Re: hard drive problem

Aragorn wrote:
On Sunday 27 August 2006 01:31, The Natural Philosopher stood up and
addressed the masses in /comp.os.linux.misc/ as follows...:

Jean-David Beyer wrote:

Most retail drives can be expected to last about 5 years in continuous
service. Possibly a bit less if they are turned on and off frequently.
And their warrenty is less than this (1 - 3 years?). So you may want to
figure yours are nearing the end of their useful life. So you better be
making daily backups of everything that would be difficult to replace.
For a 24x7 server 5 years is pretty good life 2-3 is nearer the mark...
For something that is using the disk continuosly, really continuously, a
year is not unknown as failure point.

That's why servers typically employ SCSI disks. You know that there's a
reason why they're about four times as expensive as an IDE disk, don't
you? ;-)

Desktop machines should do about 5 years.

The machines themselves, yes. The hard disks mounted in them typically have
a warranty period of one to three years, and almost die within months after
that period expires, if not sooner.

IDE hard disks are intended to be used only about 8 hours per day, with
about 2 hours in total during that 8-hour timespan of being under heavy
load.

SCSI hard disks are intended to be up 24/7, with about 80% of that time
being under heavy load. Their MTBF - I know, it's just a number - is
usually around 1'000'000 hours, which boils down to about 138 years under
continuous service.


Mm.. However what makes a disk less than useful is not gross failure, but a high error rate..repeated read/writes eventually degrade the magnetic material..also operation in dusty places (under the deak, not in a clean server room) will push failure rates way up..

OK its been several years since I had a farm under my management, but even so one or two failures in 60-70 SCSI drives - failures *bad enough to need a disk replacement* - were about the annual rate. All SCSI.

I guess thats makes the AVERAGE failure rate something like 30 years...


My personal advice would be this... *Always* make backups, regardless of
the type of hard disks you're using, and regardless of whether they're used
as standalone disks or in RAID arrays. Use SCSI disks for systems that
require a standalone disk - e.g. because you can't afford two SCSI disks -
in a mission-critical environment - with the word "mission-critical" in a
very subjective context.

If you would rather use IDE disks - be it PATA or SATA - then at least get
two of them and put them in a RAID 1 array. Less storage percentage is
lost and a more comfortable handling of risks is achieved with higher RAID
levels, e.g. RAID 5 or RAID 10.

IDE disks are cheap these days, dirt cheap. Buy at least two disks at the
time and put them in a RAID set-up - _note:_ RAID 0 does *not* offer
redundancy and even increases the risk of dataloss by another 100%! - and
put the IDE disks in removable trays. They tend to break down quite soon
when under heavy load, and this way, it'll be easier to replace them.

If an IDE hard disk in a RAID array breaks down, replace the faulty disk,
rebuild the array and then replace the other disk. If one of them goes,
the second one won't be far behind anyway. You can run the array in
degraded mode for a day or two when unable to go to the shop to buy new
disks.

Most people find SCSI far too expensive. So plainly and simply, buy IDE
disks and use RAID. They're dirt-cheap these days, so best is to make use
of it while you can...

Just my two cents... ;-)



The rest I endorse wholeheartedly.
.

Relevant Pages