Richard Yen <dba at richyen.com> writes:
> I've got a monitor running on my Slony cluster, and it notifies me
> when lag_num_events and/or lag_time exceeds a particular threshold.
> For the purposes of business requirements, I've set the threshold to
> 10 seconds and 20 events, with a sync_time of 500ms in my conf file.
> With time, the frequency of the notifications of lag have increased.
> Consequently, I was able to convince the authorities to let me
> increase the thresholds to 30 seconds and 60 events.  More time passed
> (months), and now I'm getting pages again.
>
> Is there a known bloating issue with slony and its lag times?  If so,
> how can we un-bloat?

Well, I'd start by running a VACUUM VERBOSE against the Slony-I
tables.

Thus...

PGDATABASE=something
PGPORT=something
PGUSER=suitable
CLUSTER=foo
export CLUSTER
TABLES="sl_archive_counter sl_config_lock sl_confirm sl_event
sl_listen sl_log_1 sl_log_2 sl_node sl_nodelock sl_path sl_registry
sl_seqlog sl_sequence sl_set sl_setsync sl_subscribe sl_table"
export TABLES
for table in `echo ${TABLES}`; do
   vacuumdb -t "\"_${CLUSTER}\".${table}" -v
done

And see if there's some table that has grown precipitously.  If you're
on Slony-I 1.2, then sl_log_1 and sl_log_2 should periodically be
getting truncated, so they oughtn't be a problem.

I could see there being growth of sl_seqlog, if you have a lot of
sequences, but that shouldn't cause the problem you are describing.

You might find a table deserving a reindex.  It would be useful to see
the stats, if that's the case.
-- 
(reverse (concatenate 'string "ofni.secnanifxunil" "@" "enworbbc"))
http://cbbrowne.com/info/internet.html
Rules  of the  Evil Overlord  #41. "Once  my power  is secure,  I will
destroy all those pesky time-travel devices."
<http://www.eviloverlord.com/>