Puppet users - May 2011 - Continuous Integration, Puppet, dynamic VMs etc

Hi there,

Have any of you Puppet users/abusers out there worked on or even
succeeded at some sort of Continuous Integration of your
Puppet-language codebase? I''m very interested to hear about the
experiences of other people in the community on some of this stuff. If
you were in Amsterdam for Puppet Camp you might have seen my talk on
these kinds of things and I''m sure there are more of you who are crazy
enough to try similar ideas.

Last week I finally got around to coding up the rake tasks to support
dynamic VM creation with Vagrant to run our "Faces" compile tests on
modules that had already passed syntax checks and been packaged. There
were a few sticky points (and still are some). In particular:

* Information flow is a little bit awkward when you run jobs from
Jenkins on Vagrant VMs. It looks a bit like this:
Jenkins master -> Jenkins slave (in my case, the Virtualbox machine)
-> Vagrant VM.
 - since the slave checks out the repository, you need to copy it to
the vagrant shared folder root before running any of your build tasks
inside the VM
 - VM RPC is slightly awkward in general (basically all via SSH, as
explained in the next step)

* Vagrant libraries wrap around Net::SSH::Connection::Session.exec!
which means all output is saved up until the end of the command. Not
great when you want to watch the output in realtime in the Jenkins
console view
 - operating on the session object directly means you can use the exec
call, but need to handle output and exit codes yourself with
callbacks, which I''ve done but output is still being sent slightly out
of order... weird.

* I set out to have tests complete as fast as possible, so I
pre-create my VMs and use the next available one for the test runs.
Asynchronously they are destroyed and returned to the pool, but it''s
not pretty. I guess this is a common problem (startup time of new
instances) so I''m sure there are some good solutions.

* Passing information between Jenkins jobs is always slightly painful.
I was briefly tempted by Go for its pipelining but ultimately put off
by its insistence on XML for all config. Right now I''m looking to
store information about the module that just passed a certain stage of
testing in a file (perhaps YAML format) and archiving it as an
artifact, then copying that artifact in the next stage job... but
again this seems a bit clunky (and a pain to duplicate to the 20-odd
jobs for each stage I currently have).

I''m sure there were a few other things that bugged me but I''d
mostly
like to start up a discussion on any of this stuff with community
members who might have dabbled in this kind of stuff (or
alternatively, are utter mavens at it). Alternatively I''m happy to
share some of what I''ve been doing if it interests anyone who wants to
do more testing.

Best Regards,
Oliver

-- 
You received this message because you are subscribed to the Google Groups
"Puppet Users" group.
To post to this group, send email to puppet-users@googlegroups.com.
To unsubscribe from this group, send email to
puppet-users+unsubscribe@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/puppet-users?hl=en.