Peter Jeremy peter at rulingia.com
Fri Feb 22 18:56:39 UTC 2019

On 2019-Feb-22 11:47:37 -0600, Karl Denninger <karl at denninger.net> wrote:
>Uh uh.  Yes, a scrub (or just checksum check on the read) will catch the
>"aw crap!" situation where write amplification and read/rewrite hoses
>you during a power failure but there's nothing zfs can do to fix it
>since there's no second copy available.

That's not completely true.  By default ZFS stores 2 copies of most
metadata and can be configured to store up to 3 copies (on top of pool
redundancy) of all data (check the "copies" and "redundant_metadata"
properties of the filesystem).  Obviously, if all the copies are on
one device, it won't help if the device bricks itself or all the
copies wind up on the corrupt sectors but it can help recovering from
a glitch that took out a page or two.  (It would also be possible to
setup a ZFS mirror between a SDcard and a USB flash stick).

Peter Jeremy
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 963 bytes
Desc: not available
URL: <http://lists.freebsd.org/pipermail/freebsd-arm/attachments/20190223/f24408db/attachment.sig>

More information about the freebsd-arm mailing list