Software RAID issues (was Re: Suggestions solicited, server bring up)

[Ben Scott]

On Tue, Nov 24, 2009 at 2:32 PM, Ken D'Ambrosio <ken at jots.org> wrote:
> UDNRC [sic].  With all due respect to the encyclopedic knowledge of Ben, I
> took this one with a grain of salt.  And again, Wikipedia to the rescue:

  I trust my own experience a lot more than I trust a Wikipedia
article tagged as needing more citations.  But hey, what you trust is
up to you.  But more than either of those two, I trust the Linux
sources, and investigation there finds there's no checksum storage or
calculation (outside of support for the LFN hacks).  So no CRC in FAT
filesystem.

  Back in the bad old days of DOS, occasionally you'd get a message
along the lines of

	Data CRC error reading drive A:
	Abort, Retry, Fail?

trying to read that really important file you didn't have a backup of,
so there was a CRC *somewhere*.  Perhaps it was a floppy thing?

> long filename FAT had checksums for the *filename*, not the file data.
> This was to to ensure that the 8.3 filename matched the long filename.

  MS-DOS doesn't support long file names, so whatever LFNs do is
pretty much irrelevant to what MS-DOS does (did).

>> You can do this on Linux with LVM snapshots.
>
> And, indeed, it may actually be implemented via btrfs in the next release
> of Rawhide:
> http://lwn.net/SubscriberLink/362500/8331dbd6172b5b85/

  Which is nice and all, but LVM has the advantage of having been
available for several years now.  </curmudgeon>

  (All this filesystem neophilia needed a counterpoint.  ;-)  )

-- Ben