vagrantfiles | A Virtual Machine for every language

I have had a number of issues recently with vagrant-berkshelf not syncing Chef cookbooks on an existing machine reliably. And, basically, when doing research on workarounds, I see something like:

vagrant-berkshelf is deprecated, use Test Kitchen instead.

My use case is that I have Vagrantfiles, used to build VMs and DigitalOcean droplets, that are hand-written and only use Chef to provision the VMs. I am most definitely approaching Chef as a user, not an author or tester of cookbooks.

So, I am in a case of Vagrant -> Chef, not Chef -> Vagrant.

When looking at Kitche-Vagrant, I see that:

The kitchen-vagrant driver for Kitchen generates a single Vagrantfile for each instance of Kitchen in a sandboxed directory..

My question is: if my workflow relies on hand-written, complex, Vagrantfiles, can I continue to use Chef as a provisioner without having to rely on vagrant-berkshelf?

Some of the possible alternatives I see are:

• mangle Test Kitchen configuration to work with my exiting Vagrantfile. I fear that this is not the intent of this tool and will not end well.

• use chef.cookbooks_path attribute in vagrant and let it take the place of vagrant-berkshelf.

• switch out provisioners and use say Vagrant->Ansible.

The Vagrantfile below is somewhat simplified, but the gist is that the Vagrantfile is in charge and Chef is just used to provision.

I am currently setting up a development environment whereby there are two vagrant machines. The first hosts a website etc while the second hosts an API which the website calls. Currently my vagrantfiles are as follows:

API

I am trying to set up a server environment as follows:

• Physical Host machine: (192.168.0.106)
• Physical Laptop: (192.168.0.130)

On the host machine, I am trying to use vagrant to set up 2 Virtual Machines

• Virtual Machine 1 (VM_IP_1). Vagrant is set up to port forward 5555:8080. This machine will be running some sort of webserver, on port 8080
• Virtual Machine 2 (VM_IP_2).

I am trying to accomplish the following:

1. From Physical Host Machine, I want to be able to open up a web browser, navigate to VM_IP_1:5555, and connect to the web server, displaying some web page
2. From Laptop Machine, I want to be able to open up a web browser, navigate to VM_IP_1:5555, and connect to the web server, displaying some web page (same as 1)
3. From Virtual Machine 2, I want to be able to open up a web browser, navigate to VM_IP_1:5555, and connect to the web server, displaying some web page (same as 1).

Is this even possible? I've spent the last 2 days, tinkering, trying to automate it, but no luck.

Basically, I want to simulate a non-trivial network architecture consisting of web servers, load balancers, databases, etc. I want to be able to test as if my single powerful server was a simulation of a network (i.e it would run 4 VMs) but also have the flexibility and re usability of my vagrant automation to be able to deploy to production something like (2 powerful servers, each running 2 VMS for a total of 4)

Everyone should be able to communicate with everyone (assuming I have the ports forwarded correctly - routing tables too)

Here are the VagrantFiles I've been playing around with. Tried with private_network but I don't think that's right. There's Host Only but I don't think that's right eitherI am thinking that maybe I need to have some sort of reverse port_forwarding from VM 2. But that doesn't sound right.

I use nano for git commit messages. Short summary (<=50 chars) plus a new line before the description is relatively straightforward to stick to. However, when it comes to wrapping at 72 chars in the description body I just go off what seems to look right, making for inconsistent logs.

In Vagrantfiles I've seen this sort of thing to tell the editor what to do for vi/vim: