Re: pgsql 9.0.1 table corruption

Benjamin Smith <lists@benjamindsmith.com> writes:
> On Friday, April 15, 2011 09:50:57 AM Tom Lane wrote:
>> If you simply unpacked the tar archive and started a postmaster on that,
>> you'd be pretty much guaranteed to get a corrupt database.  The tar
>> archive is not a valid snapshot by itself --- you have to replay
>> whatever WAL was generated during the archiving sequence in order to get
>> to a consistent database state.

> I have, more than once, "moved" a PG instance from one machine to another with
> the following sequence, without apparent issue. is there anything I'm missing
> and/or need to be concerned with?

> 1) service postgresql stop;
> 2) rsync -vaz /var/lib/pgsql root@newserver:/var/lib/pgsql;
> 3) ssh root@newserver;
> 4) (edit postgresql.conf, set IP addresses, memory, etc)
> 5) service postgresql start;

That way is fine --- what you can't do is take a tarball while the
server is running, unless you do the full nine yards with WAL replay.

            regards, tom lane