[ 
https://issues.apache.org/jira/browse/YARN-3212?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14368118#comment-14368118
 ] 

Ming Ma commented on YARN-3212:
-------------------------------

bq. Do we want to consider DECOMMISSIONING nodes as not active? There are 
containers actively running on them, and in that sense they are participating 
in the cluster (and contributing to the overall cluster resource). I think they 
should still be considered active, but I could be persuaded otherwise.

Do we need to support the scenario where NM becomes dead when it is being 
decommissioned? Say decommission timeout is 30 minutes larger than the NM 
liveness timeout.  The node drops out of the cluster for some time and rejoin 
later all within the decommission time out. Will Yarn show the status as just 
dead node, or {dead, decommissioning}? Seems useful for admins to know about 
it. If we need that,  we can consider two types of NodeState. One is liveness 
state, one is admin state. Then you will have different combinations.

> RMNode State Transition Update with DECOMMISSIONING state
> ---------------------------------------------------------
>
>                 Key: YARN-3212
>                 URL: https://issues.apache.org/jira/browse/YARN-3212
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: resourcemanager
>            Reporter: Junping Du
>            Assignee: Junping Du
>         Attachments: RMNodeImpl - new.png, YARN-3212-v1.patch, 
> YARN-3212-v2.patch
>
>
> As proposed in YARN-914, a new state of “DECOMMISSIONING” will be added and 
> can transition from “running” state triggered by a new event - 
> “decommissioning”. 
> This new state can be transit to state of “decommissioned” when 
> Resource_Update if no running apps on this NM or NM reconnect after restart. 
> Or it received DECOMMISSIONED event (after timeout from CLI).
> In addition, it can back to “running” if user decides to cancel previous 
> decommission by calling recommission on the same node. The reaction to other 
> events is similar to RUNNING state.



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

Reply via email to