Can you find some relation between the memory usage and insert statements? 9.1.2 has memory problems with inserts (even the simplest ones) on Linux and Windows too, I could produce it. Using pgbench also shows it. Some memory is not reclaimed.
I could produce it also with 8.4.9 on Linux, I haven't tried 8.4.10 yet.
Best regards,
Otto
2011/12/21 Rafael Martinez <r.m.guerrero@xxxxxxxxxxx>
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
On 12/21/2011 12:48 AM, Craig Ringer wrote:Hello
> On 19/12/2011 11:04 PM, Rafael Martinez wrote:
>> Any ideas about why this dramatic change in memory usage when the only
>> thing apparently changed from our side is the postgres version?
>>
> It'd be interesting to know how much of your workload operates with
> SERIALIZABLE transactions, as the behavior of those has changed
> significantly in 9.1 and they _are_ more expensive in RAM terms now.
>
As long as I know, all the databases are using the default, "read
committed".
We have almost 500 databases across all our servers, but we are only
dbas. We provide the infrastructure necessary to run this and help users
when they need it but we have not 100% control over how they are using
the databases ;-)
Version: GnuPG v2.0.14 (GNU/Linux)
regards,
- --
Rafael Martinez Guerrero
Center for Information Technology
University of Oslo, Norway
PGP Public Key: http://folk.uio.no/rafael/
-----BEGIN PGP SIGNATURE-----
iEYEARECAAYFAk7yHHAACgkQBhuKQurGihQz1gCdGJY6vk89lHKMldkYlkxOeJYJ
GSMAoKDRCRo1UpqlUgItzCm/XV9aCbb8
=7f6R
-----END PGP SIGNATURE-----
--
Sent via pgsql-performance mailing list (pgsql-performance@xxxxxxxxxxxxxx)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-performance
