Github user markap14 commented on the issue: https://github.com/apache/nifi/pull/2405 Hey @adfel70 I think I was looking at this PR the wrong way. I didn't notice the addition of the POST request, which is very vital. Because of that, I thought this was making a very large number of requests, but after reviewing the confluent API docs and reviewing the code again it all makes sense. The caching in this layer is different than the caching at the higher layer but was only really necessary for the routine that was there previously, before the POST request. I think your solution handles this case much better than the previous solution. And more importantly your solution appears to be correct :) I was able to test locally and see that if i changed the schema reference to an older version of an existing schema that things work correctly with this PR and did not without the PR. So I've merged this to master. Thanks for the fix and for the patience in getting it all sorted out!
---