[
https://issues.apache.org/jira/browse/HDFS-14059?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16687212#comment-16687212
]
Chao Sun commented on HDFS-14059:
---------------------------------
Nice finding [~zero45]. It seems we do need a mechanism to let ZK know about
the state change, so it will not transition the observer to standby. One
straightforward way is to add a method in {{ZKFCProtocol}} and use that to set
the state in ZK, though I'm not sure whether this is the cleanest solution.
Regarding auto failover for observers, I'm thinking that with ZK be aware of
observer state, maybe we can forbid observer from participating the election
(e.g., asking it to cede from active as long as it remains in observer state).
This seems a simpler than dealing with an election process where active,
standby, observer are all involved. Just some initial thoughts and would love
to hear thoughts from [~shv], [~xkrogen] and [~vagarychen].
> Test reads from standby on a secure cluster with Configured failover
> --------------------------------------------------------------------
>
> Key: HDFS-14059
> URL: https://issues.apache.org/jira/browse/HDFS-14059
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: test
> Reporter: Konstantin Shvachko
> Assignee: Plamen Jeliazkov
> Priority: Major
>
> Run standard HDFS tests to verify reading from ObserverNode on a secure HA
> cluster with {{ConfiguredFailoverProxyProvider}}.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
