[
https://issues.apache.org/jira/browse/MAPREDUCE-6353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14529450#comment-14529450
]
Anubhav Dhoot commented on MAPREDUCE-6353:
------------------------------------------
Its a good discussion to have whether we should allow zero vcores. I think its
ok if jobs do not care about getting any CPU allocated to them and just use
whatever leftover CPU there exists.
Irrespective I think we should not be crashing in the AM when calculating
available containers. If we need to enforce it I agree job submission is a
better way to rejecting those values.
> Divide by zero error in MR AM when calculating available containers
> -------------------------------------------------------------------
>
> Key: MAPREDUCE-6353
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-6353
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Components: mr-am
> Reporter: Anubhav Dhoot
> Assignee: Anubhav Dhoot
> Attachments: MAPREDUCE-6353.001.patch
>
>
> When running a sleep job with zero CPU vcores i see the following exception
> 2015-04-30 06:41:06,954 ERROR [RMCommunicator Allocator]
> org.apache.hadoop.mapreduce.v2.app.rm.RMContainerAllocator: ERROR IN
> CONTACTING RM.
> java.lang.ArithmeticException: / by zero
> at
> org.apache.hadoop.mapreduce.v2.app.rm.ResourceCalculatorUtils.computeAvailableContainers(ResourceCalculatorUtils.java:38)
> at
> org.apache.hadoop.mapreduce.v2.app.rm.RMContainerAllocator$ScheduledRequests.assign(RMContainerAllocator.java:947)
> at
> org.apache.hadoop.mapreduce.v2.app.rm.RMContainerAllocator$ScheduledRequests.access$200(RMContainerAllocator.java:840)
> at
> org.apache.hadoop.mapreduce.v2.app.rm.RMContainerAllocator.heartbeat(RMContainerAllocator.java:247)
> at
> org.apache.hadoop.mapreduce.v2.app.rm.RMCommunicator$1.run(RMCommunicator.java:282)
> at java.lang.Thread.run(Thread.java:745)
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
