On 1/7/2005 3:45 PM, James Black wrote:
> Our environment: Linux 2.6 on x86, Postgres 7.4.6, Slony 1.0.5.
> 
> Every 10 minutes, almost exactly, the slon daemon on our subscribing 
> node shuts down, as follows:
> 
> FATAL  cleanupThread: "delete from "_tii".sl_log_1 where log_origin = 
> '1' and log_xid < '222620814'; delete from "_tii".sl_log_1 where 
> log_origin = '1' and log_xid < '222620814'; delete from 
> "_tii".sl_seqlog where seql_origin = '1' and seql_ev_seqno < '203472'; 
> " - ERROR:  operator does not exist: _tii.xxid < "unknown"
> 
> Obviously the XXIDs are different.  The daemon restarts without 
> problems.  What's going on?  Has anybody else experienced this? Is it a 
> configuration problem?
> 
> Thanks,
> jfb
> 

Somehow you managed to drop the < operator for the xxid data type. I 
know that pg_dump has a problem when one dumps a namespace alone 
(pg_dump -n ...) in that it doesn't output the create operator commands. 
Could it be that the database in question was dumped and restored in 
some way?

As a cure I would suggest initiating an empty new node. No need to 
subscribe it, just "createdb" and via slonik "store node". Can be well 
on one of the existing machines, no need to ever start a slon for it or 
even store paths. Then take a full schema dump of that database and drop 
the node again. From that dump edit out the definitions for the xxid 
data type and feed them via psql into the DB having the problems.


Jan

-- 
#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me.                                  #
#================================================== JanWieck at Yahoo.com #