You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@storm.apache.org by "Ethan Li (Jira)" <ji...@apache.org> on 2020/04/16 13:58:00 UTC

[jira] [Resolved] (STORM-3588) RAS scheduler should not pre-empt and evict topologies due to generic resource

     [ https://issues.apache.org/jira/browse/STORM-3588?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Ethan Li resolved STORM-3588.
-----------------------------
    Fix Version/s: 2.2.0
       Resolution: Fixed

Thanks [~li530]. Merged this to master

> RAS scheduler should not pre-empt and evict topologies due to generic resource
> ------------------------------------------------------------------------------
>
>                 Key: STORM-3588
>                 URL: https://issues.apache.org/jira/browse/STORM-3588
>             Project: Apache Storm
>          Issue Type: Improvement
>            Reporter: Rui Li
>            Assignee: Rui Li
>            Priority: Major
>              Labels: pull-request-available
>             Fix For: 2.2.0
>
>          Time Spent: 13h
>  Remaining Estimate: 0h
>
> Currently, when we enabled generic resource support in our cluster, RAS scheduler will evict scheduled topologies if new inappropriate asks for crazy amount of generic resources.
> We have identified the currently getScore() function in DefaultSchedulingPriorityStrategy does not consider the generic resource factor. Ideally, if user topology is asking way too much generic resources, it should be assigned lower priority in scheduling and won't affect other running topologies.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)