Re: measure database contention

On Wed, Dec 17, 2008 at 2:34 AM, Albe Laurenz <laurenz(dot)albe(at)wien(dot)gv(dot)at> wrote:
> Jaime Casanova wrote:
>> we have a some bad queries (developers are working on that), some of
>> them run in 17 secs and that is the average but when analyzing logs i
>> found that from time to time some of them took upto 3 mins (the same
>> query that normally runs in 17secs).
>>
>> so my question is: how could i look for contention problems?
>
> A good first step is to identify the bottleneck.
>
> Frequently, but not always, this is I/O.
> Do you see a lot of I/O wait? Are the disks busy?
>

the disks are an RAID 10 with 4 sata disks of 15000rpm
and nop iostat reports avg of 0.12 iowait

> I don't know anything about your system, but I once experienced a
> similar problem with a 2.6 Linux system where things improved considerably
> after changing the I/O-scheduler to "elevator=deadline".
>

i don't understand I/O-schedulers at all... anyone knows what is the
recommended for postgres?

--
Atentamente,
Jaime Casanova
Soporte y capacitación de PostgreSQL
Asesoría y desarrollo de sistemas
Guayaquil - Ecuador
Cel. +59387171157