Till Rohrmann created FLINK-7295:
------------------------------------
Summary: Add callback for proper RpcEndpoint shut down
Key: FLINK-7295
URL: https://issues.apache.org/jira/browse/FLINK-7295
Project: Flink
Issue Type: Sub-task
Components: Distributed Coordination
Affects Versions: 1.4.0
Reporter: Till Rohrmann
Assignee: Till Rohrmann
In order to properly shut down {{RpcEndpoints}} it is necessary to have a
method which is called by the main thread in case of a shut down and allows to
properly close and clean up internal state. At the moment, this clean up work
is done by overriding the {{RpcEndpoint#shutDown}} method which can be called
by a different thread than the main thread. This is problematic since it
violates the {{RpcEndpoint}} contract.
I propose to change the behaviour of {{RpcEndpoint#shutDown}} to be
asynchronous. Calling this method will send a message to the {{RpcEndpoint}}
which triggers the call of the clean up method and the termination of the
endpoint.
In order to obtain the same behaviour as before, the user can obtain the
termination future on which it can wait after sending the request to shut down
the {{RpcEndpoint}}.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
