[Pgpool-general] Pgpool and/or Slony
Andrew Sullivan
ajs at crankycanuck.ca
Wed Jan 17 21:19:59 UTC 2007
On Wed, Jan 17, 2007 at 04:12:52PM -0500, Brad Nicholson wrote:
> pgpool docs clearly state:
>
> "Please note that certain queries are physically dependent to a server
> or non-deterministic. These include random functions, OID, XID, and
> timestamps may not be replicated in exactly same value among two
> servers. This is because query results for these are local to each
> PostgreSQL server."
Yes, but I'm pretty sure (I'm not looking at the docs now, and I'm
too lazy to test) that if you do
SELECT CURRENT_TIMESTAMP
or friends through pgpool (any flavour), it will catch and
re-interpret it. The problem comes when you have, for instance, a
DEFAULT on a table that is CURRENT_TIMESTAMP. Slony will handle this
correctly, but pgpool might cause the tables to drift. I'm happy to
be corrected in the details; it makes no difference to your larger
point, which is certainly true. There is no way in principle, for
instance, for a statement replicator to know that your stored
procedure calls another stored procedure that happens to call now().
It's always possible to cause statement-based replication to fail in
this way using _some_ method. The question is mostly whether your
application trips over that limitation.
A
--
Andrew Sullivan | ajs at crankycanuck.ca
I remember when computers were frustrating because they *did* exactly what
you told them to. That actually seems sort of quaint now.
--J.D. Baldwin
More information about the Pgpool-general
mailing list