[
https://issues.apache.org/jira/browse/HDFS-15830?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17281826#comment-17281826
]
Stephen O'Donnell commented on HDFS-15830:
------------------------------------------
[~ferhui] Thanks for alerting me. Just so I understand, is the use case:
1. The NN is started with dfs.image.parallel.load = false
2. You want to enable the feature without restarting the NN, so make it
reconfigurable, the further checkpoints will have parallel images.
{quote}
If something unexpected happens, we have to load old image to restart namenode.
{quote}
What "unexpected" things can happen here? I would like to understand if you are
seeing any problems with the parallel loading feature.
Making the option reconfigurable does seem like a good improvement - I will try
to review the change in the next day or two.
> Support to make dfs.image.parallel.load reconfigurable
> ------------------------------------------------------
>
> Key: HDFS-15830
> URL: https://issues.apache.org/jira/browse/HDFS-15830
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: namenode
> Reporter: Hui Fei
> Assignee: Hui Fei
> Priority: Major
> Labels: pull-request-available
> Time Spent: 20m
> Remaining Estimate: 0h
>
> After HDFS-14617, loading fsimage improve a lot.
> If something unexpected happens, we have to load old image to restart
> namenode.
> So advise that we make dfs.image.parallel.load reconfigurable, then we can
> save new fsimage.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
