[jira] [Commented] (YARN-1105) RM or YarnClient to notify AMs if resource requests cannot be satisfied

2015-03-17 Thread Steve Loughran (JIRA) 

[ 
https://issues.apache.org/jira/browse/YARN-1105?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14365080#comment-14365080
 ] 

Steve Loughran commented on YARN-1105:
--

This can actually be done in the client, which lets you do other policies such 
as decide when to relax placement, rather than rely on the cluster scheduler. 

 RM or YarnClient to notify AMs if resource requests cannot be satisfied
 ---

 Key: YARN-1105
 URL: https://issues.apache.org/jira/browse/YARN-1105
 Project: Hadoop YARN
  Issue Type: New Feature
  Components: client, nodemanager
Affects Versions: 2.1.0-beta
Reporter: Steve Loughran
Priority: Minor

 Long-lived services may want to know when resource requests can't be 
 satisfied, when they are stuck in a queue -so that they can cancel them, 
 change the requirements or react in another applications-specific way



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

[jira] [Commented] (YARN-1105) RM or YarnClient to notify AMs if resource requests cannot be satisfied

2013-08-27 Thread Steve Loughran (JIRA) 

[ 
https://issues.apache.org/jira/browse/YARN-1105?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13751669#comment-13751669
 ] 

Steve Loughran commented on YARN-1105:
--

This could be done in the AM by tracking when requests were made  acting when 
some haven't been satisfied after a given time period -which implies that the 
{{AMRMClientAsyncImpl}} could be extended to do this as it knows what is 
happening.

However, that will make it stateful, on an AM restart the list of outstanding 
requests would be lost, and we can't use {{AMRMClient. getMatchingRequests()}} 
as you need to know the {{Resource}} capability you asked for.


 RM or YarnClient to notify AMs if resource requests cannot be satisfied
 ---

 Key: YARN-1105
 URL: https://issues.apache.org/jira/browse/YARN-1105
 Project: Hadoop YARN
  Issue Type: New Feature
  Components: client, nodemanager
Affects Versions: 2.1.0-beta
Reporter: Steve Loughran
Priority: Minor

 Long-lived services may want to know when resource requests can't be 
 satisfied, when they are stuck in a queue -so that they can cancel them, 
 change the requirements or react in another applications-specific way

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira