On Sat, Jun 30, 2012 at 3:55 PM, Jerry Sievers <gsievers19@xxxxxxxxxxx> wrote:
Hi,
Many thanks for your reply.
Doing a:
psql -U postgres -d postgres -c "select * from pg_prepared_xacts;"
I see that there are 72 transactions (across the 5 databases in this cluster) with the 'prepared' values listed as various dates in april 2012,
ie. over 2 months old.
I guess the solution could that for each of the 'gid' values listed from the query above, do a:
psql -U postgres -d postgres -c "ROLLBACK PREPARED '<gid>';"
And then do a "vacuumdb -a" to see if it fixes the problem.
Thank again,
Christian
Christian Rosnes <christian.rosnes@xxxxxxxxx> writes:> ? WARNING:? oldest xmin is far in the past
> Hi,
>
> Coming back from vacation, I find a PostgreSQL database cluster (Pg version: 9.0.4)
> with 5 databases, which for the past 5 days have logged these messages:
>
> ? HINT:? Close open transactions soon to avoid wraparound problems.
><snip>>
> The cluster has been running continously for over 9 months with autovacuum enabled.
See if there is a prepared transaction still open. Query pg_prepared_xacts.
> I cannot find any 'idle in transaction' connected, and I can see one 'idle' connection
> (but not 'idle in transaction') which is almost 24 hours old, the rest is from early this morning.
>
Hi,
Many thanks for your reply.
Doing a:
psql -U postgres -d postgres -c "select * from pg_prepared_xacts;"
I see that there are 72 transactions (across the 5 databases in this cluster) with the 'prepared' values listed as various dates in april 2012,
ie. over 2 months old.
I guess the solution could that for each of the 'gid' values listed from the query above, do a:
psql -U postgres -d postgres -c "ROLLBACK PREPARED '<gid>';"
And then do a "vacuumdb -a" to see if it fixes the problem.
Thank again,
Christian
