I recently updated a CDT, which is referenced in a process model subprocess. The

The problem text on the generated alert reads: "The sub-process could not be started because the input mapping contained process parameters with incorrect types. "

The recommendation reads: "Modify the input mapping and start a new instance of the sub-process node."

I hope I don't have to do that for all processes that reference the previous version of the CDT!

I tried modifying the input mapping on the sub-process node, but there is no way to map the existing process variables to the latest type of CDT expected by the sub-process node.

I did notice in the XSD for the updated CDT, that besides the updated VARCHAR length there was a differing <xsd:schema/> parameter.

Instead of reading "xmlns:types1=..." like on the DEV server, it reads: "xmlns:types2=..." on PROD.

But all the CDT's seem to have similar variations.

CDTs should not be passed by reference as they result in the issues that you are now experiencing. Appian's documentation mentions this specifically. Here is the reference - forum.appian.com/.../Sub-Process_Activity

Check the Restrictions section in the above link.
There are quite a few earlier posts related to this. Here is one of those reference posts - forum.appian.com/.../

It's odd that I haven't experienced this before. Can I change the CDT XSD back to avoid problems during a push?

Out of curiosity, did you run impact analysis on the updated CDT? (i guess this wouldn't apply if you're talking about running instances, but I'm unclear as to whether you are or not)

I always run impact analysis, and then update. In addition to that I usually check the application dependencies to make sure there isn't a dependency on a previous version of the CDT: someCdt^7 for instance.

This particular application relies on CDT's by reference because the process contains inner loops: one Review to many Review Stages, one Stage to many Cycles.

It's absolutely necessary to encapsulate the stages and cycles as sub-processes, and it's also necessary that the base process model be updated real time with the changes to the CDT's from each cycle.

When I decided to pass by reference, it seemed like the right decision because of the nature of this particular business process and because the CDT's can be large and I wanted to conserve memory usage.

Fortunately these CDT's are pretty static. I hope I can reverse the changes by reverting the XSD's or I'm going to have a very big problem with production processes.

I wish the warning not to pass CDT's by reference wasn't buried in the documentation. I went to training AND read through the docs when I started developing on the platform, and still somehow missed this.