You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "Paul Lin (Jira)" <ji...@apache.org> on 2020/11/24 04:11:00 UTC

[jira] [Commented] (FLINK-20185) Containerized environment variables may be misinterpreted in launch script

    [ https://issues.apache.org/jira/browse/FLINK-20185?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17237845#comment-17237845 ] 

Paul Lin commented on FLINK-20185:
----------------------------------

After some invesgation, I find this should be a problem on Hadoop side. Before MAPREDUCE-7069 Hadoop would set the conflicted env twice, appending the second value with a classpath separetor (which is `:` in Linux). 

> Containerized environment variables may be misinterpreted in launch script
> --------------------------------------------------------------------------
>
>                 Key: FLINK-20185
>                 URL: https://issues.apache.org/jira/browse/FLINK-20185
>             Project: Flink
>          Issue Type: Bug
>          Components: Runtime / Configuration
>    Affects Versions: 1.11.0
>         Environment: YARN 2.6.5
>            Reporter: Paul Lin
>            Priority: Major
>
> Containerized environment variables forwarding may be broken when conflicting with existing variables. For example, setting `MALLOC_ARENA_MAX` through `containerized.taskmanager.env.MALLOC_ARENA_MAX=1` results in  `export MALLOC_ARENA_MAX="1:1"` in the launch script.



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