# Workspace proxies

Workspace proxies provide low-latency experiences for geo-distributed teams.

<Callout type="info" title="Note">
  Workspace proxies require a [Premium license](https://coder.com/pricing#compare-plans).
  For more details, [contact your account team](https://coder.com/contact).
</Callout>

Coder's networking does a best effort to make direct connections to a workspace.
In situations where this is not possible, such as connections via the web terminal and [web IDEs](/docs/@main/workspaces/workspace-access/#other-web-ides), workspace proxies are able to reduce the amount of distance the network traffic needs to travel.

A workspace proxy is a relay connection a developer can choose to use when connecting with their workspace over SSH, a workspace app, port forwarding, and so on.
Dashboard connections and API calls (for example, the workspaces list) are not served over workspace proxies.

## Deploy a workspace proxy

Each workspace proxy should be a unique instance.
At no point should two workspace proxy instances share the same authentication token.
They only require port 443 to be open and are expected to have network connectivity to the control plane dashboard.
Workspace proxies **do not** make any database connections.

Workspace proxies can be used in the browser by navigating to the user `Account -> Workspace Proxy`

## Requirements

* The [Coder CLI](/docs/@main/reference/cli/) must be installed and authenticated as a user with the Owner role.

<div className="fd-steps">
  <div className="fd-step">
    ## Create the proxy

    Create the workspace proxy and make sure to save the returned authentication token for said proxy.
    This is the token the workspace proxy will use to authenticate back to primary coderd.

    ```sh
    $ coder wsproxy create --name=newyork --display-name="USA East" --icon="/emojis/2194.png"
    Workspace Proxy "newyork" created successfully. Save this token, it will not be shown again.
    Token: 2fb6500b-bb47-4783-a0db-dedde895b865:05271b4ef9432bac14c02b3c56b5a2d7f05453718a1f85ba7e772c0a096c7175
    ```

    To verify it was created.

    ```sh
    $ coder wsproxy ls
    NAME         URL                    STATUS STATUS
    newyork                             unregistered
    ```
  </div>

  <div className="fd-step">
    ## Deploy the proxy

    Deploying the workspace proxy will also register the proxy with the control plane and make the workspace proxy usable.
    If the proxy deployment is successful, `coder wsproxy ls` will show an `ok` status code:

    ```sh
    $ coder wsproxy ls
    NAME              URL                         STATUS STATUS
    primary           https://coder.example.com   ok
    brazil-saopaulo   https://brazil.example.com   ok
    europe-frankfurt  https://europe.example.com   ok
    sydney            https://sydney.example.com   ok
    ```

    Other Status codes:

    * `unregistered` : The workspace proxy was created, and not yet deployed
    * `unreachable` : The workspace proxy was registered, but is not responding.
      Likely the proxy went offline.
    * `unhealthy` : The workspace proxy is reachable, but has some issue that is preventing the proxy from being used.
      `coder wsproxy ls` should show the error message.
    * `ok` : The workspace proxy is healthy and working properly!

    ### Configuration

    Workspace proxy configuration overlaps with a subset of the `coderd` configuration.
    To see the full list of configuration options: `coder wsproxy server --help`

    ```sh
    # Proxy specific configuration. These are REQUIRED
    # Example: https://coderd.example.com
    CODER_PRIMARY_ACCESS_URL="https://<url_of_coderd_dashboard>"
    CODER_PROXY_SESSION_TOKEN="<session_token_from_proxy_create>"

    # Runtime variables for "coder start".
    CODER_HTTP_ADDRESS=0.0.0.0:80
    CODER_TLS_ADDRESS=0.0.0.0:443
    # Example: https://east.coderd.example.com
    CODER_ACCESS_URL="https://<access_url_of_proxy>"
    # Example: *.east.coderd.example.com
    CODER_WILDCARD_ACCESS_URL="*.<app_hostname_of_proxy>"

    CODER_TLS_ENABLE=true
    CODER_TLS_CLIENT_AUTH=none
    CODER_TLS_CERT_FILE="<cert_file_location>"
    CODER_TLS_KEY_FILE="<key_file_location>"

    # Additional configuration options are available.
    ```

    <a id="running-on-kubernetes"></a>

    ### Run on Kubernetes

    Make a `values-wsproxy.yaml` with the workspace proxy configuration.

