puppetlabs-puppetserver_perf_control

An r10k control repo for Puppet Server (and PE) perf testing and memory analysis

What is this repo?

This repo is an r10k control repo that we will be using to generate catalogs for automated gatling performance testing, as well as for other memory analysis and performance testing work.

~branching strategy

The production branch of this repo should be thought of as the mainline/HEAD for development, and it should be assumed that it is subject to change at any time.

Therefore, for any kind of perf or memory test that we expect to be repeatable, we should create a branch off of production at a specific point in time, and treat these branches as immutable going forward. Recommended branch naming convention is:

<YYYYMMDD>_<description>

So, for example,

20160614_memtest

At some point we may need to re-evaluate this strategy if we start piling up a huge number of these branches, but for now this seems like the best plan to ensure that we have a fixed reference point for reliable, repeatable test outcomes.

How to use this repo during development

This workflow assumes that you will be running puppetserver from source on your local machine. Further, it assumes that after cloning this control repo, on the branch you're working on, you've created an environment-specific manifests/site.pp file, which contains a node 'default' block that brings in the role that you are interested in. With this pattern, any node that is using the specified environment is going to get that role as its classification, but presumably the only nodes that will be using the environment are nodes where this behavior is desirable. We may want to look into a more sophisticated way to handle classification in the future.

Steps:

• Create a local r10k.yaml config file that looks something like this:

:cachedir: '/my/scratch/dir/perf-control-repo/cache'

:sources:
  :puppetserver-perf:
    remote: 'git@github.com:puppetlabs/puppetlabs-puppetserver_perf_control'
    basedir: '/my/puppetserver/src/dir/target/master-code-dir/environments'

• Install r10k as a gem
• Make any desired changes to your local copy of the control repo, commit them, and push them up to the hubz.
• To deploy the r10k environment to your local Puppet Server, run something like r10k deploy environment 20160614_memtest -p -v debug -c /my/scratch/dir/perf-control-repo/r10k.yaml. (You can leave out the environment name if you want to deploy all environments.)
• Copy the hiera.yaml file from ./root_files into your confdir (presumably target/master-conf-dir), and set the datadir to a value appropriate for your setup (e.g. /path/to/puppet/server/sourcecode/target/master-code-dir/environments/%{environment}/hieradata. (NOTE: there is a slightly confusing comment in the hiera.yaml file, which is copy/pasted from the actual default hiera.yaml that we ship. It tells you that you might not need to set your datadir. For dev, you almost always will need to set it, so just ignore the comment.)
• Spin up a vmpooler VM to use as a test agent; install puppet-agent, configure the server to point to your host, and then run puppet agent -t --environment 20160614_memtest
• Lather, rinse, repeat