[
https://issues.apache.org/jira/browse/HDFS-10912?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15530843#comment-15530843
]
Jing Zhao commented on HDFS-10912:
----------------------------------
Thanks for the work, [~anu]. For SCM, I think we may have two different types
of "safemode":
# the first one is to make sure the SCM receives enough DN registration. If we
persist the container-node mapping in SCM, we do not need to wait for full
container reports. Also SCM does not take the responsibility for maintaining
the container states/durability, thus this type of safemode is very lightweight
compared with the current NN safemode. (maybe we can rename it ...)
# the second one is the manual safemode (triggered by {{forceEnterSafeMode}}).
This safemode is actually against the whole SCM instead of its node manager
(just like in today's HDFS the manual safemode is for the whole NN instead of
the blockmanager).
Therefore, to me
{{forceExitSafeMode}}/{{forceEnterSafeMode}}/{{isInManualSafeMode}} can be
moved to SCM level. {{forceExitSafeMode}} will reset both the manual safemode
and the safemode in nodemanager.
> Ozone:SCM: Add safe mode support to NodeManager.
> ------------------------------------------------
>
> Key: HDFS-10912
> URL: https://issues.apache.org/jira/browse/HDFS-10912
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: ozone
> Affects Versions: HDFS-7240
> Reporter: Anu Engineer
> Assignee: Anu Engineer
> Attachments: HDFS-10912-HDFS-7240.001.patch
>
>
> Add Safe mode support : That is add the ability to force exit or enter safe
> mode. As well as get the current safe mode status from node manager.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
