On Tue, Dec 15, 2009 at 1:33 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Robert Haas <robertmhaas@gmail.com> writes:
>> On Tue, Dec 15, 2009 at 1:23 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>>> Alvaro Herrera <alvherre@commandprompt.com> writes:
>>>> If we're to do anything about this, it is spilling the trigger queue so
>>>> it doesn't eat an unbounded amount of memory.
>>>
>>> Of course, the reason nothing much has been done about that is that
>>> by the time your trigger queue is long enough to cause such an issue,
>>> you're screwed anyway --- actually executing all those triggers would
>>> take longer than you'll want to wait.
>
>> What is the best way to go about doing that, anyway?
>
> Well, we added conditional triggers which provides a partial fix. =A0The
> only other idea I've heard that sounds like it'd really help is having
> some sort of lossy storage for foreign-key triggers, where we'd fall
> back to per-block or whole-table rechecking of the constraint instead of
> trying to track the exact rows that were modified. =A0Not sure how you
> apply that to non-FK triggers though.
Err, sorry, I quoted the wrong part. I meant, how would you rlimit
the server memory usage?
...Robert