[ 
https://issues.apache.org/jira/browse/YARN-4635?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15132322#comment-15132322
 ] 

Sunil G commented on YARN-4635:
-------------------------------

Hi [~jian he]
bq. I said this because I feel the API may look simpler and we don't need a 
separate nested AMBlackListingRequest class
For this feature, if we have a class for {{AMBlackListingRequest}}, I think it 
will be more clear to end user. Few minor advantages.

If enable/disable flag is not present in AMBlackListingRequest, then as you 
mentioned we ll be using threshold alone. Now it has to convey 4 cases.
- If user is not configuring this information from AM end. (a default value is 
needed, -ve value may be).
- User wants to disable Blacklisting for this specific application. (threshold 
need to 0)
- User wants to use this blacklisting and will configure a value for this.
- User wants this feature, but do not know a good threshold. Wants to use 
global threshold. So he need to give some big values more than 1.0f to get this 
behavior.
If we have a flag, we can simply turn on or off this feature per app. And if 
user doesnt wants  this feature, user need not have to set this blacklist 
object in context (null will set). 

These are not very strong reasons, and as you told we can achieve the current 
behavior in  both ways. No problem in choosing either of these options.  :)

> Add global blacklist tracking for AM container failure.
> -------------------------------------------------------
>
>                 Key: YARN-4635
>                 URL: https://issues.apache.org/jira/browse/YARN-4635
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: resourcemanager
>            Reporter: Junping Du
>            Assignee: Junping Du
>            Priority: Critical
>         Attachments: YARN-4635-v2.patch, YARN-4635.patch
>
>
> We need a global blacklist in addition to each app’s blacklist to track AM 
> container failures in global 
> affection. That means we need to differentiate the non­-succeed 
> ContainerExitStatus reasoning from 
> NM or more related to App. 
> For more details, please refer the document in YARN-4576.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to