Hello,

I have two minor problems with Slony (in addition to my earlier
problem that I resolved to live with with sl_status having a spurious
event lag, which is apparently a known bug where confirmation events
sometimes don't get back).

I have a fairly standard setup, 1 master and 3 slaves. The master runs
postgresl 8.4.0 on windows, and each of the Slaves run 8.3.* on Linux.
I'm using Hiroshi Saito's Slony 2.0.2 package for MS Windows.

My problems are:

1. Sometimes, sl_status on the master omits the status of 2 out of 3
of the slaves that are being replicated to. When this happens, it
alternates between which slave node (which st_received) is displayed.
The problem then goes away without any intervention from me, and all
three slaves are listed. Should I worry about this?

2. I'm having the problem discussed here on the list back in July:
http://www.nabble.com/pg_autovacuum-table-gone-in-8.4--td24516041.html
, with the following error appearing in my event log:

ERROR cleanupThread: "select nspname, relname from
"_lustre_cluster".TablesToVacuum();" - ERROR: relation
"pg_catalog.pg_autovacuum" does not exist LINE 1: select enabled from
"pg_catalog".pg_autovacuum where vacreli... ^ QUERY: select enabled
from "pg_catalog".pg_autovacuum where vacrelid = $1 CONTEXT: PL/pgSQL
function "shouldslonyvacuumtable" line 25 at SQL statement PL/pgSQL
function "tablestovacuum" line 6 at IF

The problem is caused by postgres 8.4 not having a
pg_catalog.pg_autovacuum table.


Chris Browne said :

"Happily, when I "forked" the function into 8.3 and 8.4 versions, it
turns out that the API was perfectly fine for either; indeed, the
change to pull data from pg_class rather than pg_autovacuum is
relatively minor within the function.

I just committed that to 2.0+HEAD. "

Since I'm evidently missing changes from that revision, is it possible
for me to update my Slony stored procedures or something so I don't
have this problem? How?

Regards,
Peter Geoghegan