Local preview

Coder is typically deployed to a remote data center, but you can use Docker to create a lightweight preview deployment of Coder.

Coder currently supports local preview only on workstations running macOS or Linux.

Coder automatically uploads a single-seat license upon installation.

Prerequisites

Before proceeding, please make sure that you have the following installed:

1. Docker
2. helm
3. kind or Docker Desktop
4. kubectl

Limitations

We do not recommend using local previews for production deployments of Coder.

Resource allocation and performance

Your experience with the local Coder preview is dependent on your system specs, but please note that you can expect slightly degraded performance due to the deployment running entirely inside a Docker container.

CVMs

The kind deployment supports CVMs if you meet the following requirements (if you choose not to try out CVMs, these requirements do not apply):

1. Your Linux hosts must be running Linux Kernel 5 and above.

2. You must have the linux-headers package corresponding to your Kernel version installed. You should see the following folders all corresponding to your Kernel version:

   $ uname -r
   5.11.4-arch1-1
   $ ls /usr/lib/modules
   5.11.4-arch1-1
   $ ls /usr/src/
   linux  linux-headers-5.11.4-arch1-1
   

3. Docker Desktop for Mac must use version 2.5.0.1. This specific version is required because of a recent downgrade to Linux Kernel 4.9 due to a bug.

Dev URLs

Currently, the local preview doesn't support dev URLs. Instead, you can use tools like ngrok to preview webpages from inside an workspace.

We are working on bringing Dev URL support to local previews in later releases.

SSH

When using kind for your local preview, SSH is not configured to run by default.

When using Docker Desktop for your local preview, SSH works as long as your machine does not have an existing SSH server running on port 22.

Air-gapped clusters

The local preview option does not work in an air-gapped deployment.

Option 1: Kind

To install Coder, run:

curl -fsSL https://coder.com/try.sh | PORT="8080" sh -s --

Note: you can edit the value of PORT to control where the Coder dashboard will be available.

When the installation process completes, you'll see the URL and login credentials you need to access Coder:

You can now access Coder at

    http://localhost:8080

You can tear down the deployment with

    curl -fsSL https://coder.com/try.sh | sh -s -- down

Platform credentials
User:     admin
Password: yfu...yu2

Visit the URL, and log in using the provided credentials. The platform is automatically configured for you, so there's no first-time setup to do.

Option 2: Docker Desktop

Docker Desktop includes a standalone Kubernetes server and client that you can use to run Coder.

1. Enable the Kubernetes cluster inside Docker.

2. Ensure that Docker has enough resources allocated to meet Coder's requirements (you can do so by going to Docker preferences).

   

   

3. Install metrics-server so that Coder gets valid metrics from your cluster:

   helm repo add bitnami https://charts.bitnami.com/bitnami
   

4. Install Coder on to your cluster.

If you run into OutOfmemory errors when installing, try increasing your resource allocation in Docker. If increasing the resource allocation doesn't fix the error, reinstall Coder using the following Helm values:

helm upgrade --install coder \
    --set ingress.useDefault=false \
    --set cemanager.resources.requests.cpu="0m" \
    --set cemanager.resources.requests.memory="0Mi" \
    --set envproxy.resources.requests.cpu="0m" \
    --set envproxy.resources.requests.memory="0Mi" \
    --set timescale.resources.requests.cpu="0m" \
    --set timescale.resources.requests.memory="0Mi" \
    coder/coder

Removing Coder

To remove the local Coder deployment, run:

curl -fsSL https://coder.com/try.sh | sh -s -- down

Because Coder runs inside Docker, you should have nothing left on your machine after tear down.