[
https://issues.apache.org/jira/browse/BIGTOP-313?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14053236#comment-14053236
]
Roman Shaposhnik commented on BIGTOP-313:
-----------------------------------------
Well, if you look at the suggested way of executing puppet you'll see that it
expects configs to be in /etc/puppet/config. That makes our
bigtop-deploy/puppet/config just a place for sample configs.
A different way to look at it, of course, is --confdir and then it doesn't
matter where you place your configs, *except* for the fact that Puppet (at
least 2.x) expects your actual configuration to be under
$CONFDIR/config. IOW, when you say something like
--confdir=bigtop-deploy/puppet you still need that extra config folder at the
end.
> Move puppet configuration out of manifest and into extlookup() CSV
> ------------------------------------------------------------------
>
> Key: BIGTOP-313
> URL: https://issues.apache.org/jira/browse/BIGTOP-313
> Project: Bigtop
> Issue Type: Improvement
> Components: Deployment
> Reporter: Patrick Taylor Ramsey
> Assignee: Roman Shaposhnik
> Priority: Minor
> Labels: deploy, deployment, puppet
> Fix For: 0.3.0, 0.4.0
>
> Attachments: patch.txt
>
>
> Previously, bigtop-deploy puppet could only be configured by editing puppet
> manifests, or through a small number of environment variables that set facter
> facts. I've rewritten cluster.pp to use extlookup to read most configuration
> options from a site-specific CSV file with sane defaults for most options
> (only hadoop_head_node and bigtop_yumrepo_uri are required to be set). These
> changes should make it easier to automate deploying multiple clusters with
> different configurations from the same manifests.
> As written, the CSV files must live in $confdir/config/, where $confdir is
> the puppet config directory. This means that if you are manually calling
> 'puppet agent' on these manifests (rather than serving them from a puppet
> master), you are now required to specify --confdir in addition to
> --modulepath.
> At some point in the future, may be worth moving to hiera, as yaml is a lot
> prettier and more flexible than CSV. I did this using extlookup so as not to
> introduce external dependencies (since extlookup is now in puppet core).
--
This message was sent by Atlassian JIRA
(v6.2#6252)
