Drop unused ports and switch run_init to false

[ehelms]

From my testing and talking to Candlepin team:

• Calling init is really a development task and not needed for production
• AJP connector seems to be only useful if using Apache and even then not if you are using mod_proxy
• The init can be called via HTTPS so 8080 is not needed at all

This reduces Candlepin setup time, and reduces the number of ports that Candlepin takes up.

[ehelms]

@kahowell Would you mind verifying my assumptions here are correct?

I tested all these changes on Katello nightly and saw no degradation of testing.

[ehelms]

I don't know that I understand these tests well enough to know why this fails to detect 8443 as the available port.

[ekohl]

We're using a testing pattern that's not really the standard puppet way. Since we decided that this isn't for 1.20 I'll revisit this when RC1 is out.

[kahowell]

From my point of view, this PR is 👍

[ekohl]

You'll want to rebase on #116. If you could also hide this behind if $candlepin::run_init that'd make the default installation slightly smaller:

puppet-candlepin/manifests/install.pp

Line 7 in 4c9a8ab

ensure_packages(['wget'], { ensure => $candlepin::wget_version, })

[ehelms]

Still a bit stuck after rebase on how to get this updated test to pass

[ekohl]

From the logs:

Nov 07 17:31:03 centos7-64.example.com server[1730]: SEVERE: Failed to load keystore type PKCS12 with path conf/keystore due to /usr/share/tomcat/conf/keystore (No such file or directory)
Nov 07 17:31:03 centos7-64.example.com server[1730]: java.io.FileNotFoundException: /usr/share/tomcat/conf/keystore (No such file or directory)

I suspect the issue is that we never set up a keystore with a certificate in our tests so we can't listen on 8443. Previously we listened on 8080 and it was no problem.

[ekohl]

@ehelms I've updated the example to generate a keystore as well

[ekohl]

I've opened #119 to remove the lint check

[ekohl]

Should be good to go when the tests are green.

[ehelms]

Thanks for all the help @ekohl ! I am thinking I either hold on merging this until Katello 3.10 is branched, or merge and don't update this module in 3.10

[ekohl]

I'm fine with holding off on merging. Otherwise 3.10 will be a downgrade compared to nightly and I think that's uglier than the alternatives.

[ekohl]

Before I forget: this will need an installer migration for upgrading users.

[ekohl]

Before I forget: this will need an installer migration for upgrading users.

On second thought, it would if it was a top level module but it isn't.