[ https://issues.apache.org/jira/browse/YARN-4108?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14731183#comment-14731183 ]
Wangda Tan commented on YARN-4108: ---------------------------------- [~sunilg], bq. Or is there any other advantage of making to-be-preempted containers from allocation logic such as user-limit The biggest benefit is, we will automatically check: which containers to preempt if we want to ask some other containers. This is quite important to me: with this behavior, all preempted containers are confirmed could be leveraged by less-satisfied applications. If we don't do this in scheduler allocation logic, we may end up with copying some of these logics to preemption policy OR do something like "dry-run" from outside, which is not as straightforward as doing that directly in allocation logic to me. > CapacityScheduler: Improve preemption to preempt only those containers that > would satisfy the incoming request > -------------------------------------------------------------------------------------------------------------- > > Key: YARN-4108 > URL: https://issues.apache.org/jira/browse/YARN-4108 > Project: Hadoop YARN > Issue Type: Bug > Components: capacity scheduler > Reporter: Wangda Tan > Assignee: Wangda Tan > > This is sibling JIRA for YARN-2154. We should make sure container preemption > is more effective. > *Requirements:*: > 1) Can handle case of user-limit preemption > 2) Can handle case of resource placement requirements, such as: hard-locality > (I only want to use rack-1) / node-constraints (YARN-3409) / black-list (I > don't want to use rack1 and host\[1-3\]) > 3) Can handle preemption within a queue: cross user preemption (YARN-2113), > cross applicaiton preemption (such as priority-based (YARN-1963) / > fairness-based (YARN-3319)). -- This message was sent by Atlassian JIRA (v6.3.4#6332)