Sebastian Rother wrote:
Hello everybody,
I would like to know if the spare-disk (specified in the raid.conf of
raidctl) gets powered off until it's needed.
As far as I understood the manpage the spare-disk is used to rebuild
the raid if a HDD fails.
And example what happened to a company I know:
A company I know had 2 Servers, redundant, for working.
As the mainserver crashed (HDDs) they used the secondary Server.
This crashed also 34 Minutes later.
The company bought the Servers at the same time and they where running
all the time.
In this case they had no Spare-Disks.
I ask if the spare-disk gets powered off because the spare-disk could
crash too (lifecricle). Normaly you buy a Server and the disks and also
the (maybe then used) spare disk at the same time.
So the Disks run exactly the same time and this could be a risk (for
the spare disk).
So if this option isn't supported yet (as I said I found nothing at the
manpage) would it make sense to support it (forcing a HDD to suspend
until it's needed (sleep mode))?
I REALLY don't think so.
All you will end up doing is trading one mode of failure for another.
I've heard stories about the all the drives failed at once!, too.
I've also seen cases where a bunch of powered-on drives run fine, until
you spin them down for an hour, and they won't spin back up. So now you
will want to periodically spin up and test drives. That will bring on
other failure modes.
Now, after 23 years in the computer industry, if those drives lasted
more than a few months in service before failure, I would suspect
something in the environment got them. Quality control just isn't that
good in the computer industry to have a whole batch of parts fail after
some large value of X hours of operation to all fail within a fraction
of an hour. If X is small, sure... If the failure time is more
spread out, sure. IF it was an environmental problem, spinning down
drives just opened you up to NEW modes of failure without solving the
problem you were fearing.
Complexity is the enemy of total uptime. Well-designed simple systems
are easy to fix and restore. Complex systems that aren't ever supposed
to go down tend to go down with a most spectacular boom, and stay that
way for days while the maintainers try to figure out what went wrong and
how to work their way through all the twisted wreckage to bring up
something functional again...
When you design for a very special case failure, you have to be aware
you may be opening yourself up for a lot of /other/ special case
failures. It is like the person who fears a plane crash, so they drive
everywhere, exposing themselves to higher risks.
Nick.
