The following bug has been logged on the website:
Bug reference: 7889
Logged by: Rodrigo Garcia
Email address: rodrigo.garcia@cross-join.com
PostgreSQL version: 9.2.2
Operating system: Darwin 12.2.1 - OSX Mountain Lion
Description: =
Recently we've migrated from Oracle 10g to PostgreSQL 9.2.2. =
Our problem is related to memory consumption, we've a mac mini server =
mountain lion server with 16 gb of RAM, each PostgreSQL server process is
consuming the RAM defined in shared_buffers (shouldn't this be a shared
memory???) resulting in the following memory consumption (top command):
PhysMem: 2438M wired, 5816M active, 7663M inactive, 16G used, 458M free
We're forced to detected when the memory is below a threshold and invoque
the "purge" command at OS level. This brings performance issues because
we're clearing out not only OS cache but also disk cache data that we think
PostgreSQL uses.
Here is the top -o rsize in the system (shared_buffers =3D 512MB,work_mem =
=3D
32MB,maintenance_work_mem =3D 256MB):
PID COMMAND %CPU TIME #TH #WQ #PORT #MRE RPRVT RSHRD RSIZE
4125 postgres 0.0 12:08.75 1 0 9 81 25M 558M 575M =
86560 postgres 0.0 34:43.97 1 0 9 81 25M 558M 575M =
38534 postgres 32.9 11:03.16 1/1 0 9 80 26M+ 558M 575M+
57374 postgres 0.0 17:00.25 1 0 9 81 25M 558M 572M =
84932 postgres 8.6 06:33.90 1 0 9 77 21M 558M 564M =
61888 postgres 0.0 06:43.98 1 0 9 77 21M 558M 564M =
14401 postgres 0.0 03:58.94 1 0 9 77 21M 558M 564M =
84952 postgres 5.3 06:36.73 1/1 0 9 77 21M 558M 564M+
84927 postgres 0.0 06:09.82 1 0 9 77 20M 558M 563M =
84936 postgres 0.0 04:04.06 1 0 9 77 19M 558M 562M =
73230 postgres 0.0 12:37.65 1 0 9 59 1360K 558M 538M =
73229 postgres 0.0 02:06.01 1 0 9 63 5568K 558M 512M =
477 Finder 0.1 12:08.66 5 4 222+ 777+ 391M- 10M- 418M+ =