Thread: Postgres problem with partition queries

Hi All,
We are having a partition table and we are querying some rows in the
partition.
select * from events where
     events.evt_time >= '2010-11-29 00:00:00'
    AND events.evt_time <= '2010-11-30 23:59:59'
While this query is in running state, I am issuing a drop statement on a
partition of this table.

drop table events_p_20101207130000

Please note that the query is running on a different partition and drop is
happening on a different partition.
Now, in a different session, I am running a different query,
select * from events where
     events.evt_time >= '2010-11-28 00:00:00'
    AND events.evt_time <= '2010-11-29 23:59:59'

Now the issue is both the drop and my 2nd query is in waiting state and both
these queries started running after my first queries completion.
We have set constraint_exclusion to on and we used, select * from
pg_stat_activity where waiting = 'true'  query to check the status of the
waiting queries.

Now, can you please explain me on why it is behaving like this. Is this a
postgres bug or any configuration needs to be done to avoid this waiting.

Any help is really appreciated.

Thanks,
Bakki

--
View this message in context:
http://postgresql.1045698.n5.nabble.com/Postgres-problem-with-partition-queries-tp3287234p3287234.html
Sent from the PostgreSQL - admin mailing list archive at Nabble.com.

bakkiya <bakkiya@gmail.com> writes:
> We are having a partition table and we are querying some rows in the
> partition.
> select * from events where
>      events.evt_time >= '2010-11-29 00:00:00'
>     AND events.evt_time <= '2010-11-30 23:59:59'
> While this query is in running state, I am issuing a drop statement on a
> partition of this table.
> [ and it blocks ]

This is hardly surprising.  A query on the inheritance tree has to
take AccessShare lock on every member table, because it has to inspect
each table to determine whether it should scan it.  So it's going to
conflict with the exclusive lock needed to drop one of the tables.
That means first that the DROP gets blocked by a pre-existing query,
and second that the waiting DROP also blocks other queries of the same
kind.

            regards, tom lane