[jira] [Created] (FLINK-10413) requestPartitionState messages overwhelms JM RPC main thread

["Zhu Zhu (JIRA)" <ji...@apache.org>]

Zhu Zhu created FLINK-10413:
-------------------------------

             Summary: requestPartitionState messages overwhelms JM RPC main thread
                 Key: FLINK-10413
                 URL: https://issues.apache.org/jira/browse/FLINK-10413
             Project: Flink
          Issue Type: Bug
          Components: Distributed Coordination
    Affects Versions: 1.7.0
            Reporter: Zhu Zhu


We tried to benchmark the job scheduling performance with a 2000x2000 ALL-to-ALL streaming(EAGER) job. The input data is empty so the tasks finishes soon after started.

 

In this case we see slow RPC responses and TM/RM heartbeats to JM will finally timeout.

We find ~2,000,000 requestPartitionState messages triggered by 

triggerPartitionProducerStateCheck in a short time, which overwhelms JM RPC main thread. This is due to downstream tasks can be started earlier than upstream tasks in EAGER scheduling.

 

We's suggest no partition producer state check to avoid this issue. The task can just keep waiting for a while and retrying if the partition does not exist. There are two cases when the partition does not exist:
 # the partition is not started yet
 # the partition is failed

In case 1, retry works. In case 2, a task failover will soon happen and cancel the downstream tasks as well.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)