Hierarchy load causes QlikView to crash

[#1]

Hello,

I have an issue with creating a large hierarchy in QlikView.

My script loads data from Oracle db into a temporary table and then using Hierarchy function to create a tree.

The data has grown to about 6 000 000 rows and now QV stops responding when I try to load the data.

Via debug I found out it's the Hierarchy function that causes QV to crash.

Any tips how can I actually load all the data without leaving anything out?

Is there a limitation somewhere inside QV that causes the application to crash?

Thanks,

Miro

[#2]

Hi Steve,

it is exactly 6 247 804 lines.

There are no rows with PARENTNODE = -1 in the db.

There are 69 rows where PARENTNODE is null.

The null rows are converted to -1 in the TABLEtmp load.

I have accomplished a successful hierarchy load without crash, but it took more than 24 hours.

The hardware I'm using:

CPU: Intel Core i7-3770 @ 3.40 GHz

RAM: 16 GB of RAM (15,8 GB usable)

OS: Windows 7 Professional 64-bit

I need to somehow optimize the load because 24 hours of processing is just too much.

Miro

[#3]

Wow, this is a tough one.

If you have a load model, this is your opportunity to look at the data and see how it is hanging together.

A couple of further questions;  how many records are in the data table that you are connecting to?  And how are you connecting to the other table (a join or simply allowing QV to perform the association)? 

I would strip the null values out of your load, they could cause issues (although I'm not sure 24hr worth of issues but it would be good to eliiminate this issue).

I would question the 6million+ records for the hierarchy; what is this a hierarchy for?  I find it difficult to imagine your company's structure has 6million+ org units.  Also, try using a LOAD DISTINCT on your TABLE load.

One last question, how long does the load take if you use a QUALIFY *; at the beginning of your code?

In terms of hardware, it looks like you got enough memory for a reasonable model.

Keep me updated - hopefully we're getting closer to the issue.

Steve

[staelanicet]

Did you find a solution ?