Re: For index bloat: VACUUM ANALYZE vs REINDEX/CLUSTER - Mailing list pgsql-general

From Sander Steffann
Subject Re: For index bloat: VACUUM ANALYZE vs REINDEX/CLUSTER
Date
Msg-id 006c01c7f9f2$1b4a1430$51de3c90$@steffann@computel.nl
Whole thread Raw
In response to Re: For index bloat: VACUUM ANALYZE vs REINDEX/CLUSTER  ("Phoenix Kiula" <phoenix.kiula@gmail.com>)
List pgsql-general
Hi,

> Can I add SHM with merely by managing the entry in sysctl.conf? My
> current values:
>
> kernel.shmmax = 536870912
> kernel.shmall = 536870912
>
> My "shared_buffers" in postgresql.conf is "20000". From the website
> http://www.desknow.com/kb/idx/12/061/article/ I notice that shmmax
> should be sharedbuffer*8192, so I suppose my shmmax can be much lower
> than the above, but I raised it for performance. Am I wrong to do so?

You need to configure the kernel so it allows processes to use more shared
memory. This does not mean that a process automatically uses it. For
PostgreSQL you will need to increase shared_buffers to make it use the extra
available shared memory. With your shared memory settings you can probably
increase shared_buffers to about 65000.

With the 'ipcs' command you can see how much shared memory PostgreSQL uses.
Look under 'Shared Memory Segments' to memory owned by user postgres.

- Sander



pgsql-general by date:

Previous
From: Devrim GÜNDÜZ
Date:
Subject: New PostgreSQL RPM sets are available for Fedora / RHEL
Next
From: Ron Johnson
Date:
Subject: Re: Recall: August Monthly techdata split fi