Jinglun created HDFS-15899:
------------------------------
Summary: Remove rpcThreadPool from DeadNodeDetector.
Key: HDFS-15899
URL: https://issues.apache.org/jira/browse/HDFS-15899
Project: Hadoop HDFS
Issue Type: Improvement
Reporter: Jinglun
The DeadNodeDetector uses a thread pool to do all the probe rpc calls. The
purpose is to use the thread pool timeout to monitor the probe timeout. But the
rpc client already has a timeout. We can use the rpc client timeout instead of
the thread pool timeout and remove the rpcThreadPool.
The rpcThreadPool introduces additional complexity for probing the DataNode.
The probe task waiting in the busy rpcThreadPool might exceed the configured
timeout. The probe task will be marked as failed even it is not scheduled.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-dev-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-dev-h...@hadoop.apache.org
