Re: incremental backups

On Jan 30, 2006, at 6:58 PM, Tom Lane wrote:

> Rick Gigger <rick@alpinenetworking.com> writes:
>> And here is the real million dollar question.  Let's say for some
>> reason I don't have the last WAL file I need for my backup to be
>> valid.  Will it die and tell me it's bad or will it just start up
>> with a screwed up data directory?
>
> It'll restore up to the end of the data it has.  The only case that's
> actually "invalid" is not restoring far enough to cover the time
> window
> that the original base backup was taken over.  Otherwise it's just a
> situation of restoring up to a particular point in time...
>

That's what I mean by invalid.  Let's say I do something stupid and
do a physical backup and I don't grab the current WAL file.  All I
have is the last one to be archived before I did my backup, which is
not late enough to do a valid restore.  Will postgres know that the
restore process failed because I didn't have that last necessary WAL
file or will it just start up in a potentially inconsistent state.
Obviously that would be my fault not postgres' since I am the one
that didn't give it the data it needed to do a full restore.  But I
am just wondering if that is a potential area to shoot yourself in
the foot or if postgres will put the safety on for me.

Rick