drive to last any particular life
span. You are also probably wrong if you expect a large quantity of
drives to actually demonstrate the rated MTBF, but that's another rant.
Stuff happens, it has to be part of your plans.
Nick.