    Notice the `workspaceProxy` configuration which is `false` by default in the Coder Helm chart:

    ```yaml
    coder:
      env:
        - name: CODER_PRIMARY_ACCESS_URL
          value: "https://<url_of_coderd_dashboard>"
        - name: CODER_PROXY_SESSION_TOKEN
          value: "<session_token_from_proxy_create>"
        # Example: https://east.coderd.example.com
        - name: CODER_ACCESS_URL
          value: "https://<access_url_of_proxy>"
        # Example: *.east.coderd.example.com
        - name: CODER_WILDCARD_ACCESS_URL
          value: "*.<app_hostname_of_proxy>"

      tls:
        secretNames:
          - kubernetes-wsproxy-secret

      # enable workspace proxy
      workspaceProxy: true
    ```

    Using Helm, install the workspace proxy chart

    ```sh
    helm install coder coder-v2/coder --namespace <your workspace proxy namespace> -f ./values-wsproxy.yaml
    ```

    Test that the workspace proxy is reachable with `curl -vvv`.
    If for some reason, the Coder dashboard still shows the workspace proxy is `UNHEALTHY`, scale down and up the deployment's replicas.

    <a id="running-on-a-vm"></a>

    ### Run on a VM

    ```sh
    # Set configuration options via environment variables, a config file, or cmd flags
    coder wsproxy server
    ```

    <a id="running-as-a-system-service"></a>

    ### Run as a system service

    If you've installed Coder via a [system package](/docs/@main/install/), you can configure the workspace proxy by settings in `/etc/coder.d/coder-workspace-proxy.env`

    To run workspace proxy as a system service on the host:

    ```sh
    # Use systemd to start workspace proxy now and on reboot
    sudo systemctl enable --now coder-workspace-proxy

    # View the logs to ensure a successful start
    journalctl -u coder-workspace-proxy.service -b
    ```

    To restart workspace proxy after applying system changes:

    ```sh
    sudo systemctl restart coder-workspace-proxy
    ```

    <a id="running-in-docker"></a>

    ### Run in Docker

    Modify the default entrypoint to run a workspace proxy server instead of a regular control plane.

    #### Docker Compose

    Change the provided [`compose.yml`](https://github.com/coder/coder/blob/main/compose.yaml) file to include a custom entrypoint:

    ```diff
      image: ghcr.io/coder/coder:${CODER_VERSION:-latest}
    + entrypoint: /opt/coder wsproxy server
    ```

    #### Docker run

    ```sh
    docker run --rm -it --entrypoint /opt/coder ghcr.io/coder/coder:latest wsproxy server
    ```

    #### Custom Dockerfile

    ```dockerfile
    FROM ghcr.io/coder/coder:latest
    ENTRYPOINT ["/opt/coder", "wsproxy", "server"]
    ```

    <a id="selecting-a-proxy"></a>

    ### Select a proxy

    Users can select a workspace proxy at the top-right of the browser-based Coder dashboard.
    Workspace proxy preferences are cached by the web browser.
    If a proxy goes offline, the session will fall back to the primary proxy.
    This could take up to 60 seconds.

    ![Workspace proxy picker](/docs/@main/images/admin/networking/workspace-proxies/ws-proxy-picker.png)
  </div>
</div>

## Multiple workspace proxies

When multiple workspace proxies are deployed:

* The browser measures latency to each available proxy independently.
* Users can select their preferred proxy from the dashboard.
* The system can automatically select the lowest-latency proxy.
* The dashboard latency indicator shows latency to the currently selected proxy.

## Observability

Coder workspace proxy exports metrics via the HTTP endpoint, which can be enabled using either the environment variable `CODER_PROMETHEUS_ENABLE` or the flag `--prometheus-enable`.

The Prometheus endpoint address is `http://localhost:2112/` by default.
You can use either the environment variable `CODER_PROMETHEUS_ADDRESS` or the flag `--prometheus-address <network-interface>:<port>` to select a different listen address.
