[
https://issues.apache.org/jira/browse/EAGLE-946?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15925483#comment-15925483
]
ASF GitHub Bot commented on EAGLE-946:
--------------------------------------
Github user qingwen220 commented on the issue:
https://github.com/apache/eagle/pull/869
@haoch my idea is to split the timeline into different ranges, which is
slightly different from your understanding.
Give an example, `endpointConfig.requestTimeInHour=3h, requestNum=4`, the
possible ranges are
(-inf, currentTime - 3h), (currentTime-3h, current-2h), (currentTime-2h,
currentTime-1h), (currentTime-1h, +inf)
With this division, ideally we won't miss any applications.
As for the algorithm you mentioned, I have a question. For example,
(-inf, currentTime - 3h): 4 apps
(currentTime-3h, current-2h): 0
(currentTime-2h, currentTime-1h): 0
(currentTime-1h, +inf): 23 apps
If my understanding is right, we will miss the 4 apps in (-inf, currentTime
- 3h)
> Refactor MRRunningJobApp & HadoopQueueApp
> ------------------------------------------
>
> Key: EAGLE-946
> URL: https://issues.apache.org/jira/browse/EAGLE-946
> Project: Eagle
> Issue Type: Improvement
> Affects Versions: v0.5.0
> Reporter: Zhao, Qingwen
> Assignee: Zhao, Qingwen
>
> Requirements
> 1. REST apis to the remote cluster should be called only once.
> 2. For each request, the fetch running apps should be limits
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
