[
https://issues.apache.org/jira/browse/BIGTOP-1497?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14196429#comment-14196429
]
Sean Mackrory commented on BIGTOP-1497:
---------------------------------------
Looks good so far. A few things that jump out are
1) The install section where directories are created and populated with files
from the build should be abstracted into an install_tachyon.sh script - like
do-component-build we implement that for each component so as to minimize the
stuff in platform-specific files like .spec files.
2) The .spec file has a reference directly to Amp Lab's Github for the tarball.
Information about where to obtain tarball releases should be in the bigtop.mk
file, and package.mk or the Gradle will download the tarball so that neither
the .spec file or the Debian packaging has to know specific about where it's
coming from.
3) This might be a case where Bigtop has deviated from what the Linux
distributions do, but most of the bits like JARs, shell scripts from upstream,
etc. go under /usr/lib/<component>/, and /usr/bin contains wrappers that set up
the environment, detect Java, etc. and then invoke scripts
in/usr/lib/<component>/bin.
Are there init scripts, etc. for the services that make up this system? We have
a template (look for uses of init.d.tmpl) for generating init scripts that are
aware of the little quicks on various distros.
> Add tachyon to bigtop
> ----------------------
>
> Key: BIGTOP-1497
> URL: https://issues.apache.org/jira/browse/BIGTOP-1497
> Project: Bigtop
> Issue Type: Bug
> Components: build
> Reporter: jay vyas
> Fix For: 1.0.0
>
>
> Lets package apache tachyon into bigtop. Its synergistic for both mapreduce,
> yarn, and spark , which are the dominant processing engines in big data.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
