Raid 6 & URE question

[ToTheMax]

I've been trying to research how raid 6 functions in regards to raid 6 during a drive failure and encountering a URE (unrecoverable read error)

If we assume the following;
- 12TB drive array
- Using WD red drives with a manufacture stated 10^14 URE rate, this will give a 100% chance of encountering a URE while the array is being rebuilt
- Raid 5 will typically stop a rebuild if it encounters a URE loosing all data
- Raid 6 has 2 parity drives so can deal with 2 drive failures
- A URE isn't a dead disk, just a sector that cannot be read

My question is will a rebuilding raid 6 array with a failed disk fail another disk when it encounters a URE or will it continue without failing the disk?

Thxs.

 

[nutcase]

Probably depends on the RAID controller, but the likelyhood is it will dump the array. 12TB RAW or usable, and what size drives? RAID 5 it is common to see recommendations not to use drives over 1TB, not really sure about R6 as never used it.

I only ever use R10 these days for performance and failure tolerance. Still got 2 servers running R5 but both use 300GB drives so not too worried.

edit: sorry complete brain fart. I was thinking a third disk hit an URE. It will very likely spit second drive. Long day yesterday trying to coax some more speed out of a full backup of 1.8TB when for some reason it's only running at about 7MB/Min so looking at not finishing until Monday...

 

[Mercutio]

Pretty sure with raid 6 it would fail the second drive (so the array would be readable but show as degraded). I can't think of another sensible outcome for a controller if it knows a drives dead AND finds a URE on another drive while trying to rebuild. It's basically out of redundancy options at that point so should degrade the array.

If you can wipe the drive with the URE (as it was maybe just "bit rot" rather than an actual hardware failure) and force it to accept the drive again it should still rebuild onto it but you're looking at a rebuild/re-sync with the URE drive being seen as a new raid member I'd think.

I'd probably want to (at least) throw a bunch of diags at the drive with the URE to ensure there's not the start of issues with it before forcing the controller to accept it again (or assume it's suspect and replace both).

One of these days I'm going to do something crazy like a distributed raid (something like hadoop) with a friend or 2 on good broadband. It's almost certainly overkill but is probably about the best way to ensure a (literally) bomb proof backup.

Hadoop's awesome but not really intended for anything other than endless archiving. (Basically you have multiple machines with storage attached, the whole cluster of machines then presents it's storage as a single mount. Anything saved, there's at least 1 other copy of the blocks of data on another entire machine. Spread the clustered machines over multiple sites and you can take a building burning down with all kit destroyed and still have data )

 

[ToTheMax]

Thanks for the very useful info. Specifically I was weighing up whether a 6 * 6TB disk raid 6 setup was worthwhile with drives stated at a 10^14 URE rate. If I go by the manufacturers stated URE rate I would see multiple URE events in the event of a rebuild (1 failed disk).

I might just go with a raid 10 setup which avoids the rebuild issues parity setups have.