A deployment of mine is hitting the cassandra.max_queued_native_transport_requests limit quite often. I'd like to trace which batch of requests caused Cassandra to go over the limit, by adding some logs to Cassandra 2.2. I was considering setting some thread local flag when we block and verifying it when we execute the request's corresponding Message, but the maze of work stealing in SharedExecutorPool is confusing me. Where could I plug in to log, for example, the keyspace and table name of any request that increased the queued tasks over the limit?
