Mailing-List: contact yarn-issues-help@hadoop.apache.org; run by ezmlm
Precedence: bulk
Date: Tue, 24 Oct 2017 19:06:00 +0000 (UTC)
From: "Wangda Tan (JIRA)" <jira@apache.org>
To: yarn-issues@hadoop.apache.org
Message-ID: <JIRA.13110811.1508448629000.48830.1508871960349@Atlassian.JIRA>
In-Reply-To: <JIRA.13110811.1508448629000@Atlassian.JIRA>
References: <JIRA.13110811.1508448629000@Atlassian.JIRA> <JIRA.13110811.1508448629077@jira-lw-us.apache.org>
Subject: [jira] [Commented] (YARN-7370) Preemption properties should be
 refreshable
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
archived-at: Tue, 24 Oct 2017 19:06:05 -0000


    [ https://issues.apache.org/jira/browse/YARN-7370?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16217478#comment-16217478 ] 

Wangda Tan commented on YARN-7370:
----------------------------------

Offline discussing this JIRA with [~GergelyNovak], I felt it might be better to do this discussion on community.

YARN-6142, we will take care of all scheduling edit policy refresh. So for this JIRA, I would like to keep it as simple as possible: 

In current implementation. PCPP#init is called only when RMActiveService start. What we can do here is: 

1) Move all config-related updates such as: 

    maxIgnoredOverCapacity = csConfig.getDouble(
        CapacitySchedulerConfiguration.PREEMPTION_MAX_IGNORED_OVER_CAPACITY,
        CapacitySchedulerConfiguration.DEFAULT_PREEMPTION_MAX_IGNORED_OVER_CAPACITY);

    naturalTerminationFactor = csConfig.getDouble(
        CapacitySchedulerConfiguration.PREEMPTION_NATURAL_TERMINATION_FACTOR,
        CapacitySchedulerConfiguration.DEFAULT_PREEMPTION_NATURAL_TERMINATION_FACTOR);
  
   To a separate method: updateConfigs(CapacitySchedulerConfiguration conf). 

2) Call updateConfigs at the beginning of editSchedule. 

3) Add a local variable to store CapacitySchedulerConfiguration such as lastCSConf, and inside updateConfigs, compare if scheduler.getConfiguration == lastCSConf. Only update configs when scheduler.getConfiguration != lastCSConf. With this change, we can avoid config version, etc. 

4) And we should not change any logics of RMActiveService, etc. It should be sufficient to change PCPP only.

> Preemption properties should be refreshable
> -------------------------------------------
>
>                 Key: YARN-7370
>                 URL: https://issues.apache.org/jira/browse/YARN-7370
>             Project: Hadoop YARN
>          Issue Type: Bug
>          Components: capacity scheduler, scheduler preemption
>    Affects Versions: 2.8.0, 3.0.0-alpha3
>            Reporter: Eric Payne
>
> At least the properties for {{max-allowable-limit}} and {{minimum-threshold}} should be refreshable. It would also be nice to make {{intra-queue-preemption.enabled}} and {{preemption-order-policy}} refreshable.


--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: yarn-issues-help@hadoop.apache.org