Wols Lists wrote:
On 10/07/17 20:48, Per Jessen wrote:
Raid 1? Make sure if you intend to run with broken disks that you have a
3-way mirror. I'd rather run raid-6 ...
Yeah. The bigger the disk, the bigger the risk.
The other thing about raid 1 or raid 6, is that linux raid-1 always assumes the first disk is correct in the case of any discrepancy. So if you have a write problem on that disk, your data is corrupt!
If Linux always assumed the 2nd drive was correct, the situation wouldn't be much better :-)
Although with 4/5/6 it also assumes that the data is correct and if there's any error, that it's the parity at fault, there is a utility (raid6check) that you can run over a raid-6 to fix it provided only one disk is playing up.
Raid-6 provides you with two bits of redundant data, so assuming any ONE random block gets corrupted, raid-6 can work out which block is corrupt, and what the correct value is.
Raid-5 can only recover if a disk is failed, because it only has one bit of redundant data, so it needs you to tell it which block is corrupt before it can work out what the correct value is.
We have a disk fail about once a month, on average I think. Most in the bigger/newer RAID6 arrays, surprisingly rarely in RAID1 or RAID5, but they're usually smaller and older drives. Modern 2/4/6Tb SAS drives are a lot more prone to failure than ancient Compaq (9/18/36Gb drives :-) -- Per Jessen, Zürich (21.6°C) http://www.hostsuisse.com/ - dedicated server rental in Switzerland. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org