[
https://issues.apache.org/jira/browse/RATIS-573?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16868169#comment-16868169
]
Supratim Deka commented on RATIS-573:
-------------------------------------
The StateMachine will handle the log fail notification and (should) stop
accepting new transactions.
That was the plan, make the ContainerStateMachine inside Ozone stop accepting
any more transactions on the pipeline - make startTransaction return an error.
If we want to continue throw the exception, there has to be a graceful handling
without shutting down the node.
Until we define that, do you think the behaviour in the current patch is
reasonable?
> Handle Raft Log Append Failure
> ------------------------------
>
> Key: RATIS-573
> URL: https://issues.apache.org/jira/browse/RATIS-573
> Project: Ratis
> Issue Type: Improvement
> Components: server
> Reporter: Supratim Deka
> Assignee: Supratim Deka
> Priority: Major
> Attachments: RATIS-573.00.patch, RATIS-573.01.patch,
> RATIS-573.02.patch, RATIS-573.03.patch
>
>
> As part of Handling IO Failures, HDDS-1595.
> The scope of this jira is to handle failure in RAFT log append by:
> 1. notify the error to the state machine for consumer specific handling
> 2. propagate the error to the initiator (to the client from leader, to the
> leader from follower).
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
