GCP Compute Engine
In this guide, you will learn how to deploy the Coder control plane instance and your first template.
Requirements
This guide assumes you have roles/compute.instanceAdmin.v1 access to your
Google Cloud Platform project.
Launch a Coder instance from the Google Cloud Marketplace
We publish an Ubuntu 22.04 VM image with Coder and Docker pre-installed. Search
for Coder v2 in the GCP Marketplace or
use direct link.
Be sure to keep the default firewall options checked so you can connect over HTTP, HTTPS, and SSH.
We recommend keeping the default instance type (e2-standard-4, 4 cores and 16
GB memory) if you plan on provisioning Docker containers as workspaces on this
VM instance. Keep in mind this platforms is intended for proof-of-concept
deployments and you should adjust your infrastructure when preparing for
production use. See: Scaling Coder
Be sure to add a keypair so that you can connect over SSH to further configure Coder.
After launching the instance, wait 30 seconds and navigate to the public IPv4 address. You should be redirected to a public tunnel URL.
That's all! Use the UI to create your first user, template, and workspace. We recommend starting with a Docker template since the instance has Docker pre-installed.
Configuring Coder server
Coder is primarily configured by server-side flags and environment variables. Given you created or added key-pairs when launching the instance, you can configure your Coder deployment by logging in via SSH or using the console:
ssh ubuntu@<gcp-public-IPv4>
sudo vim /etc/coder.d/coder.env # edit config
sudo systemctl daemon-reload
sudo systemctl restart coder # restart Coder
Give developers VM workspaces (optional)
Instead of running containers on the Coder instance, you can offer developers full VM instances with the gcp-linux template.
Before you can use this template, you must authorize Coder to create VM instances in your GCP project. Follow the instructions in the gcp-linux template README to set up authentication.