Horton

Re: Re: [COMMITTERS] pgsql: Fix plpgsql to release SPI plans when a function or DO block is

Jan Urbański <wulczer@wulczer.org>

View thread
From: Jan Urbański <wulczer@wulczer.org>
To: Stephen Frost <sfrost@snowman.net>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, David Fetter <david@fetter.org>, PG Hackers <pgsql-hackers@postgresql.org>
Date: 2011-03-30T18:21:05Z
Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  1. Fix plpgsql to release SPI plans when a function or DO block is freed.

Attachments

On 28/03/11 17:25, Stephen Frost wrote:
> * Jan Urbański (wulczer@wulczer.org) wrote:
>> On 28/03/11 04:31, Tom Lane wrote:
>>>> Do the other PLs we ship need similar fixes?
>>>
>>> Offhand I think the other PLs leave management of prepared plans to the
>>> user.  If there are any places where they cache plans behind the scenes,
>>> maybe a similar fix would be appropriate.
>>
>> FWIW I executed
>>
>> do $$ plpy.execute("select 1 from pg_class") $$ language plpythonu;
>>
>> 10k times in a session and the backend grew a lot in memory and never
>> released it. I can't offhand see where the memory went, I'll try to
>> investigate in the evening.
> 
> I'm about 90% sure that they all have this problem..  I havn't had a
> chance to look at how Tom fixed pl/pgsql (I didn't think it'd be easy to
> do w/o coming up with a way to explicitly tell the PL to release
> something) so perhaps I'm mistaken, but they all share very similar
> code..

Valgrind showed me the way. PFA a trivial patch to avoid leaking a
PLyProcedure struct in inline blocks.

Prepared plans get cleaned up currectly, the SPI plan is deallocated
when the Python plan object is garbage collected.

Cheers,
Jan