Hi, all:
When we consume multiplue topic in one flinkKafkaConsumer, each topic
partition count is less than the parallelism of the consumer, there is a
problem we encounted is that, currently the partitionAssigner can evenly assign
partition of one topic to subtask, but the assignment process is indepent
between topic, so finally , we get some subtasks is in charge of much more
partitions while some subtask is total free! To slow this issuse, i have
reported the issue by https://issues.apache.org/jira/browse/FLINK-22840.
I worked on this issue by add one assign strategy wich can deal with this
requirement , and want to contibue to flink guys!
Is there someone can talk about this issuce with me ?
--
顺祺！
徐小龙
同济大学软件学院
上海市曹安公路4800号济事楼508 201804
电话/传真：13671633451
E-MAIL:xiaolongsy2...@163.com
