Jeff Janes <jeff.janes@gmail.com> wrote:
> On Mon, Aug 5, 2013 at 11:16 AM, Joshua D. Drake <jd@commandprompt.com>
> wrote:
>> I seem to recall autovacuum changes landing for 9.2.4. Can someone please
>> describe what those changes were and how they could affect usage?
>
> Those landed in 9.2.3, see release notes for that version:
> Fix performance problems with autovacuum truncation in busy workloads
> (Jan Wieck)
> Fix error in vacuum_freeze_table_age implementation (Andres Freund)
>
> There should be no change in usage, unless you were taking some heroic
> methods to overcome the problems and can now discontinue them.
Well, that was the intent, but there was an unintended increase in
the frequency with which an autovacuum which attempts to truncate a
heap may fail to set new statistics, and the logging around
truncation got a bit too "chatty". These issues will be corrected
with the next minor release, but until then some users may need to
run ANALYZE commands in some cases to prevent tables with large
swings in size from developing stale statistics, and there may be
some new LOG entries which users wonder about -- they can safely be
ignored.
Overall, the autovacuum changes in 9.2.3 put an end to some
debilitating problems with blocking and load related to overly
aggressive and eager autovacuum runs. Jan's fix addressed problems
with tables used for queues, as in slony and some JMS
implementations. Andres fixed a bug which caused wraparound
prevention autovacuum runs to occur too frequently.
--
Kevin Grittner
EDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
