VMs created using Vagrant with a badly configured vagrantfile may have DNS issues

[JNasri]

Description

When setting up the local multi-node cluster using Vagrant, the VMs lose domain name resolution (DNS). Commands like apt-get update or ping google.com fail to resolve, even though raw internet access via IP address still works (e.g., ping 8.8.8.8 succeeds).

Why this happens

When Vagrant configures the network interface required for internal Kubernetes traffic, the OS network scripts overwrite the DNS settings originally provided by the primary NAT internet adapter. Because the private network has no DNS server assigned, the VMs are left with an empty or broken configuration.

The Fix

To prevent this, your Vagrantfile needs to be configured to force VirtualBox (or any virtualization software you use) to proxy the host machine's DNS down to the guest.

Updating the machines.each block in the vagrantfile to include these settings permanently fixes the issue:

  machines.each do |name, specs|
    config.vm.define name do |machine|
      machine.vm.hostname = name

      machine.vm.network "private_network",
        ip: specs[:ip]

      machine.vm.provider "virtualbox" do |virtualbox|
        virtualbox.name = "kthw-#{name}"
        virtualbox.cpus = 1
        virtualbox.memory = specs[:memory]
        
        # Force VirtualBox to proxy host DNS directly to the VM
        virtualbox.customize ["modifyvm", :id, "--natdnshostresolver1", "on"]
        virtualbox.customize ["modifyvm", :id, "--natdnsproxy1", "on"]
      end

      # Force a public fallback DNS inside the guest OS config file
      machine.vm.provision "shell", inline: <<-SHELL
        echo "nameserver 8.8.8.8" | sudo tee /etc/resolv.conf
      SHELL
    end
  end

Note: The forcing of a public nameserver during the shell provisioning step is optional. I added it to double check.