Telepresence with Google Cloud Kubernetes Engine (GKE)
In my current project Qwiz’n’Buzz we are actively working on a discord integration as an Discord Activity. In sake of user protection, Discord uses a proxy as a middleman for requests to our services. Additionally, the Discord SDK relies on your application been integrated in a iframe provided by Discord. This brings challenges for fast local development processes.
To test the integration locally Discord suggests cloudflared to tunnel the local service to a public endpoint. Unless you are using a paid plan, the endpoint URL is ephemeral and changes between restarts. This requires you have to update the Discord Activity URL Mapping settings every time you restart the tunnel.
This is where I remembered Telepresence. Telepresence allows you to proxy a local development environment into a remote Kubernetes cluster. This enables you to test and debug services within the context of the full system without deploying the service to the cluster. This way, we can provision stable development domains and cluster infrastructure to iterate quickly on the Discord integration locally.
Telepresence brings two ways for redirecting traffic from a kubernetes service to your local machine. The first way replaces the service-backing pod with a Telepresence pod that forwards traffic to your local machine. The second pattern adds a sidecar container (traffic-agent) to the service-backing pod that forwards traffic to your local machine. The second pattern is the default behavior and is the one I will focusing in this post.
