Github user mxm commented on the pull request: https://github.com/apache/flink/pull/934#issuecomment-125561321 All in all, nice work @kl0u. Do you think we could change this pull request such that we can handle "small/large" or "message/blobcached" accumulators more transparently? It would be great if we encapsulated the two types within one message and only have one method `getUserAccumulators` to retrieve both types. The internal logic could be handled by either the `AccumulatorRegistry` or the `AccumulatorSnapshot` class. That would make the code easier to understand and remove some code from the JobManager and ExecutionGraph classes.
--- If your project is set up for it, you can reply to this email and have your reply appear on GitHub as well. If your project does not have this feature enabled and wishes so, or if the feature is enabled but not working, please contact infrastructure at infrastruct...@apache.org or file a JIRA ticket with INFRA. ---