Strange behavior with pg_locks and partitioning

On Fri, Jun 15, 2012 at 12:42 PM, Josh Berkus <josh@agliodbs.com> wrote:

In the course of debugging why a particular server required increasing
max_locks_per_transation, I found a peculiar behavior.  If you do an
UPDATE which doesn't match any CE constraint on the parent table in an
inheritance chain, you get a RowExclusiveLock on every partition and
every index on every partition.  However, these rowexclusivelocks have
no page or tuple reference; it's a RowExclusiveLock with no row.

Is this intentional?

RowExclusiveLock is a type of table lock, not a lock on a row.

You're going to get that on all tables (and their indexes) involved in
any write query.

So it sounds unsurprising to me.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company

So it sounds unsurprising to me.

OK, I'll just submit a note for the docs for max_locks_per_transaction,
then.

--
Josh Berkus
PostgreSQL Experts Inc.
http://pgexperts.com