Re: postgres bogged down beyond tolerance

Tena Sakai wrote:

> I checked kernel parameter
> shmmax and it was set as 33554432. I "fixed" it as
> suggested by the manual:
> http://www.postgresql.org/docs/8.2/static/kernel-resources.html

Note that just by changing shmmax you're not doing anything to Postgres
itself. If you want Postgres to use more memory, you need to increase
shared_buffers in postgresql.conf. Changing shmmax is only required so
that the kernel allows Postgres to allocate all those shared_buffers.

> Namely, I shutdown the database, issued two commands:
> /sbin/sysctl -w kernel.shmmax=134217728
> /sbin/sysctl -w kernel.shmall=2097152
> and rebooted the computer.
>
> After it came up, I checked the shmmax and it is set
> as 33554432. Which surprised me. Since I used -w
> flag, I thought it should've written to /etc/sysctl.conf,
> but there is no such entry at all and the data of this
> file is from 2006.

No, the -w actually means "write to the kernel", which is state that
doesn't persist by itself. It's your own responsability to write it to
sysctl.conf.

Anyhow, I don't think any of this has anything to do with a performance
problem. If you haven't ever vacuumed your database, do so now and try
again. (Actually, have a read of the "maintenance" chapter in the
manual).

--
Alvaro Herrera Valdivia, Chile ICBM: S 39º 49' 18.1", W 73º 13' 56.4"
Voy a acabar con todos los humanos / con los humanos yo acabaré
voy a acabar con todos / con todos los humanos acabaré (Bender)