Re: Slow query (planner insisting on using 'external merge' sort type) - Mailing list pgsql-performance

From Igor Neyman
Subject Re: Slow query (planner insisting on using 'external merge' sort type)
Date
Msg-id A76B25F2823E954C9E45E32FA49D70ECCD489DB0@mail.corp.perceptron.com
Whole thread Raw
In response to Re: Slow query (planner insisting on using 'external merge' sort type)  (Ian Pushee <ian@intuvisiontech.com>)
Responses Re: Slow query (planner insisting on using 'external merge' sort type)
List pgsql-performance

-----Original Message-----
From: pgsql-performance-owner@postgresql.org [mailto:pgsql-performance-owner@postgresql.org] On Behalf Of Ian Pushee
Sent: Friday, June 19, 2015 10:54 AM
To: pgsql-performance@postgresql.org
Subject: Re: [PERFORM] Slow query (planner insisting on using 'external merge' sort type)



On 6/19/2015 10:46 AM, Igor Neyman wrote:
>
> Probably events_confidnce index is not very selective, that's why optimizer prefers seq scan.
> I'd try to create an index on (name, eventspy_id, camera_id, type, status).
>
> Also, the recent 9.2 is 9.2.13, you should upgrade.
>
> Regards,
> Igor Neyman

Hi Igor,

I already have an index for (name, eventspy_id, camera_id, type, status)... that is the index being used (apparently
silently)when I set random_page_cost=1.0.
 

Thanks,
-Ian


--

Well, having 8GB Ram on the machine you probably should not be using default config parameters.
Depending on what else is this machine is being used for, and depending on queries you are running, you should
definitelymodify Postgres config.
 
If this machine is designated database server, I'd start with the following parameters modified from default values:

shared_buffers = 1024MB
temp_buffers = 8MB
work_mem = 64MB                
effective_cache_size = 1024MB
random_page_cost = 2.5
cpu_tuple_cost = 0.03
cpu_index_tuple_cost = 0.05

and see how it goes.

Regards,
Igor Neyman




pgsql-performance by date:

Previous
From: Ian Pushee
Date:
Subject: Re: Slow query (planner insisting on using 'external merge' sort type)
Next
From: Igor Neyman
Date:
Subject: Re: Slow query (planner insisting on using 'external merge' sort type)