[jira] [Comment Edited] (BIGTOP-1908) Move bigtop-deploy's dockerfiles into a centralized docker place

["Evans Ye (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/BIGTOP-1908?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14734406#comment-14734406 ] 

Evans Ye edited comment on BIGTOP-1908 at 9/8/15 8:21 AM:
----------------------------------------------------------

* Oops, should remove the image section in Vagrantfile, sorry.
* That's an option. What you need in BIGTOP-1936 is the apt module. If we do it in this way, we need to separate out the rng-tool, enable local repo, and mkdir code snippets, otherwise our bigtop/puppet images will be tainted. 
* Follow the logic above, I think its the time to refactor and isolate out puppet and puppet module installation code from setup-env*.sh. We can put that code in *bigtop_toolchain/bin/puppetize-{RHEL, DEB}.sh* so that we can reuse it anywhere. After all bigtop_toolchain needs someone to setup puppet for it, hence providing puppetize.sh as its init script is natural. By having puppetize.sh we can better support to setup bare-metal, VM build slaves as well.
* In bigtop/puppet's Dockerfiles, execute *puppetize-{RHEL, DEB}.sh* and ./gradlew toolchain-puppetmodules to make iimages puppetized.
* For Docker provisioner, since we already separated out puppet code from setup-env*.sh, there should be no code change.
* For VM provisioner, executing puppetize.sh script must be added in its Vagrantfile.

Any suggestion to make it better?


was (Author: evans_ye):
* Oops, should remove the image section in Vagrantfile, sorry.
* That's an option. What you need in BIGTOP-1936 is the apt module. If we do it in this way, we need to separate out the rng-tool, enable local repo, and mkdir code snippets, otherwise our bigtop/puppet images will be tainted. 
* Follow the logic above, I think its the time to refactor and isolate out puppet and puppet module installation code from setup-env*.sh. We can put that code in *bigtop_toolchain/bin/puppetize-{RHEL, DEB}.sh* so that we can reuse it anywhere. After all bigtop_toolchain needs someone to setup puppet for it, hence providing puppetize.sh as its init script is natural. By having puppetize.sh we can better support to setup bare-metal, VM build slaves as well.
* In bigtop/puppet's Dockerfiles, execute puppetize-{RHEL, DEB}.sh and ./gradlew toolchain-puppetmodules to make iimages puppetized.
* For Docker provisioner, since we already separated out puppet code from setup-env*.sh, there should be no code change.
* For VM provisioner, executing puppetize.sh script must be added in its Vagrantfile.

Any suggestion to make it better?

> Move bigtop-deploy's dockerfiles into a centralized docker place
> ----------------------------------------------------------------
>
>                 Key: BIGTOP-1908
>                 URL: https://issues.apache.org/jira/browse/BIGTOP-1908
>             Project: Bigtop
>          Issue Type: Improvement
>    Affects Versions: 1.0.0
>            Reporter: Evans Ye
>            Assignee: Evans Ye
>             Fix For: 1.1.0
>
>         Attachments: BIGTOP-1908.patch, BIGTOP-1908.patch, BIGTOP-1908.patch
>
>
> It would be great if we can have all the docker image building scripts consolidated into a centralized docker place. The docker image for bigtop docker provisioner can be built and pushed to dockerhub as well so that the docker provisioner can be easily consumed by bigtop users without requiring users to build image by themselves.



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