[ 
https://issues.apache.org/jira/browse/FLINK-10868?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16759111#comment-16759111
 ] 

Zhenqiu Huang commented on FLINK-10868:
---------------------------------------

[~till.rohrmann]

Thanks for reviewing the PR. According to your suggestions. I changed as
 # As the feature is generic for both Yarn and Mesos, add only the maximum 
failure rate config option in resource manager config.
 # Put failure rate related logic into the TimestampBasedFailureRater which 
implements the FailureRater interface. As I don't want to mix two changes with 
different purpose in the same PR, we can make other code 
FailureRateRestartStrategy use it in another small PR.
 # For the failure rate test for RM, I tried to do in ResourceManagerTest. I 
found it is hard to mimic the behavior of registerSlotRequest without mocking 
lots components. And I also have to setup Test RM exactly like what 
YarnResourceManagerTest is doing. Thus, I still put the test cases separately 
in YarnResourceManagerTest and MesosResourceManagerTest.

> Flink's JobCluster ResourceManager doesn't use maximum-failed-containers as 
> limit of resource acquirement
> ---------------------------------------------------------------------------------------------------------
>
>                 Key: FLINK-10868
>                 URL: https://issues.apache.org/jira/browse/FLINK-10868
>             Project: Flink
>          Issue Type: Bug
>          Components: Mesos, YARN
>    Affects Versions: 1.6.2, 1.7.0
>            Reporter: Zhenqiu Huang
>            Assignee: Zhenqiu Huang
>            Priority: Major
>              Labels: pull-request-available
>          Time Spent: 10m
>  Remaining Estimate: 0h
>
> Currently, YarnResourceManager does use yarn.maximum-failed-containers as 
> limit of resource acquirement. In worse case, when new start containers 
> consistently fail, YarnResourceManager will goes into an infinite resource 
> acquirement process without failing the job. Together with the 
> https://issues.apache.org/jira/browse/FLINK-10848, It will quick occupy all 
> resources of yarn queue.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to