[
https://issues.apache.org/jira/browse/YARN-389?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13718673#comment-13718673
]
Hitesh Shah commented on YARN-389:
----------------------------------
The static limit is there for a reason. No application should ask for a
container above a certain limit as defined by the admins. For example, if most
nodes in a cluster have 4 GB resources, it can be used to set the cap to 4 GB
to ensure that even if all large nodes ( ones with more than 4 GB ) disappear,
the cluster is still healthy.
The issue at hand is a scheduling/allocation problem:
- can this allocation request be fulfilled?
- can it be fulfilled now?
- can it be fulfilled within a short window?
- can it ever be fulfilled?
- When an allocation request deemed to be non-fulfill-able?
- is this based on static configuration?
- is this based on a single snapshot of the dynamic view of the cluster?
- is this based on snapshots over a period of time?
- If time, on what basis is time defined?
- clock time?
- no. of rounds of heartbeats by all healthy nodes in the cluster.
How is the application informed of the necessary information that it needs to
make a decision? Information could be:
- your request could not be fulfilled
- your partial request could not be fulfilled
- the reason why it could not be fulfilled
- the current view of the cluster such as max available container size
> Infinitely assigning containers when the required resource exceeds the
> cluster's absolute capacity
> --------------------------------------------------------------------------------------------------
>
> Key: YARN-389
> URL: https://issues.apache.org/jira/browse/YARN-389
> Project: Hadoop YARN
> Issue Type: Bug
> Reporter: Zhijie Shen
> Assignee: Omkar Vinit Joshi
>
> I've run wordcount example on branch-2 and trunk. I've set
> yarn.nodemanager.resource.memory-mb to 1G and
> yarn.app.mapreduce.am.resource.mb to 1.5G. Therefore, resourcemanager is to
> assign a 2G AM container for AM. However, the nodemanager doesn't have enough
> memory to assign the container. The problem is that the assignment operation
> will be repeated infinitely, if the assignment cannot be accomplished. Logs
> follow.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
