PG optimization on CentOS Linux file system with SCSI disks

Поиск
Список
Период
Сортировка
От Phoenix Kiula
Тема PG optimization on CentOS Linux file system with SCSI disks
Дата
Msg-id e373d31e0907302244q86f299ayf7406c56ce8063fd@mail.gmail.com
обсуждение исходный текст
Список pgsql-general
I read here in a different context

http://www.redhat.com/docs/manuals/cms/rhea-dpg-cms-en-6.1/ch-config.html

About /etc/security/limits.conf and /etc/sysctl.conf files.

For the last year or so I've had this on a CentOS 32 bit system with
4GB of RAM and SATA II disks, only Postgres relevant stuff mentioned
here:


   # in /etc/security/limits.conf
   postgres        soft    nofile  8192
   postgres        hard    nofile  8192

   # in /etc/sysctl.conf
   fs.file-max = 32768
   kernel.shmmax = 1073741824
   kernel.shmall = 536870912


I am on CentOS 5, 64 bit now with 8GB of RAM and SCSI 15k/rpm disks
with RAID1. Two core2duo CPUs (Intel 5130). Given that this is an
upgrade in pretty much every way, can I increase the values above to
better maximize my PG stuff? Apache/MySQL also run from the same box
(can't afford a dedicated DB server yet), and pretty much the whole
other shebang, but their requirement is not huge. PG is typically the
master of the machine, or at least I'd like to structure it that way.

With that in mind, questions:

1. Can I increase the values in "limits.conf" from 8192 to something
much higher given that these are now SCSI hard disks, or is it totally
separate stuff?

2. I'm thinking of increasing the values in "/sysctl.conf" like this:

   # These values were already here! So high because of 64bit?
   kernel.shmmax = 68719476736
   kernel.shmall = 4294967296

   # Also increase fs.file-max because of SCSI disks?
   # This is the current value, can I increase and expect any
performance gains?
   fs.file-max = 131072

3. In postgresql.conf, increase values accordingly:

    max_connections              = 400
    shared_buffers               = 700MB
    effective_cache_size         = 1.5GB
    work_mem                     = 80MB
    temp_buffers                 = 4096



Will this improve performance?

Thanks.

В списке pgsql-general по дате отправления:

Предыдущее
От: Scott Marlowe
Дата:
Сообщение: Re: why is autovacuum still ongoing even if set to off ?
Следующее
От: Tom Lane
Дата:
Сообщение: Re: Make check fails on 8.3.7