Skip to content
Sample code for Cloud Run to use a static IP for outgoing requests via a SSH tunnel over a GCE instance
Python Shell Dockerfile
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.
.gitignore Initial import Jul 24, 2019
LICENSE Initial import Jul 24, 2019
requirements.txt Initial import Jul 24, 2019

Static Outbound IP example for Cloud Run applications

This repository contains an example of a Google Cloud Run application that runs an SSH tunnel through a GCE instance within the container to route outbound requests of the Cloud Run application through the static IP of the GCE instance.

⚠️ Read the accompanying blog post as well:

Before you begin

  1. Launch Google Cloud Shell (recommended, as it has al the tools required pre-installed).

  2. Clone this repository and cd into it.

Create a tunnel instance on GCE

  1. Create a set of ssh key pairs so that your container can SSH into the VM.

    ssh-keygen -q -f ssh_key

    Note that the private SSH key, which is a secret, will be bundled into the container image, which can be compromised if anyone gets access to your source code/build system. You can also use other means of delivering this key to the container in the runtime (e.g. by downloading from a GCS bucket, or using a secrets manager).

  2. Create a Google Compute Engine instance (f1-micro in us-central1 with name "tunnel"):

    gcloud compute instances create "tunnel" \
        --zone=us-central1-b \
  3. (Optional) You can go to the Cloud Console and promote this VM’s ephemeral IP address to be a "static IP address". But, long as you don't delete this VM, its IP address will not change.

  4. Upload the SSH public key (not a secret) to the VM to authenticate as user "tunnel":

    gcloud compute instances add-metadata "tunnel" \
        --metadata-from-file ssh-keys=<(echo "tunnel:$(cat")

(Optional) Inspect the application source code

Take time to understand:

  • runs a SSH client (as SOCKS5 TCP proxy server via GCE VM) and the flask Python application server.

    This script sets HTTPS_PROXY=socks5://localhost:5000 environment variable to the Python app to use the proxy. However, this HTTPS_PROXY environment variable works the same way on many other languages, including Go as well.

    By setting HTTPS_PROXY environment variable you don't need to update your code to use the SOCKS5 proxy.

  • waits for the SSH port-forwarding server configured via HTTPS_PROXY to be accessible.

  • starts a flask app querying and sends its result back.

  • Dockerfile invokes via tini init system.

Deploy Cloud Run application

  1. Set up $PROJECT variable in your shell to your current project.

    PROJECT="$(gcloud config get-value core/project -q)"
  2. Build and push the container image to Google Conatiner Registry.

    gcloud builds submit --tag$PROJECT/sample-tunnel
  3. Find the EXTERNAL_IP address of the Compute Engine VM named "tunnel" you created earlier:

    gcloud compute instances list --filter=name=tunnel
  4. Deploy to Cloud Run, by setting GCE_IP environment variable to the IP of the VM:

    gcloud beta run deploy sample-tunnel \
        --set-env-vars="GCE_IP=x.y.z.t" \
        --platform=managed \
        --region us-central1 \
        --allow-unauthenticated \$PROJECT/sample-tunnel

Query the application

When you visit the application’s public URL, you will see that the IP address that it used to query is the IP address of the GCE instance.

You can visit the application’s /exit endpoint to crash the Cloud Run container instance, which would normally have a new dynamic IP assigned to the instance, but in this case it remains the same as Cloud Run holds onto the request.

Don't forget to check out the accompanying blog post:

You can’t perform that action at this time.