[
https://issues.apache.org/jira/browse/FLINK-16006?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17038672#comment-17038672
]
Zhenqiu Huang commented on FLINK-16006:
---------------------------------------
[~rmetzger] Thanks for the feedback. blocklist should be common logic in
FlinkResourceManager, so that every type of cluster manager frameworks can
inherit from it. For now, I know yarn and mesos has the blacklist functionality
support.
In yarn, If the number of task-failures on a Node Manager of a given job/Yarn
application reaches 3(mapreduce.job.maxtaskfailures.per.tracker), Application
Master of that job will ask Resource Manager to blacklist that Node Manager for
that job only. It means, new tasks of that job will not be assigned to that
blacklisted Node Manager.
For Kubernetes, I need more investigation. [~fly_in_gis] Do you have any
experience of blacklist support in Kubernetes?
> Add host blacklist support for Flink YarnResourceManager
> --------------------------------------------------------
>
> Key: FLINK-16006
> URL: https://issues.apache.org/jira/browse/FLINK-16006
> Project: Flink
> Issue Type: Improvement
> Components: Deployment / YARN
> Affects Versions: 1.10.0
> Reporter: Zhenqiu Huang
> Priority: Minor
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
