[
https://issues.apache.org/jira/browse/TAJO-1423?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14389666#comment-14389666
]
Jakob Homan commented on TAJO-1423:
-----------------------------------
Hey Dongjoon-
Great work. Tested the patch and was able to get it running with a few
modifications.
Patch review:
bq. Oh, tajo account in Dockerhub is taken by somebody else. Too bad
The account name should probably be ApacheTajo in any event.
* Need ASF Headers on all the files
* Rather than having a single person as the maintainer (Hyunsik), it might be
better if the dev list and the project were listed. Not sure if that's kosher
within Docker.
* The Hadoop config files are duplicated between centos and Ubuntu, and the
actual Dockerfiles are almost duplicated (about six lines different). Would be
better to have one set of files rather than two, and may be better to do the
needful to generate the platform-specific Dockerfiles. This could be done as
part of the build.sh script.
* Are the two flavors on supported OS (Centos vs Ubuntu) just to satisfy
organizations that might be based on one or the other? Docker should be able to
run either on either, so I'm not sure of their utility.
* run-cluster.sh should probably be have an env line to bash since it requires
on bash variable expansion (sh doesn't support this and cannot execute the
script). Also, see the next suggestion
* Rather than have a separate run-cluster.sh, would something like
[fig|http://www.fig.sh/] be more useful? Testing the patch I ran into a bunch
of issues with some containers not being up, rm'ed, etc. Might be useful to
move to cluster orchestration logic to a more developed tool. Also, this may
sidestep the slightly hacky curling of a commited file to bring in
run-cluster...
* Dockerfiles don't chmod +x on any of the init-nn, init-tajo, start-tajo
scripts. After adding these steps, was able to get the cluster up and running
as described in the Dockerfile.
> Support official docker images
> ------------------------------
>
> Key: TAJO-1423
> URL: https://issues.apache.org/jira/browse/TAJO-1423
> Project: Tajo
> Issue Type: Task
> Reporter: Dongjoon Hyun
> Assignee: Dongjoon Hyun
> Attachments: TAJO-1423.patch
>
>
> This issue will be discussed in Tajo mailing list. The following is my
> proposal for official Docker.
> First of all, the official Tajo docker will be here.
> https://registry.hub.docker.com/u/tajo/centos6/
> https://registry.hub.docker.com/u/tajo/ubuntu14.10/
> (Please refer https://registry.hub.docker.com/u/sktelecom/ubuntu14.10-hdw/ )
> To do that, we need to make a offical `tajo` account in dockerhub.com and set
> an automatic build by specifying Tajo Docker source folder location.
> Docker source folder might have a similar level with `dev-support` folder.
> The structure will look like the followings.
> {code}
> dockerfiles
> ├── centos6
> │ ├── Dockerfile
> │ ├── ...
> │ └── README.md
> └── ubuntu14.10
> ├── Dockerfile
> ├── ...
> └── README.md
> {code}
> After thatn, whenever Tajo Github is updated, Dockerhub will automatically
> rebuild the docker images.
> Note that Dockerfile usually use the binary distribution of Tajo, e.g. 0.10.0.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
