[
https://issues.apache.org/jira/browse/MAPREDUCE-2917?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Mahadev konar updated MAPREDUCE-2917:
-------------------------------------
Status: Open (was: Patch Available)
@arun,
I think this heuristic might not work well. If you already have 20
reservations, it will take you 4000s to reserve another container.
What we might want to try out is:
{code}
number of re reservations/number of reservations * (1 - (size of container
needed/size of node))
{code}
This will penalize really huge reservations and will also provide a decent time
frame on re reserving the next container.
Also, we might want to bias towards nodes not already running high RAM
containers.
> Corner case in container reservations
> -------------------------------------
>
> Key: MAPREDUCE-2917
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-2917
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Components: mrv2, resourcemanager
> Affects Versions: 0.23.0
> Reporter: Arun C Murthy
> Assignee: Arun C Murthy
> Fix For: 0.23.0
>
> Attachments: MAPREDUCE-2917.patch
>
>
> Saw a corner case in container reservations where the node on which the AM is
> running was reserved, and hence never fulfilled leaving the application
> hanging.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
