Github user StephanEwen commented on the pull request:
https://github.com/apache/incubator-flink/pull/254#issuecomment-66356037
The `ConsumableOnceInMemoryOnlyPartitionQueue` uses a bit of an unorthodox
combination of atomics, a blocking queue, and sometimes a lock. I am still
digging through the interactions with the netty channels, but I think in its
current state this one is not quite safe to concurrent operations (by the task
thread that adds / cancels / closes and the netty event loop). May be simpler,
safer, and actually even more efficient to just use a monitor for
synchronization and an ArrayDeque for the buffers.
---
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 [email protected] or file a JIRA ticket
with INFRA.
---