I am trying to iron out my understanding of how replacement pipelines for ontology objects work. Recently, I made changes to the schema (deleted properties) and backing dataset of of a streaming object. Once I saved my changes, I saw that a replacement pipeline kicked off and my live pipeline was running as usual. The schema status of the object was changed to updating as well. My understanding is that the live pipeline should still be able to serve all active workflows as usual until the replacement pipeline completes. However, we saw a number of OSDK based apps failing since they couldn’t load the objects (the root cause error was that it was unable to find the deleted properties). This doesn’t make sense to me and I’m wondering what the best practice is here when pushing schema changes to objects.
In this particular case, it was a streaming backed object and the replacement pipeline takes around 10-12 hours to run fully so its not really easy to find a downtime window where we can push this change.