CreateHierarchyByAttribute cause server to crash

[Francois]

Hi all

We recently upgraded a local PA installation to TM1 2.0.97.6. We hoped to use the hierarchy create by attribute function via TI or through PAW, but each time it's created the server crash. This was supposed to be fixed in 2.0.9.1 LC. https://www.ibm.com/support/pages/syste ... 91v5_0.pdf

Server : Microsoft Windows Server 2016.
Server type : VM
TM1 : 2.0.97.6
PAW : 2.0.63

Any idea what I'm doing wrong other than an upgrade?

Tx.

[Wim Gielis]

Hello,

Could you reproduce the behavior on any dimension (even new, small, dimensions) that have a suitable attribute ?
Is it done in TI or manually ? If TI, what's the function call you use ?

[tony.wijaya]

Hi,

We encountered same issue before with PA 2.0.9.8 and PAW 2.0.64.

For our case, we noticed that the affected dimension had a lot of complex attribute rules which apparently cause the server to crash every time we created hierarchy (either from CreateHierarchyByAttribute or from PAW front end). In the end, we managed to resolve the issue by simplifying the attribute rule and converting the rule to TI.

I would suggest you to check if you have attribute rules in the affected dimension and try removing the rule to test it out.

[lotsaram]

I would as a rule, always avoid rules in attribute cubes

I would also ever using the CreateHierarchyByAttribute function. This is IMO pure demoware. Much better to create your own TI to maintain attribute based hierarchies (or use the bedrock one).

[Francois]

Hello,

Could you reproduce the behavior on any dimension (even new, small, dimensions) that have a suitable attribute ?
Is it done in TI or manually ? If TI, what's the function call you use ?

Wim, I did that straight after I posted the question. To my surprise, it did work through a TI on a small dimension with 5 elements and no rules. The dimension I used it on was on the IBM blueprint time dimension, the one with the many complex rules. We need to create alternate hierarchies for reporting on a calendar and fiscal year. My next course of action will be to remove the rules and replace attribute maintenance with a TI.

Thanks for the help.

[Francois]

For all of you who are interested, createhierarchybyattribute should be used with care and demo purposes. This is limited to a more once-off use case rather than using it on a reporting hierarchy that is ever growing in size. In a production environment, creating and maintaining a virtual hierarchy, one needs to consider the size and maintenance frequency of a dimension.

The best way to do is by:

• Using HierarchyExist and HierarchyCreate to create hierarchy and HierarchyDeleteAllElements to clean the set

• Assign a Subset as a datasource as opposed to looping through the dim

• Insert elements into the hierarchy using HierarchyElementInsert based on a attribute or another lookup

• Link children to parents using HierarhcyElementComponentAdd

[gtonkin]

Not sure that I would use HierarchyDeleteAllElements, would rather unwind the hierarchy like you would a rollup.

[Francois]

Not sure that I would use HierarchyDeleteAllElements, would rather unwind the hierarchy like you would a rollup.

Hi gtonkin, yes you are right but in my case, all elements will be mapped in the hierarchy in every run. Those with no attributes will default to the 'Other' consolidator. As data is stored on a leaf and HieararchyDeleteAllElements does not delete any leafs this option achieves the same result.