[jira] [Commented] (BIGTOP-1752) Add Zinc to bigtop_toolchain so that it may be used by Spark builds

["Olaf Flebbe (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/BIGTOP-1752?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15720602#comment-15720602 ] 

Olaf Flebbe commented on BIGTOP-1752:
-------------------------------------

Hey [~jonathak]

Right now our CI is sharing network interfaces (because of --net=container:nexus). 
Parallel builds with zinc will fail badly with this setup.

I would have to dig deep into docker networking to figure out an alternate solution with private docker networks, not sharing interfaces between containers.

If we agree to proceed with zinc enabled, this has to be solved.

Olaf

> Add Zinc to bigtop_toolchain so that it may be used by Spark builds
> -------------------------------------------------------------------
>
>                 Key: BIGTOP-1752
>                 URL: https://issues.apache.org/jira/browse/BIGTOP-1752
>             Project: Bigtop
>          Issue Type: New Feature
>          Components: spark, toolchain
>    Affects Versions: 0.8.0
>            Reporter: Jonathan Kelly
>            Assignee: Jonathan Kelly
>            Priority: Minor
>             Fix For: backlog
>
>         Attachments: BIGTOP-1752.patch, BIGTOP-1752.patch
>
>
> The Zinc Nailgun server can be used to speed up compilation of Scala projects (like Spark) considerably.  For example, building Spark without Zinc running takes about 26 mins in my experience, but with Zinc running it takes only about 14 mins.
> This patch causes Zinc to be installed in /usr/local/zinc by the bigtop_toolchain, though it must still be started by whatever process (e.g., Jenkins) executes the build by running "/usr/local/zinc/bin/zinc -start" before kicking off the build.
> See https://github.com/typesafehub/zinc and http://spark.apache.org/docs/latest/building-spark.html#speeding-up-compilation-with-zinc for more information about Zinc and using it with a Spark build.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)