Re: Geoserver-PostGIS performance problems - Mailing list pgsql-performance

From Riaan van den Dool
Subject Re: Geoserver-PostGIS performance problems
Date
Msg-id 500E7478020000DB000A78E9@pta-emo.csir.co.za
Whole thread Raw
In response to Re: Geoserver-PostGIS performance problems  (Brett Walker <brett.walker@geometryit.com>)
Responses Re: Geoserver-PostGIS performance problems
List pgsql-performance

Thank you for this response.

It turns out our performance problems were solved when I switched off 'Prepared statements' in Geoserver for the PostGIS data store. It makes quite a huge difference.

Riaan

>>> Brett Walker <brett.walker@geometryit.com> 7/24/2012 09:54 AM >>>

Hi Riaan,

I am familiar if the Geoserver/Geotools package and I believe that the problem is not Postgres/PostGIS but rather Geoserver.

The DB Connection Closed message is not sent at the end of the query, but rather at the end of the rendering. There is more than just querying happening between the Select message and the Closed message.

Brett

From: pgsql-performance-owner@postgresql.org [mailto:pgsql-performance-owner@postgresql.org] On Behalf Of Riaan van den Dool
Sent: Tuesday, 24 July 2012 5:21 PM
To: pgsql-performance@postgresql.org
Subject: [PERFORM] Geoserver-PostGIS performance problems

 

 

Hi

 

We have noticed something strange in the interaction between our Geoserver instance and Postgres/PostGIS.

 

After setting Geoserver's log level to include developer debugging, I managed to capture a single request from Geoserver WMS to PostGIS.

 

The (shortened) sequence of events and their timestamps:

 

12:31:22,658 - SELECT query for MSG is sent to Postgres

12:32:10,315 - Rendering for MSG layer starts

12:32:10,356 - DB Connection Closed

========

~ 48 seconds

 

Interestingly enough, when I execute the same query (MSG) directly from PgAdmin3:

 

SELECT "frp_mw",encode(ST_AsBinary(ST_Force_2D("the_geom")),'base64') as "the_geom" FROM "public"."af_msg_abba_datetime_today" WHERE ("the_geom" && GeometryFromText('POLYGON ((-27.67968749408379 -46.92207325648429, -27.67968749408379 -6.186892358058866, 75.67968748740275 -6.186892358058866, 75.67968748740275 -46.92207325648429, -27.67968749408379 -46.92207325648429))', 4326) AND (("frp_mw" >= -1 AND "frp_mw" <= 150) OR ("frp_mw" >= 151 AND "frp_mw" <= 300) OR ("frp_mw" >= 301 AND "frp_mw" <= 600) OR ("frp_mw" >= 601 AND "frp_mw" <= 50000)));

 

I get 6515 rows in 380 ms.

 

Ie Postgres is able to return the results of the query within 380ms if queried from PgAdmin3 but Geoserver takes about 48 seconds to get hold of the same resultset.

 

Is this some kind of JDBC problem perhaps?

 

Some details about our setup:

 

Master Postgres database is on a separate VM from Geoserver, but we replicate to a slave Postgres cluster on the Geoserver VM (same host). So Geoserver is referencing the 'localhost' read-only Postgres cluster for its queries.

 

The 380 ms response time shown above was from the slave Postgres cluster, same one that Geoserver is using.  

 

All Linux (Ubuntu 11.10) based. Postgres 9.1 PostGIS 1.5 Geoserver 2.1.3

 

Riaan






--
This message is subject to the CSIR's copyright terms and conditions, e-mail legal notice, and implemented Open Document Format (ODF) standard.
The full disclaimer details can be found at http://www.csir.co.za/disclaimer.html.


This message has been scanned for viruses and dangerous content by MailScanner,
and is believed to be clean.


Please consider the environment before printing this email.


--
This message is subject to the CSIR's copyright terms and conditions, e-mail legal notice, and implemented Open Document Format (ODF) standard.
The full disclaimer details can be found at http://www.csir.co.za/disclaimer.html.


This message has been scanned for viruses and dangerous content by MailScanner,
and is believed to be clean.


Please consider the environment before printing this email.

--
This message is subject to the CSIR's copyright terms and conditions, e-mail legal notice, and implemented Open Document Format (ODF) standard.
The full disclaimer details can be found at http://www.csir.co.za/disclaimer.html.


This message has been scanned for viruses and dangerous content by MailScanner,
and is believed to be clean.


Please consider the environment before printing this email.

pgsql-performance by date:

Previous
From: Brett Walker
Date:
Subject: Re: Geoserver-PostGIS performance problems
Next
From: Jim Vanns
Date:
Subject: Re: Odd blocking (or massively latent) issue - even with EXPLAIN