[
https://issues.apache.org/jira/browse/YARN-6793?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Yufei Gu updated YARN-6793:
---------------------------
Description:
There is a delay between preemption happen and containers are killed. If some
resources released from nodes which are supposed to be preempted at that time
are not enough for the resource request, reservation happens again at that node.
E.g. scheduler reserves <memory 2048, vcore 2> in node 1 for app 1. It will
take 15s by default to kill containers in node 1 for fulfill that resource
requests. If <memory 1024, vcore 1> released from node 1 before the killing,
scheduler reserve <memory 2048, vcore 2> for app1 again in node 1.
was:
There is a delay between preemption happen and containers are killed. If some
resources released from nodes which are supposed to be preempted at that time
are not enough for the resource request, reservation happens again at that node.
E.g. scheduler reserve <memory 2048, vcore 2> in node 1 for app 1. It will take
15s by default to kill containers in node 1 for fulfill that resource requests.
If <memory 1024, vcore 1> released from node 1 before the killing, scheduler
reserve <memory 2048, vcore 2> for app1 again in node 1.
> Duplicated reservation in Fair Scheduler preemption
> ----------------------------------------------------
>
> Key: YARN-6793
> URL: https://issues.apache.org/jira/browse/YARN-6793
> Project: Hadoop YARN
> Issue Type: Bug
> Components: fairscheduler
> Affects Versions: 2.8.1, 3.0.0-alpha3
> Reporter: Yufei Gu
> Assignee: Yufei Gu
> Priority: Critical
>
> There is a delay between preemption happen and containers are killed. If some
> resources released from nodes which are supposed to be preempted at that time
> are not enough for the resource request, reservation happens again at that
> node.
> E.g. scheduler reserves <memory 2048, vcore 2> in node 1 for app 1. It will
> take 15s by default to kill containers in node 1 for fulfill that resource
> requests. If <memory 1024, vcore 1> released from node 1 before the killing,
> scheduler reserve <memory 2048, vcore 2> for app1 again in node 1.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
