[
https://issues.apache.org/jira/browse/HDFS-12467?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16186603#comment-16186603
]
Chen Liang commented on HDFS-12467:
-----------------------------------
Thanks [~nandakumar131] for the quick update!
Sorry, again I just noticed another thing...this change is actually trickier
than it seemed...below is not necessarily an issue, the patch seems to be doing
the right thing. But since it changes the syntax, so I just want to clarify on
this.
In the current code, the following steps:
1. call forceExitChillMode() : this will set inManualChillMode to false
2. call clearChillModeFlag() : this will set inManualChillMode to empty
3. call isOutOfNodeChillMode() : since inManualChillMode is empty, it purely
depends on totalNodes.get() >= getMinimumChillModeNodes() or not, it *can
return either true or false here*
In the new code, taking same steps
1. call forceExitChillMode() : this will set both inManualChillMode and
inStartupChillMode to false
2. call clearChillModeFlag() : this will set inManualChillMode to false
3. call isOutOfNodeChillMode() : at this point, it is guaranteed that both
inStartupChillMode and inManualChillMode are false, so this will *only return
true here*
This basically answering the question that: should forceExitChillMode force
exit startup chill mode as well? OR alternatively, should clearChillModeFlag
clear startup chill mode? Seems force chill mode is not being used a lot, so
this syntax detail probably doesn't even matter much, just that we should
probably keep in mind about this. I want to leave this JIRA some more time just
in case someone has a different opinion. Otherwise I'll commit it.
Another thing for follow-up, is that after it exits startupChillMode, what if
later on there are node dead and {{totalNodes.get()}} goes below
{{getMinimumChillModeNodes()}} again, should it go back to chill mode again? We
can leave this for a separate JIRA.
> Ozone: SCM: NodeManager should log when it comes out of chill mode
> ------------------------------------------------------------------
>
> Key: HDFS-12467
> URL: https://issues.apache.org/jira/browse/HDFS-12467
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: ozone
> Reporter: Nandakumar
> Assignee: Nandakumar
> Priority: Minor
> Attachments: HDFS-12467-HDFS-7240.000.patch,
> HDFS-12467-HDFS-7240.001.patch, HDFS-12467-HDFS-7240.002.patch,
> HDFS-12467-HDFS-7240.003.patch, HDFS-12467-HDFS-7240.004.patch
>
>
> {{NodeManager}} should add a log message when it comes out of chill mode.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
