On Fri, Apr 8, 2011 at 4:00 PM, Jeff Davis <pgsql@j-davis.com> wrote:
> On Fri, 2011-04-08 at 15:03 -0400, Bruce Momjian wrote:
>> A fix will be included in upcoming Postgres releases 8.4.8 and 9.0.4.
>> These releases will remove the need for the above script by correctly
>> updating all TOAST tables in the migrated databases.
>
> You might want to clarify that the fix may be required if you ever used
> pg_upgrade before. Using the new version of pg_upgrade/dump when you
> still have a bad relfrozenxid doesn't help.
>
> Regards,
> Jeff Davis
>
I've been noticing in my logs for the past few days the message you
note in the wiki. It seems to occur during a vacuum around 7:30am
every day. I will be running the suggested script shortly, but can
anyone tell me in how bad of shape my db is in? This is our production
db with two hot standby's running off it.
grep -i 'could not access status of transaction' postgresql-2011-04*.log
postgresql-2011-04-06.log:2011-04-06 07:28:27 PDT [15882]: [1-1]
(user=postgres) (rhost=[local]) ERROR: could not access status of
transaction 1273385235
postgresql-2011-04-07.log:2011-04-07 07:27:14 PDT [29790]: [1-1]
(user=postgres) (rhost=[local]) ERROR: could not access status of
transaction 1273385235
postgresql-2011-04-08.log:2011-04-08 07:26:35 PDT [2402]: [1-1]
(user=postgres) (rhost=[local]) ERROR: could not access status of
transaction 1273385235