From: | Joe Conway <mail(at)joeconway(dot)com> |
---|---|
To: | "Hackers (PostgreSQL)" <pgsql-hackers(at)postgresql(dot)org> |
Subject: | memory leak regression 9.1 versus 8.1 |
Date: | 2012-05-09 20:36:25 |
Message-ID: | 4FAAD549.2070107@joeconway.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
I'm working on an upgrade of PostgreSQL embedded in a product from
version 8.1.x to 9.1.x. One particular PL/pgSQL function is giving us an
issue as there seems to be a rather severe regression in memory usage --
a query that finishes in 8.1 causes an out of memory exception on 9.1.
Using the same data on the same machine I see memory use stay steady at
a reasonably low value on the 8.1 installation but steadily climb on 9.1
(I watched it go over 2 GB and canceled the query -- the production
machines are 32 bit)
The attached standalone script seems to reproduce the effect. On 8.1
memory usage remains steady and low, on 9.1 I watched it climb past 1.1
GB and canceled the query.
I suspect the append node to be the culprit because if I skip the "UNION
ALL", i.e. if I use one generate_series with 20 million rows instead of
10 with 2 million each, then I do not see the memory leak. The real
function is actually selecting over many inherited tables (i.e. a
partitioned table).
Thoughts?
Thanks,
Joe
--
Joe Conway
credativ LLC: http://www.credativ.us
Linux, PostgreSQL, and general Open Source
Training, Service, Consulting, & 24x7 Support
Attachment | Content-Type | Size |
---|---|---|
9.1-memory-leak-regression.sql | text/x-sql | 5.2 KB |
From | Date | Subject | |
---|---|---|---|
Next Message | Josh Berkus | 2012-05-09 21:23:30 | Re: synchronous_commit and remote_write |
Previous Message | Michael Nolan | 2012-05-09 18:51:54 | Re: problem/bug in drop tablespace? |