[
https://issues.apache.org/jira/browse/HDFS-5835?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13883558#comment-13883558
]
Jing Zhao commented on HDFS-5835:
---------------------------------
+1 the Patch looks good to me.
Some questions not related to the patch:
# So when we start the SBN, the SBN already has been upgraded?
# Is it possible that the NN failover just happens when we start the SBN? Or
the other NN is in standby state at this time, and this NN will become active
in the end? In that case this "STARTED' option may also be applied to the ANN?
# If we allow SBN to do checkpoint during the rolling upgrade, the SBN may not
hit the upgrade marker in the editlog when it restarts. Thus the current
document said we would disable the checkpoint. But this may also cause issue if
the time between "start" and "finalize" is long. Since we do not delete old
editlog and fsimage during checkpointing, an alternative way is to scan the
editlog even across the fsimage?
> Add a new option for starting standby NN when rolling upgrade is in progress
> ----------------------------------------------------------------------------
>
> Key: HDFS-5835
> URL: https://issues.apache.org/jira/browse/HDFS-5835
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: namenode
> Reporter: Tsz Wo (Nicholas), SZE
> Assignee: Tsz Wo (Nicholas), SZE
> Attachments: h5835_20130127.patch
>
>
> When rolling upgrade is already in-progress and the standby NN is not yet
> started up, a new startup option is needed for the standby NN to initialize
> the upgrade status.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
