You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@ratis.apache.org by "Lokesh Jain (JIRA)" <ji...@apache.org> on 2018/07/09 05:56:00 UTC

[jira] [Created] (RATIS-268) Handle committed client requests on leader crash and shutdown

Lokesh Jain created RATIS-268:
---------------------------------

             Summary: Handle committed client requests on leader crash and shutdown
                 Key: RATIS-268
                 URL: https://issues.apache.org/jira/browse/RATIS-268
             Project: Ratis
          Issue Type: Bug
            Reporter: Lokesh Jain
            Assignee: Lokesh Jain


Currently when a leader shuts down all the pendingRequests in its LeaderState are rejected with NotLeaderException. This should not happen for requests which have already been committed to the log. These requests will be applied to the state machine and along with that the same request would be retried by the client because of NotLeaderException. The same request would be applied to the state machine twice which can lead to an inconsistent state.

This scenario would also occur in case the leader crashes. In such a scenario the leader will not send NotLeaderException but the client request would timeout leading to the same results.



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