Recently I needed a quick way to spin-up a VPN Gateway and generate the openvpn config clients could use to connect.

There is a good guide to generating the necessary certificates and manually editing the openvpn config you can download from the portal in the official docs.

Being a sucker for punishment I wondered if I could automate the process (mainly because I always forget the openssl commands) and secondly it was to be run by someone else (not me) so I wanted it to be as simple as possible.

Warning: Please take time to understand the limitations of this certificate generation before production use (TF State file containing CA private keys) and review the limitations in the TLS provider for Terraform.

So how would this work? First choice is Terraform for the automation. Luckily that has a provider for generating certs! So the flow goes like this:

1. Create the Root CA
2. Use that to generate a client cert
3. Output the client cert for use in the openvpn config file
4. Inject the CA into the Azure VPN configuration and create it
5. Run a script to fetch the Azure VPN OpenVPN configuration file (as this contains the Pre-shared key we don’t set) then inject the client cert we outputted from the Terraform.
6. Connect

One gotcha – The TLS provider only outputs standard pems. Azure VPN requires the CA to be in a specific format outputted by the following command.

openssl x509 -in caCert.pem -outform der | base64 -w0 > caCert.der

As a result there is a null_resource and local_file resource to handle this translation.

So all you should need is Terraform, Bash, OpenSSL and Azure CLI – not perfect but it’s the best I could do! (This is currently a working draft – please use with caution).

[Brain dump so I don’t forget this one]

So you want your bash script to exit on an error but you’d like it to clean some stuff up before it closes after the error occurs.

No problem a TRAP can do this for you (read detailed docs for caveats).

In a very simple form it looks like this:

Learn more here: https://www.linuxjournal.com/content/bash-trap-command

For some testing I’m doing I need a set of images of a specific size to simulate pulling larger vs smaller image.

Here is a quick script I put together for generating a 200mb, 600mb, 1000mb and 2000mb image (tiny bit larger as alpine included). Took a while to work out best to use /dev/urandom not /dev/zero as with zero the images got compressed for transfer.