Yi Zhao <yi.zhao at alibaba-inc.com> writes:
> when I do some insert after the switch of the master node in replication
> cluster, slony alway give a notice as follows:
> ------------------
> ERROR:  cache lookup failed for type 17140
> CONTEXT:  SQL statement "INSERT INTO _replication.sl_log_1 (log_origin,
> log_xid, log_tableid, log_actionseq, log_cmdtype, log_cmddata) VALUES
> (1, $1, $2, nextval('_replication.sl_action_seq'), $3, $4);"
> ------------------
> so, I must quit psql client, and relogin, it's works.
>
> what's wrong with slony and what I should do except quit and relogin
> psql client?

Nothing is wrong with Slony-I; this is an issue with PostgreSQL where
the problem is that old cached query plans are still being run even
after the underlying object (e.g. - _replication.sl_log_1) has
changed.

As you have observed, this may be resolved by closing the connection
and opening a new one.

In PostgreSQL 8.3, this is supposed to be resolved; see the release
notes for:

   "Automatically re-plan cached queries when table definitions change
   or statistics are updated"
-- 
(reverse (concatenate 'string "ofni.sailifa.ac" "@" "enworbbc"))
<http://dba2.int.libertyrms.com/>
Christopher Browne
(416) 673-4124 (land)