Building Throwaway Machines with Terraform

Building Throwaway Machines with Terraform

8/27/2019 • 3 min read

Looking for an easy way to quickly try out some new whiz-bang software safely and securely? Look no further!

At my day job at Molecula, I often find myself needing to perform a simple experiment and I run off and install a bunch of tools, run a bunch of tests, make brilliant observations and then go onto the next Big Thing. The problem with this approach is if I run my experiments on my local laptop, it becomes cluttered and the cruft builds up into an unmanageable rats nest. When I run my experiments on machines in the cloud, many juicy artifacts are often lost when my cloud machine is destroyed. I usually capture the major focus of the experiement in notes, but later I realize that I missed something that was contained in the output of a command or even what commands I actually executed. So following the grunt work principle, this task must be automated.


First, install the awesome Terraform, a cloud provisioning tool from HashiCorp. This basically requires putting an executable in your path, but the install process is described in detail at HashiCorp.

Once this is installed, you have to craft a few files that describe which cloud provider to connect to and how to set it up. The links below are the ones I use for GCP (Google Cloud Platform), and those files assume you have exported your google credentials to a local file credentials.json file from the GCP dashboard, you use an RSA key (~/.ssh/ in your home directory and you have generated a gist cli access token (~/.gist), the process to generate is described here. After those files are in place simply init and apply and you are off to the races. You will have to replace the project value in the with project_id value located your credentials.json

terraform init
terraform apply

Once it’s all up and running you can access that machine by

ssh ubuntu@`terraform output ip`

And finally cleanup when you’re done with

terraform destroy

All ssh sessions are recorded to log files which reside in the ubuntu home directory /home/ubuntu of the generated machine. When the machine is destroyed via terraform those logs are uploaded to a gist which has the nice feature of making the content of the gist indexed and searchable via search. Don’t forget to limit to just your gists by adding the user filter that looks something like user:tgruben. It should be noted that gists are publicly viewable. You can add the private flag -p from the destroy hook in and the content is not discoverable but it is still viewable by the public if you can find the link.

I hope you find this useful.

Get even more from Pilosa

Stay updated on our progress and upcoming events.