On Wed, 28 Aug 2002, Christian Mock wrote:
> On Tue, Aug 27, 2002 at 02:08:17PM -0400, Andrew Sullivan wrote:
>
> > > hangs until the first transaction is done; the related postgres process
> > > shows as "UPDATE waiting" in ps.
> >
> > Do you have referential integrity on any of these? That's the usual
> > suspect in these cases.
>
> I see. One of the tables that are inserted into in the long-running
> transaction references the table that gets updated in the "UPDATE waiting"
> transaction. But the referencing column is always NULL on import.
>
> Does that mean that there's a table-level lock? Can I get around this?
Hmm, I may have missed it, but can you give an example of the actual
statments and schema?
The only workarounds I know about are hacking the trigger functions and
making the constraints deferred (which is only a partial solution, but
often helps in practice).