Vinod Kumar Vavilapalli commented on YARN-3685:

bq. This was true of Linux even before YARN-316, so in that sense, YARN did 
already have some classpath logic indirectly.
bq. I was thinking of stuff like yarn.application.classpath, where values are 
defined in terms of things like the HADOOP_YARN_HOME and HADOOP_COMMON_HOME 
environment variables, and those values might not match the file system layout 
at the client side.
Hm.. YARN_APPLICATION_CLASSPATH is a simple convenience configuration property 
that the server *does not* load, but used by the applications like 
distributed-shell. And yeah, this convenience property was never assumed to 
work with variable installation layouts. Increasingly our apps are being 
migrated to a distributed-cache based deployment so as to avoid the layout 
issue, so in sum YARN_APPLICATION_CLASSPATH is essentially unused.

> NodeManager unnecessarily knows about classpath-jars due to Windows 
> limitations
> -------------------------------------------------------------------------------
>                 Key: YARN-3685
>                 URL: https://issues.apache.org/jira/browse/YARN-3685
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: nodemanager
>            Reporter: Vinod Kumar Vavilapalli
> Found this while looking at cleaning up ContainerExecutor via YARN-3648, 
> making it a sub-task.
> YARN *should not* know about classpaths. Our original design modeled around 
> this. But when we added windows suppport, due to classpath issues, we ended 
> up breaking this abstraction via YARN-316. We should clean this up.

This message was sent by Atlassian JIRA

Reply via email to