PDA

View Full Version : Trouble with Analysis Examples - Slice and Dice



emersonsz
01-25-2006, 04:22 AM
I installed Pentaho in Tomcat 5.5.9 on Windows Xp.

It's working, but I can't access Analysis Examples.

The databases (SampleData, Shark, Hibernate, Quartz) were open sucessfully.

When I try access in [Pt 0] Slice and Dice :

pt_418] Quadrant Headcount Actual and Budget
[pt_427] This page cannot be accessed directly. It must be linked to from other pages

If necessary, I sent my Server.Xml

Thanks,

Emerson.

Obs : I hope change my actual solution (MS Analysis Services, Crystal Analysis, Crystal Reports) by Pentaho.

jdixon
01-25-2006, 08:07 AM
The Analysis page needs to have a starting point provided so that it knows which datasource, and query to start with.

You get this error when the Analysis page does not know where to start.

The Analysis Examples link should give you a url such as this:

http://localhost:8080/pentaho/ViewAction?&solution=samples&path=analysis&action=query1.xaction

The Analysis component will redirect the browser to the Pivot page with a URL such as this:

http://localhost:8080/pentaho/Pivot?solution=samples&path=analysis&action=query1.xaction

If the Pivot page is unable to store attributes in the user session you will also see this error when you try to do anything after the Pivot page is first displayed.

James

emersonsz
01-25-2006, 08:36 AM
Hi James,

thanks for the answer!

How can I indicate to the Analysis page where to start ?

I used "Advanced_Install_Guide_1.0.0.pdf" to install Pentaho.
About access to data there indicate :
- Copy hsqldb.jar to Tomcat\\common\\endorsed directory. Stop and Start Tomcat - ok
- Include Datasource Definitions in the Server.Xml - ok

Are there something else configuration to do ?

Thanks a lot.

Emerson - GCI

jdixon
01-25-2006, 09:27 AM
It is specified on the URL:

e.g.

http://localhost:8080/pentaho/Pivot?solution=samples&path=analysis&action=query1.xaction

This says to use samples/analysis/query1.xaction from the pentaho-solutions folder. If you look at this file you can see that the datasource and MDX query are specified in the file.

James

andydempsey
05-22-2006, 11:37 AM
Is there a session timeout or similar that could cause this page is not accessible error?

I receive it when I drill following a slow query (35 minutes say) or similar. The drill which caused the long query worked fine, results correct, displayed correctly, but I then cannot do anything else.

guzaldon
05-22-2006, 12:21 PM
boy 35 minutes. that's a long time. is it being mapped to the OLTP? or does your schema have more than 20 or so dimensions to join acrost?

I noticed that mine with about 3 million rows and about 8 dimensions only take 35 seconds tops. the reason we are building a data warehouse and using pentaho is so we don't have to wait 30 minutes to 3 hours to get results?

just my 2 cents. Sorry it's not much help, but I would image that were would be some sort of timeout period. have you tried staying active in some part of the session while you wait for the query to finish?

Nic

andydempsey
05-22-2006, 12:28 PM
No, teeny-tiny DW really, just slightly poor SQL running against a slightly poor database on my laptop with everything else. Performance is not my interest at this time, I am focusing on the flexibility of mondrian.

As for DW size, not that I'm into comparing measurements, but I regularly work on DW in the region of 10TB plus, with market basket style analysis that can take in excess of 5 hours on a teradata box. It gets even more fun when you start including SQL generated from PMML output.

35 minutes is very quick for some systems. Piss poor for my testing though ;-)

guzaldon
05-22-2006, 12:47 PM
Oh I See what your saying. I'm actually a pretty new comer to the whole Data warehousing idea and the such, but I try to add my 2 cents but sometimes is not even worth 2 cents. and I'm really only building the first one in my experience so I guess that EXP showed in my last comment. But I'm trying to learn, and that's what makes this challege so much fun.

Our OLTP isn't even over 12 gigs of data :-( You got us beat with your 10TB, that's pretty sweet though. I must say, you gotta know what your doing.

But I see your point and you'll probably able to beat the crap out of mondrian compared to what I could only dream of.

Heck I don't even know what a PMML output is.

but the guys here at pentaho I would hope should be able to give you some insight, compared to my meger 2 cents. but on the other hand I have helped a few people get up and running on mondrian so I guess I have learned a bit.

the time out could be something that needs to be configured in jboss as well, but I'm not sure just a couple more pennies.

Don't have too much fun,

Nic

andydempsey
05-22-2006, 12:51 PM
PMML = predictive modelling markup language, or in laymans terms XML for defining data mining algorithms.

As for mondrian - I know squat currently. I am expert in other enterprise class BI tools and need to round myself out with something like pentaho.

guzaldon
05-22-2006, 01:09 PM
so PMML would be like the mondrian.xml file on crack? thanks for the info.

One of the things I like most about the mondrian is that you only have to change your mapping to the db in one place and everything else can stay the same. then you can put the {var} in there to allow for all sorts of extra fun stuff when using the widgets and that sort of and allowing extra flexability. which you probably saw in the DEMO.

mbatchelor
05-22-2006, 03:28 PM
In the web.xml, you'll find the session timeout defined as follows:

<session-config>
<session-timeout>30</session-timeout>
</session-config>


As for the performance, well - I've worked with Teradata before - many of the performance issues were due to the lack of join indexes in specific places. As I recall, they don't (or at least didn't) have materialized views. But, everything was designed with partitioning in mind - usually across dates.

In those consultancy engagements, simply giving the poor-performing SQL queries to the DBA was sufficient to get the join indexes straight so that a 30-minute query became a 20-second query - yes, that happened on a few occasions (multiple terabyte DW). Occasionally, the order of items in the where clause had to be tweaked such that the partitioned entity appeared first (it seemed to be important to the optimizer).

Anyway, you probably know all that (and more) if you're entrenched in a Teradata DW project. That's my experience anyway.

Take care,

Marc