I guess you are indicating the data shuffle process among different task managers.
While task manager(TM) registering itself to the job manager(JM), it also carries the infos of ip address and data port that it listens to. During the process of scheduling tasks, the upstream TM's address info(ip, port) would be covered inside the data structure of task deployment descriptor for respective downstream tasks. Then the downstream tasks can connect to the remote upstream TM to request data. In short words, JM knows all the addresses of TMs via registration, then these addresses would be sent to the required peers during task schedule and deployment. Best, Zhijiang ------------------------------------------------------------------ From:KristoffSC <krzysiek.chmielew...@gmail.com> Send Time:2020 Feb. 26 (Wed.) 19:39 To:user <user@flink.apache.org> Subject:Re: How JobManager and TaskManager find each other? Thanks all for the answers, One more question though. In [1] we can see that task managers are talking with each other - sending data streams. How each task manager knows the address of other task managers? [1] https://ci.apache.org/projects/flink/flink-docs-release-1.10/concepts/runtime.html#job-managers-task-managers-clients -- Sent from: http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/
