[jira] [Created] (GEODE-2304) Client operation timeouts should be honored by the server instead of the client

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

Udo Kohlmeyer created GEODE-2304:
------------------------------------

             Summary: Client operation timeouts should be honored by the server instead of the client
                 Key: GEODE-2304
                 URL: https://issues.apache.org/jira/browse/GEODE-2304
             Project: Geode
          Issue Type: Improvement
          Components: client/server
            Reporter: Udo Kohlmeyer


On client pools a read-timeout period can be set. The client will enforce the read-timeout period by timing out the operation on the client. This timeout might result in the retrying of the same action on another server.

I think that this timeout should be honored by the servers. A client should provide the timeout as part of the operation event. The server should then track if the operation has completed in that time period. If not and it is possible, the operation should be cancelled. The server can then send an "OperationTimeoutException" back to the client for it to handle.

This way, we potentially end up with less "orphaned" operations in the cluster due to the client read-timeout setting.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)