[
https://issues.apache.org/jira/browse/RATIS-709?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16983738#comment-16983738
]
Tsz-wo Sze commented on RATIS-709:
----------------------------------
Took a second look of the patch. The new code will never be executed since (e
instanceof NotReplicatedException) is never true. NotReplicatedException is
returned as a part of the RaftClientReply but not completed exceptionally.
Therefore, the client will not retry since the RaftClientReply is not null. We
have WatchRequestTests.testWatchRequestTimeout asserting it.
Are you sure that the client will retry on NotReplicatedException?
> RaftClient should not retry on a different leader on NotReplicated exception
> from leader
> ----------------------------------------------------------------------------------------
>
> Key: RATIS-709
> URL: https://issues.apache.org/jira/browse/RATIS-709
> Project: Ratis
> Issue Type: Bug
> Components: client
> Reporter: Shashikant Banerjee
> Assignee: Sammi Chen
> Priority: Blocker
> Attachments: RATIS-709.00.patch
>
>
> Currently, when a watch request times out with a NotReplicatedException on
> the leader raft client starts retrying the request on different server and
> starts failing with NotLeaderException and it goes in a loop. Ideally , when
> a watch request times out , it should not be retried automatically by raft
> client given the timeout value in the leader is sufficiently reasonable.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
