How to deal with Business Data Changes and OpenSearch Index in Axon Ivy 12?

mariano_froh

Dear Axon Ivy Community,

I have a question regarding OpenSearch indexing of Business Data after making changes to Business Data Classes.

According to the documentation, adding new attributes to a Business Data Class is supported and seems to work without any issues in our case as well.

However, we run into problems when:

changing the type of an existing attribute
removing an existing attribute

Our understanding is that this likely causes inconsistencies in the OpenSearch index, especially because older cases still contain Business Data persisted with the previous model structure.

Our scenario

We use the Engine with Bundled OpenSearch
We make use of the Business Case Data concept described here
We have old cases in the database, most of them already completed.
During development, it would be acceptable for us to delete these old cases if they are the root cause of indexing issues.
What we want is that new cases persist Business Data with the updated model and are indexed correctly in OpenSearch.

Questions

What is the recommended way to rebuild or reset the OpenSearch index after incompatible changes to Business Data Classes?
Is deleting old cases sufficient to resolve such issues, or can outdated Business Data still remain somewhere in the system?
When deleting cases via API, are the associated Business Case Data entries also removed, or do they persist independently?
Are there any best practices for evolving Business Data Classes (especially type changes or attribute removals) while avoiding indexing problems?

I’d appreciate any recommendations or insights into how others handle this during development and deployment.

Thanks in advance!

Christian Strebel

From our understanding:

Questions 1 and 2
The easiest solution is often to rename the BusinessData Class (for example by adding a postfix or changing the package name).

This separates old and new data models cleanly:

new cases use the new class and are indexed independently
old data can still be queried as long as the old classes remain available

The limitation is that you cannot query old and new data together in a single API call, since they are treated as different Business Data Classes.

Question 3
Deleting cases removes the entries in IWA_BusinessCaseData, i.e. the linkage between case and Business Data.
However, the actual Business Data in IWA_BusinessData remains unless it is explicitly deleted. That means it also remain in the search index.

Question 4 / Best practices

Prefer additive changes only
Avoid changing types of existing attributes
If a field must change incompatibly, create a new attribute and phase out the old one instead of reusing the same name with a different type

What we are not fully aware of yet is the exact impact of removing an attribute.
For type changes, the issue is clear. For removed attributes, I assume the main risk is that older persisted objects and index entries still contain fields that no longer exist in the current class definition.