bolt-lxd-lab

Explore configuring virtual machines and LXD containers with Puppet Bolt.

Requirements

You must have Bolt version 3.4.0 or higher. That is when LXD support was introduced. You also need Vagrant. Finally, you need the lxc command line client, so you can talk to LXD clusters.

You should also have some basic knowledge of LXD.

Getting Started

First, run vagrant up. Two Ubuntu virtual machines will launch. These will each be given 6 GB of RAM by default. Adjust that via the Vagrantfile if you want to use less.

Next, we install LXD in our two VMs using the bolt plan located at plans/install_lxd.pp. It's written in Puppet language. VS Code has a very good extension for Puppet, so check it out.

bolt plan run lab::install_lxd

You can read the code, or you can get some basic information about the plan with bolt plan show lab::install_lxd. It works for any documented plan or task.

Now that we've installed LXD in our VMs, let's initialize one of them (alpha VM) as the cluster leader.

bolt plan run lab::lxd_init

Then we join the other node (beta) to create a two-node LXD cluster.

bolt plan run lab::lxd_join

The lxd_join plan extracted the cert from the alpha node bootstrapped in the lxd_init plan. A yaml configuration file was templated out, and uploaded to the beta node. This was used to join beta to alpha and form the cluster.

We can see the uploaded file by running an ad hoc command.

bolt command run --target beta 'cat /tmp/join.yaml'

Bolt can help us download it as a backup, too.

bolt file download -t beta /tmp/join.yaml /tmp

We can also check that our clustering worked by running lxc inside the VMs.

bolt command run -t alpha,beta 'lxc cluster list' --run_as root

Root is required, because our ssh user isn't a member of the lxd group.

Let's fix that.

bolt command run -t alpha,beta "usermod -aG lxd $USER" --run_as root

Configure lxc

To use the Bolt+LXD inventory integration, we need to configure a new LXD remote for our little lab cluster.

Accept the certificate. The remote trust password is "demo".

lxc remote add lab 172.100.10.2:8443

If successful, switch to our new remote and print the cluster members.

lxc remote switch lab
lxc cluster list

Launch Containers

Remember the containers that were already listed in our inventory? Now it's time to create them. They're called "one" and "two", and we'll use an Ubuntu 20.10 base container image for both.

lxc launch ubuntu:20.10 one
lxc launch ubuntu:20.10 two

Now we can run a command inside both containers with bolt.

bolt command run -t one,two 'uptime'

Let's run another, but this time we can use our "containers" group. This is defined in inventory.yaml, and aggregates targets "one" and "two".

bolt command run -t containers 'apt-get update'

Finally, because we have configured clustering with a fan network overlay, our containers can route to each other, even across hosts.

# get the ip for container two
lxc info two | awk '/eth0:\s{1}inet\s+/ {print $3}'

# run a single ping to test connectivity
bolt command run -t one 'ping -c 1 <two's ip>'

TODO

• Try something more complicated