there is already checksumming of metadata. having a file-system option of
checksumming every file's contents upon closing the write would be very
convenient, too. It would only take an extra few bytes in the file descriptor,
wouldn't it? It would be nice insurance against hardware corruption. zfs has
it, too, I believe. just a suggestions from someone who is just an end user...
On Jun 13, 2008 01:33 +0000, iaw4 wrote:
> there is already checksumming of metadata.
There is only checksumming of the metadata in the journal and for the
group descriptor. There is not yet checksumming of the rest of
metadata (bitmaps, inode table, superblock, etc).
> having a file-system option of
> checksumming every file's contents upon closing the write would be very
> convenient, too. It would only take an extra few bytes in the file
> descriptor, wouldn't it? It would be nice insurance against hardware
> corruption. zfs has it, too, I believe. just a suggestions from someone
> who is just an end user...
Sure, I don't think anyone objects to it, but it is a lot of work to
implement and nobody has yet made the effort to do so. If someone is
looking to work on this, we've given a lot of thought to the design
of how to implement this.
Cheers, Andreas
--
Andreas Dilger
Sr. Staff Engineer, Lustre Group
Sun Microsystems of Canada, Inc.