Jan, all,
* Jan Wieck (jan@wi3ck.info) wrote:
> On Sat, Dec 30, 2017 at 10:27 AM, Stephen Frost <sfrost@snowman.net> wrote:
> > The checksums included in PG are page-level and therefore there simply
> > isn't one to look at if the file is zero bytes.
>
> And even if the file wasn't zero bytes you can't tell from the per page
> CRCs if you have all the pages you should have. You could have extra pages
> that aren't supposed to be there or missing some (or any mix of the two).
> A per page CRC is useless for those cases.
Right, which is why it makes sense to have whole-file checksums when
doing things like backups, when you're already reading all of the file
and can grab a checksum in stream, to detect if anything bad happens
after the backup has compeleted to the file that's been backed up.
It's unclear if that would have helped here or not since we don't know
when the file ended up being zero'd out, as I understand it.
Thanks!
Stephen