Scott G. Miller wrote:
> On Thu, Jan 8, 2009 at 8:19 AM, Scott G. Miller <sgmiller at gmail.com 
> <mailto:sgmiller at gmail.com>> wrote:
>
>     After enabling replication for a large database, I'm consistently
>     seeing a state where logs will not switch.  sl_log_1 contains some
>     decent number of rows (a few thousand), and sl_log_2 is growing. 
>     Executing _replication.logswitch_start() and finish() has no
>     effect, as both report switch still in progress.  This continues
>     for days, until eventually there are millions of rows in sl_log_2,
>     and sync times take 15s or more just to fetch rows to apply to the
>     slave.
>
>     Thoughts?
>
>
> Oh, sorry, this is Slony-I 2.0 running on Postgres 8.3.3, simple 
> master/slave configuration with one database and 5 sets.
>
What does test_slony_state.pl (or the DBI version, 
test_slony_state-dbi.pl) report?

The truncate won't happen if events aren't propagating properly 
everywhere; the output of that script may give clues as to what's wrong.

It is a good practice to run that script regularly, as it will notice 
and report on a number of sorts of problems that have caused people to 
report errors, sometimes when they had some configuration problem...

-- 
(format nil "~S@~S" "cbbrowne" "ca.afilias.info")
<http://dba2.int.libertyrms.com/>
Christopher Browne
(416) 673-4124 (land)