* Robert Treat <xzilla@users.sourceforge.net> [17.03.2003 16:49]:
> This is exactly the way pg_dump works. Essentially when you pg_dump, you
> get a "snapshot" of your database at the moment you start pg_dump. While
> other transaction can be occuring during your dump, the dump wont see
> them thanks to MVCC.
>
I see, thank you.
One more thing. I still cannot find any info on HOW pg_dump locks data it
is going to backup. Is this described somewhere in the manual? (I've found
nothing neither in "Concurency Control" nor in pg_dump description).
What I'd like is to understand the algorithm. I know how backuping is
realized in ORACLE, and I'm sure there, that when I start it, lock is set on
entire tablespace (it's put in readonly mode), so I'm shure, that nothing
will get there.
In PostgreSQL I don't know how lock is aquired. I'm trying to avoid
situations, like (assumption): lock is aquired on 'per table' level and
while first table in the list is locked, last one is being updated, so -
my db won't be consistent after backup.
Thanks in advance.
--
Victor Yegorov