Using Custom Vagrant Boxes with netlab

A friend of mine started using Vagrant with libvirt years ago (it was his enthusiasm that piqued my interest in this particular setup, eventually resulting in netlab). Not surprisingly, he’s built Vagrant boxes for any device he ever encountered, created quite a collection that way, and would like to use them with netlab.

While I didn’t think about this particular use case when programming the netlab virtualization provider interface, I decided very early on that:

  • Everything worth changing will be specified in the system defaults
  • You will be able to change system defaults in topology file or user defaults.

Changing Vagrant Box or Container Name

Parameters describing individual devices are grouped under devices dictionary in system parameters. Each device is described as another dictionary, and that dictionary contains image names (box- or container names) for every virtualization provider supported by that device, for example:

devices:
  eos:
    virtualbox:
      image: arista/veos
    clab:
      image: ceos:4.26.4M
    libvirt:
      image: arista/veos

To change a Vagrant box used by a device type, you have to change one (or all) of these settings.

Changing System Defaults in Lab Topology

You can change system defaults in your lab topology. Because you’re changing a default, you have to make that change within the defaults dictionary, for example:

Changing Arista cEOS container image in lab topology
defaults:
  devices:
    eos:
      clab:
        image: ceos:4.27.0F

You probably hate using a ruler to figure out proper YAML indentation as much as I do, so I added a neat trick to netlab YAML files – you can use dotted name syntax (similar to how sane programming languages deal with objects):

Changing Arista cEOS container image in lab topology – dotted syntax
defaults.devices.eos.clab.image: ceos:4.27.0F

Each dotted name is expanded into a dictionary, and that dictionary is merged with the lab topology (Box package does a wonderful job doing that), allowing you do use the same path in multiple settings:

Changing multiple Arista images
defaults.devices.eos.clab.image: ceos:4.27.0F
defaults.devices.eos.libvirt.image: arista/veos:4.27.0F

You can also make your future life miserable and do something like this:

Just because you could doesn’t mean that you should
defaults.devices.eos:
  clab.image: ceos:4.27.0F
  libvirt.image: arista/veos:4.27.0F

Changing System Defaults in User Defaults File(s)

If you want to use the same changed defaults for multiple lab topologies (which is what my friend would want to do), you can store the changed defaults in topology-defaults.yml (applies to all topologies in the same directory) or ~/.netlab.yml (applies to everything the current user does).

As you’re changing settings within a defaults file, you MUST NOT use the defaults prefix. For example: to change Arista images for the current user, use the following settings in ~/.netlab.yml:

Changing Arista images in user defaults
devices.eos.clab.image: ceos:4.27.0F
devices.eos.libvirt.image: arista/veos:4.27.0F

Changing Usernames and Passwords

I try to use vagrant/vagrant as username/password combination in all Vagrant boxes I’m building (Junos is the only exception as it enforces password complexity rules we all love so much). Your boxes might have different usernames.

Not a problem: you have to change the group_vars device settings – a dictionary containing Ansible variables for that device type, including authentication variables ansible_user and ansible_ssh_pass. These are the default settings for Arista EOS:

Ansible variables for Arista devices
devices:
  eos:
    group_vars:
      ansible_user: vagrant
      ansible_ssh_pass: vagrant
      ansible_network_os: eos
      ansible_connection: network_cli

Arista is a particularly tough nut: cEOS uses different default username than vEOS, so we had to add support for provider-specific group variables. Even worse, vEOS boxes I build use privileged user vagrant, cEOS uses non-privileged user admin, so we have to set ansible_become as well.

Usernames and passwords for Arista Vagrant boxes and containers
devices:
  eos:
    group_vars:
      ansible_user: vagrant
      ansible_ssh_pass: vagrant
    clab:
      group_vars:
        ansible_user: admin
        ansible_ssh_pass: admin
        ansible_become: yes
        ansible_become_method: enable    

Now that you know how to change system defaults, it shouldn’t be too hard to change username for libvirt Arista vEOS box to foobar:

Changing default username for Arista vEOS libvirt box
devices.eos.libvirt.group_vars.ansible_user: foobar

Changing Vagrantfile

The final hurdle my friend was facing: he didn’t like the way netlab sets up NX-OS boxes in generated Vagrantfile:

libvirt Vagrantfile template for NX-OS boxes
    {{ name }}.vm.synced_folder ".", "/vagrant", disabled: true
    {{ name }}.ssh.insert_key = false
    {{ name }}.ssh.shell = "run bash"
    {{ name }}.vm.boot_timeout = 360
    {{ name }}.vm.guest = :freebsd

    {{ name }}.vm.provider :libvirt do |domain|
      domain.cpus = 2
      domain.features = ['acpi']
      domain.loader = "/usr/share/OVMF/OVMF_CODE.fd"
      domain.memory = 6144
      domain.disk_bus = "sata"
      domain.disk_device = "sda"
      domain.disk_driver :cache => "none"
      domain.nic_model_type = "e1000"
      domain.graphics_type = "none"
      domain.driver = "kvm"
      {% if "amd" in defaults.processor|lower %}
      domain.cpu_mode = "custom"
      {% endif %}
    end

netlab release 1.2.4 adds support for user Jinja2 templates – all netlab code using Jinja2 templates1 looks for templates in current directory, ~/.netlab directory, and Python package2. Templates used to create Vagrantfile are within the provider-specific subdirectory (libvirt or virtualbox).

To change the template used to create Vagrantfile configuration for a Nexus OS libvirt box:

  • If you want to use the changed template for a specific topology, create libvirt directory within the directory the with lab topology.
  • If you want to use the changed template for all labs, create libvirt directory within the ~/.netlab directory.
  • Find the original template in netlab sources (example: libvirt Nexus OS template).
  • Copy that template with the same file name into libvirt directory and modify it as needed.
  • Enjoy using netlab with your custom Vagrant boxes.

  1. But not the Ansible playbooks used by netlab up or netlab initial ↩︎

  2. To see the template search path, run netlab create --debug and look for TEMPLATE PATH messages. ↩︎

1 comments:

  1. Wondering if you do any user input validation from those YAML formatted files with netsim first? Or is it user's responsibility to provide correct input values?

    Replies
    1. Yes and Yes ;)

      Sometimes I feel like half of the code of optional configuration modules is input validation. Don't take my word for it, there's a reason the code is on GitHub.

Add comment
Sidebar