Re: data integrity in btrfs

[Hugo Mills <hugo@carfax.org.uk>]

[-- Attachment #1: Type: text/plain, Size: 1742 bytes --]

On Fri, Dec 30, 2011 at 02:29:46PM +0100, Jaromir Zdrazil wrote:
> Hi again,
> 
>    I know that ZFS include data integrity verification against data corruption modes using propably SHA256.
> 
>    By sketchy readings at https://btrfs.wiki.kernel.org/index.html , http://en.wikipedia.org/wiki/Btrfs and other sources I have found just that there is Sha32C used and that it should be similar to ZFS.
> 
>    How are data faults detected and repaired ni BRTFS? If the answer could be simple and precize, I would be more than happy.
> 
> Thank you! Gone to lunch ;O)

   Every 4k block in btrfs is checksummed (using CRC32, so it's not
cryptographically robust against malicious modification, but should
spot most random errors).

   If you use RAID-1 or RAID-10 storage, then you get two copies of
each piece of data, stored on different devices. Each copy is
independently checksummed. When data is read, the checksum is verified
as well, and a failed checksum is logged to syslog. In this case, the
filesystem will attempt to read the other copy. If both copies are
bad, an I/O error is returned; if one of the copies is good, that data
is returned.

   With recent kernels and an up-to-date userspace, there is a feature
called scrub which will read both copies of all of the data blocks in
the filesystem and compare them to each other. If there is a mismatch
with a failed checksum, scrub will rewrite the broken block to fix it.

   Hugo.

-- 
=== Hugo Mills: hugo@... carfax.org.uk | darksatanic.net | lug.org.uk ===
  PGP key: 515C238D from wwwkeys.eu.pgp.net or http://www.carfax.org.uk
       --- "Are you the man who rules the Universe?" "Well,  I ---       
                              try not to."                               

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 190 bytes --]