Re: RAID5/6 permanent corruption of metadata and data extents

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Thu, Apr 2, 2020 at 5:09 AM Filipe Manana <fdmanana@xxxxxxxxx> wrote:
>
> Any thoughts? Perhaps someone else was already aware of this problem and
> had thought about this before. Josef?

Upon degraded, there is some data that is only ever written as P or Q.
This requires two stripe elements to successfully be read to
reconstruct what is missing (because it was tossed, because of the
degraded state). What about falling back to raid1 (for raid5) or
raid1c3 (for raid6)? That is, freeze the raid56 metadata block groups,
and only do new writes to raid1 or raid1c3?

Alternatively, disallow raid56 for metadata going forward. Make it not
possible to use -mraid5 or -mraid6 at mkfs time. And explicitly
recommend everyone convert. This might be the easiest fix.

-- 
Chris Murphy



[Index of Archives]     [Linux Filesystem Development]     [Linux NFS]     [Linux NILFS]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux