Re: pg_lock_status() performance

On Tue, Apr 28, 2009 at 5:41 PM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> Merlin Moncure <mmoncure(at)gmail(dot)com> writes:
>> I have a unloaded development server running 8.4b1 that is returning
>> from a 'select * from pg_locks' in around 5 ms.  While the time itself
>> is not a big deal, I was curious and tested querying locks on a fairly
>> busy (200-500 tps sustained)  running 8.2 on inferior hardware.  This
>> returned (after an initial slower time) in well under 1 ms most of the
>> time.  Is this noteworthy?  What factors slow down best case
>> pg_lock_status() performance?
>
>> edit: I bet it's the max_locks_per_transaction parameter. I really
>> cranked it on the dev box during an experiment, to 16384.
>> testing...yup that's it.  Are there any negative performance
>> side-effects that could result from (perhaps overly) cranked
>> max_locks_per_transaction?
>
> [squint...]  AFAICS the only *direct* cost component in pg_lock_status
> is the number of locks actually held or awaited.  If there's a
> noticeable component that depends on max_locks_per_transaction, it must
> be from hash_seq_search() iterating over empty hash buckets.  Which is
> a mighty tight loop.  What did you have max_connections set to?

16384 :D

(I was playing with a function that created a large number of tables/schemas)

merlin