Re: RAID1 problem - server freezes on md data-check
On Tue, 05 Jan 2010, Ross Halliday wrote:
> Technologies like RAID5 which have parity checking will tell you the
> instant one disk is behaving badly and kill it from the array - this is
> more suited to protecting against partial failures and data corruption.
You want RAID6 for that. The math is simple: you need n>1, n odd, to use a
simple majority vote to know which data set is correct. RAID5 cannot do
that. RAID1 can, but not with 2 devices.
RAID5 is only useful for known failure (i.e. you get information about WHICH
component device is bad, e.g., through sector IO errors). If it is silent
corruption, you're screwed.
> I have to seriously question the value of this once-a-month check as the
> other 27-30 days of the month your disk could be half-dead, spewing
So do I. At work use short SMART scans *daily* to locate bad sectors (it is
good at finding a cluster of weakening sectors, but not perfect), and a
repair scrub once a week to reduce bitrot. But you're better off with a
scrub once a month, than never scrubbing at all.
But that has nothing to do with silent corruption protection. If that is
what you're afraid, your problem needs a very different solution than RAID
could give you.
--
"One disk to rule them all, One disk to find them. One disk to bring
them all and in the darkness grind them. In the Land of Redmond
where the shadows lie." -- The Silicon Valley Tarot
Henrique Holschuh
Reply to: