PDA

View Full Version : Queries against Mondrian much slower in RC2



tdidomenico
09-26-2007, 04:32 PM
Greets.

I've recetly installed RC2 (was using RC1), and I've noticed that my queries against Mondrian are much slower than with RC1. Just to be sure I reinstalled RC1 and everything was back to normal.

Can anyone suggest a way of analyzing this further? I assume the issue is Mondrian related because I see Postgres working at full capacity for much more time than before.

Also, was Mondrian 2.4 included in RC2? Could this have something to do with it?


Cheers!

wgorman
09-27-2007, 10:37 AM
Hello,

RC1 uses Mondrian 2.3.2, RC2 uses Mondrian 2.4.1. The issue is most likely related to the upgrade. Mondrian 2.4.2 is out now, this may or may not solve your issue. To upgrade to Mondrian 2.4.2, simply download mondrian (http://sourceforge.net/projects/mondrian) and place the new mondrian.jar file in the pentaho.war/WEB-INF/lib directory.

If this doesn't solve your problem, there are two steps to enabling logging in Mondrian:

First, modify the pentaho-solutions/system/mondrian/mondrian.properties. Set the mondrian.trace.level= 1. That will add Mondrian's SQL statements it generates to standard out of the server.

Second, modify the jboss/server/default/conf/jboss-log4j.xml file. Set the mondrian category priority value to "DEBUG".

If you send us the logging information along with your mondrian schema, that will help us track down what the performance issue might be. Thanks for your help!

Will

tdidomenico
09-27-2007, 11:00 AM
Thank you for your quick and detailed reply, Will.

I'm actually using a manually deployed installation of Pentaho on Tomcat, but I'll have no problem finding the right files to edit and/or replace.

I'll be testing this in the next couple of days. I'll post my findings in this thread.

tdidomenico
09-27-2007, 12:54 PM
Ok, I've run some quick tests: I basically replaced mondrian.jar and tried the same query with 2.3.2, 2.4.1 and 2.4.2.

2.3 takes 17 seconds, while 2.4.x take over 4 minutes.

Before posting my schema and logs, I was wondering if this problem could be related to the loss of caching being discussed in the mailing list, here's the link:

http://forums.pentaho.org/showthread.php?t=56873

Thanks!

wgorman
09-28-2007, 04:47 PM
The cache they are discussing creating hasn't been created yet, in either 2.3 or 2.4, so that shouldn't be related. Please post more details and I'll make sure we dig into it more.

Thanks!

Will

tdidomenico
09-28-2007, 05:45 PM
I'm attaching my Mondrian schema and two log files: one when running the query with 2.3.2, and the other one when running it with 2.4.2 (2.4.1 is just the same).

I was a little lost when trying to figure out what part of the log would be useful. If what I'm sending doesn't do it, please give me some hints to improve the selection.

Thank you!

tdidomenico
10-19-2007, 03:01 PM
Just for the record, and in case anyone is still looking into this, 1.6GA is running smoothly and my queries are back to normal.

I'm still trying to understand what happened in RC2; I'll post my findings here if there are any...

Thanks!

wgorman
10-25-2007, 09:33 AM
In 1.6 GA, we made the decision to go forward with Mondrian built in JDK 1.5, vs a retroweaved version of Mondrian that is compatible with 1.4. This might explain the issues you were facing.

Will

tdidomenico
10-25-2007, 09:46 AM
Good to know, Will. Thank you!