jenkinsci/configuration-as-code-plugin

Jenkins Configuration-as-Code plugin

Joseph Petersen

@casz

@030 i was referring to the version of git plugin :)

Oleg Nenashev

@oleg-nenashev

Got the recording published for the today;s session: https://youtu.be/WBrT6xYbhm0

Jake Burns

@burnsjake

Is there a good howto on how to bootstrap a brand new jenkins installation (without docker) with the configuration as code plugin?

Tim Jacomb

@timja

https://github.com/jenkinsci/configuration-as-code-plugin#getting-started

@burnsjake just follow getting started, install the configuration-as-code plugin.
Export your desired configuration to a yaml file, then store it

Mark Waite

@MarkEWaite

@burnsjake that information is also available from https://plugins.jenkins.io/configuration-as-code, though not yet as convenient to hyperlink as the github.com page.

Jake Burns

@burnsjake

@MarkEWaite @timja , I'm confused as to how I'd install the CasC plugin without manually going through the first-use. I'm looking to do this without docker.

Mark Waite

@MarkEWaite

@burnsjake There probably is a way to initiate a Jenkins installation that also installs the configuration as code plugin, but most efforts to automate at that level of installation are done with Docker. If you're using one of the operating system packages (rpm, deb, msi), then after the installation and selecting the recommended plugins, you would use the plugin manager page to install the configuration as code plugin.

Jake Burns

@burnsjake

Yup, using aptitude to install on ubuntu. But there are no apt packages to install separate plugins. I'm fairly surprised this hasn't been sussed out yet.

Mark Waite

@MarkEWaite

Already complicated enough to maintain 3 different installer packages for the main package (deb, rpm, msi) with no solid consensus for techniques to manage add-ons between the various Linux distros and Windows. Should we style after python pip or ruby gems or golang or ...

Jake Burns

@burnsjake

My thought is to install via aptitude, stop jenkins immediately, add JAVA_ARGS="-Djava.awt.headless=true -Djenkins.install.runSetupWizard=false" to /etc/default/jenkins, add an admin user with a groovy script into /usr/share/jenkins/ref/init.groovy.d/security (optional?), Install a plugin manually (???), PROFIT!

Jake Burns

@burnsjake

A plugin being the CasC plugin.

Joseph Petersen

@casz

@burnsjake why won't you use docker? :confused:

Tim Jacomb

@timja

@burnsjake there's
https://docs.ansible.com/ansible/latest/modules/jenkins_plugin_module.html
and
https://github.com/jenkinsci/plugin-installation-manager-tool

for installing plugins

Tim Jacomb

@timja

@oleg-nenashev did you want to do a jcasc release soon?
Someone just asked on github:
https://github.com/jenkinsci/configuration-as-code-plugin/pull/1053#issuecomment-535865314

Daniel Estermann

@d.esterman_gitlab

I also wait for a release to test if ConfiguratorException: serverAuthenticationToken is required to configure class hudson.plugins.sonar.SonarInstallation has been fixed

Robert Sandell

@rsandell

Hello, I'm working on fixing JCasC support for the git-client tool installations, and it's a bit of an unusual beast to fix.

Mark Waite

@MarkEWaite

That is the understatement of the year. @rsandell .

Robert Sandell

@rsandell

My second attemt I thought was a clever and beautiful hack where I used readResolve to "fix" the things that jcasc was doing wrong when using the databound constructor logic

But alas it turns out that jcasc is not calling readResolve after creating the objects.

Mark Waite

@MarkEWaite

Ah, that hurts.

I had an alternate hack technique that came to mind. Would love to hear your thoughts on my hack idea.

Robert Sandell

@rsandell

It's not clear cut to me now that I'm thinking about it if it should or not. XStream does it after deserializing from disk. but in form submission it doesn't. And perhaps JCasC is a bit of both or somewhere in between.

Mark Waite

@MarkEWaite

There are only two special strings that identify the JGit implementations, "jgit" and "jgitapache". What if the git client plugin data bound constructor for CliGit detected one of those two strings and called out to the JGit implementation instead?

I assume (my guessing) that making JCasC call readResolve is a much more invasive change that the team will be ready to make. It will invoke many pieces of code that had not been run before in this context

Robert Sandell

@rsandell

Well you can't really construct another object from the constructor

my readresolve does that; detecting if name is the magic name and in that case returns the appropriately constructed object

Mark Waite

@MarkEWaite

Could the CliGit data bound constructor somehow "queue" a call to readResolve if it detects it is called in from JCasC?

Robert Sandell

@rsandell

My second attempt that I'm currently working on is using a custom Configurator, that would probably make it possible to have a bit more nice looking yaml, but it would mean adding an optional dependency to jcasc in the git-client plugin. And the Configurator API in jscasc is badly documented to say it mildly :) so the work is going slow on that approach.

Would you be against adding that optional dependency mark?

Mark Waite

@MarkEWaite

I don't object to adding an optional JCasC dependency into the git client plugin. A mandatory dependency would not be acceptable, but optional would be fine.

Robert Sandell

@rsandell

ok.

How do people feel about adding a call to readresolve in the handling of DataboundConstructorConfigurator? or maybe it was called DescribableConfigurator (the default one)?

I've heard a rumor that Oleg doesn't like the custom configurator approach :)

Mark Waite

@MarkEWaite

How do people feel about adding a call to readresolve in the handling of DataboundConstructorConfigurator? or maybe it was called DescribableConfigurator (the default one)?

@casz and @timja seem like the two best suited to answer that question along with @oleg-nenashev

Robert Sandell

@rsandell

Oleg is on vacation so I didn't dare pinging him :) But other's opinions are welcome ;)

oleg-nenashev @oleg-nenashev waves

Mark Waite

@MarkEWaite

Sorry for my mistake pinging him on vacation!

Mark Waite

@MarkEWaite

If that didn't work, could we modify the CliGit data bound constructor to have it detect the name argument is "jgit" (or "jgitapache") and then act as though it were JGit internally? I assume that would be a more dramatic change to the git client plugin, but would avoid inserting a readResolve call into a highly used portion of JCasC

Robert Sandell

@rsandell

possible, but quite ugly Ithink. The UI config would change as well possibly confusing users further

Mark Waite

@MarkEWaite

Good point

Oleg Nenashev

@oleg-nenashev

Adding support of readResolve() to JCasC might be doable

Robert Sandell

@rsandell

I don't think adding a readResolve call would harm anything as most of those implementations looks for things deserialized from an old format and if so converts to newer fields already set by the databound constructor. But there is always corner cases I guess :)

is the rumor true @oleg-nenashev that you preffer not to use custom Configurators if avoidable?

it would make the yaml a bit better in this case, but much more work since it's not very documented so I'm fumbling a bit in the dark trying to user it :)

Oleg Nenashev

@oleg-nenashev

@rsandell for me readResolve() is rather a part of a wider topic about retaining JCasC backward compatibility for configurations. Right now we have no answer of it, and it would be great to have one

I will not push back any Configurator implementation FTR. If a plugin maintainer is fine AND if it is really need to resolve an issue, why not?

Robert Sandell

@rsandell

OK :)

Tim Jacomb

@timja

My second attempt that I'm currently working on is using a custom Configurator, that would probably make it possible to have a bit more nice looking yaml, but it would mean adding an optional dependency to jcasc in the git-client plugin. And the Configurator API in jscasc is badly documented to say it mildly :) so the work is going slow on that approach.

Far too nice, the API really needs some docs improvements :(

Where communities thrive

People

Repo info

Activity