[Pgpool-general] Online recovery with sequence

[Jaume Sabater]

On Mon, Aug 17, 2009 at 2:37 PM, Tatsuo Ishii<ishii at sraoss.co.jp> wrote:

> In local pgpool mailing list in Japan, a user found that sequence
> values among servers are out of sync after online recovery. This is
> caused by the way how PostgreSQL logs sequence: it logs 32 on WAL
> *before* PostgreSQL uses all of 32 sequence values. So after recovery,
> recovered node's sequence advances about 32 than the one of the online
> node. To handle the problem, I recommend to place following code in
> your second stage recovery script.

If no sequences (SERIAL types and such) are used, I take it that we
would not need this "patch", would we? Because one error I got after
upgrading to 2.2.3 was queries returning different number of rows in
the two nodes (second node had just been recovered), and it's the
error I am still trying to reproduce before I report it. Perhaps it
has (or had) to do with this stuff...

-- 
Jaume Sabater
http://linuxsilo.net/

"Ubi sapientas ibi libertas"