[
https://issues.apache.org/jira/browse/HDFS-14269?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16766651#comment-16766651
]
Konstantin Shvachko commented on HDFS-14269:
--------------------------------------------
Hey this is the standard startup procedure. If you restart Active NN it will
start as Standby, and you will call transitionToActive to make it active. With
ZK failover though this is automated. HDFS-14130 targets to exclude Observer
from voting.
If you want to bypass transitioning to Observer you can start NN with
{{StartupOption.OBSERVER}}. Hope this works for you.
> [SBN read] Observer node switches back to Standby after restart
> ---------------------------------------------------------------
>
> Key: HDFS-14269
> URL: https://issues.apache.org/jira/browse/HDFS-14269
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: namenode
> Affects Versions: 3.3.0
> Reporter: Wei-Chiu Chuang
> Priority: Major
>
> The Observer state is not persistent. Once it restarts it becomes Standby
> node again. Since it does not participate in NameNode failover for now,
> should we assume an Observer node is always an Observer node? This state
> should be persisted somewhere, like in Zookeeper.
> CC: [~csun] [~shv]. Like to get inputs from you. Have you discussed this
> before?
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
