On Thu, Mar 22, 2018 at 2:24 PM, Jeremy Finzel <finzelj@gmail.com> wrote:
> I am running this on a san snapshot of our production system. I assume that
> this will give me a valid check for file-system-level corruption. I am
> going to kick it off and see if I find anything interesting.
It might. Note that SAN snapshots might have corruption hazards,
though. Postgres expects crash consistency across all filesystems, so
you might run into trouble if you had a separate filesystem for WAL,
for example. I know that LVM snapshots only provide a consistent view
of a single logical volume, even though many LVM + Postgres setups
will involve multiple logical volumes. This makes it possible for a
small inconsistency across logical volumes to corrupt data.
I don't know anything about your SAN snapshotting, but this is at
least something to consider.
--
Peter Geoghegan