[
https://issues.apache.org/jira/browse/HDFS-3026?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13218910#comment-13218910
]
Aaron T. Myers commented on HDFS-3026:
--------------------------------------
bq. Rather than waiting N seconds and running the NN in an ill-defined state
which will handle requests, seems like it would be better flag that the NN
shutdown on reply to the RPC, and the shutdown path attempt NN#stop vs exiting
hard. Would be useful for other non-HA cases where we discover in RPC context
that we'd like to shutdown. Agree?
The reason I didn't call NN#stop is because NN#stop itself tries to do a state
transition, which doesn't seem wise to do in the event of a failed state
transition.
I'm not sure what you mean by "Rather than waiting N seconds and running the NN
in an ill-definded state which will handle requests..." Do you mean that, in
addition to waiting N seconds before shutting down, we should also make sure
that no subsequent incoming RPCs will be handled by the NN?
> HA: Handle failure during HA state transition
> ---------------------------------------------
>
> Key: HDFS-3026
> URL: https://issues.apache.org/jira/browse/HDFS-3026
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: ha, name-node
> Affects Versions: HA branch (HDFS-1623)
> Reporter: Aaron T. Myers
> Assignee: Aaron T. Myers
> Attachments: HDFS-3026-HDFS-1623.patch
>
>
> This JIRA is to address a TODO in NameNode about handling the possibility of
> an incomplete HA state transition.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
