Start Multiple netlab Labs on the Same Server

A heavy netlab user sent me an email along these lines:

We’re running multiple labs in parallel on the same server, and we’re experiencing all sorts of clashes like overlapping management IP addresses. We “solved” that by using static device identifiers in our labs, but I’m wondering if there’s a better way of doing it?

That’s exactly the sort of real-life challenges I love working on, so it wasn’t hard to get me excited, and the results are bundled in netlab release 1.5.

The Problem

Let’s start with what problem are we trying to solve?:

  • netlab assumes that the underlying provisioning system creates a management network that can be used to access network devices.
  • vagrant-libvirt plugin and containerlab have a default name for the management network, and so netlab (by default) doesn’t mess with it.
  • vagrant-libvirt does not adjust the /etc/hosts file (containerlab does), so it’s crucial to have well-known management IP addresses assigned to the network devices. netlab solves that by using the default vagrant-libvirt management network subnet together with static MAC addresses and DHCP bindings.

Starting two labs on the same server will inevitably result in overlapping IP addresses unless you specify unique node id values in every lab topology. Obviously not something one can rely on for a stable production-grade solution.

But wait, it gets worse. vagrant-libvirt and containerlab augment VM names or container names with a prefix to make them unique. netlab uses the same approach to name libvirt networks or Linux bridges it needs to create. The default prefix is the current directory name, which means that when two users start the same lab all sorts of weird things could happen.

Finally, the UDP tunnels used to implement libvirt point-to-point links used IP addresses calculated from node id and interface ifindex. Starting two labs with point-to-point links might also have some interesting side effects.

Use Cases

It quickly became evident that different users might want to use the multiple labs per server functionality in different ways:

  • A single user starting multiple labs in different directories.
  • Multiple users starting parallel labs (using the same or different lab topology) on the same system.
  • Some environments would want to have an even more flexible approach with on-demand allocation of lab resources.

The solution has to be flexible enough to support all three use cases, and simple enough that it could be used without external components.

The Solution

Here’s what we decided to implement:

  • The multiple labs functionality is implemented with a new multilab plugin, making the changed behavior completely optional. If you don’t use that plugin, you get the traditional netlab setup.
  • Every lab instance using multilab plugin must have a unique lab ID. Assigning the lab ID, and making sure it’s unique1, is the users’ responsibility.
  • multilab plugin uses the lab ID to make netlab parameters like the management subnet, the management network name, the VM/container prefix, VIF interface names and libvirt tunnel endpoints unique.
  • Modified parameters are used when creating Vagrantfile, containerlab topology or Ansible inventory, ensuring that each lab instance gets a unique set of resources.

Here’s how you can use this functionality to support the three use cases:

  • Multiple labs per user: set the parameter in lab topology or as a CLI parameter.
  • Multiple users with one lab per user: set the user-wide default in ~/.netlab.yml file.
  • Flexible allocation: build an external component that keeps track of lab IDs and a wrapper around netlab command that sets CLI parameter.

If you have a better idea, please open a GitHub issue or discussion.

Finally, it would be an interesting exercise in futility to try to enforce the use of multilab plugin with every lab topology, so I added the capability to specify the default list of plugins used by all lab topologies as plugin parameter in a defaults file – add plugin: [ multilab ] to ~/.netlab.yml file and you’re good to go.

Getting Started

To get more details and learn about additional features included in release 1.5.0, read the release notes. To upgrade, execute pip3 install --upgrade networklab.

New to netlab? Start with the Getting Started document and the installation guide.

  1. In release 1.5.0, the uniqueness of the lab ID is not checked. I hope to have at least rudimentary checks in release 1.5.1. ↩︎

Add comment