Traffic Director

This snippet focus on exploring traffic director in GCP to manage communication between VM and pods in GKE through envoy proxy. Part 1 Focus in setting up with VM.

Background

Traffic Director is a managed control plane for application networking. Which allow us the user to setup communication between services in a similar manner of ‘Service Mesh’ but managed by GCP and also with several additional features beyond service mesh such as:

Multi Cluster Kubernetes
Virtual Machine
Proxyless gRPC
Ingress and Gateways
Multiple Environment

One of my interests is how we can establish communication between VM and Containers within Kubernetes which some of us may not have expertise in creating configuration or exposing services from Kubernetes to an external environment (inside VM). Within this story, I want to focus on how we can create communication between services in GCE as a Virtual Machine and in GKE as Pods.

Notes: The objective for this story is for a conceptual test, for production configuration please do check with the official documentation.

Test

To run the test of the traffic director which creates communication between VM and Pods I test on the following setup:

Simplified version of VM (with Envoy) communicating with Pods (with Envoy and only utilize ClusterIP as service type) through traffic director

VM with Traffic Director

So the first setup is to deploy a VM template with Envoy installed and communicate with the Traffic Director. The simplest way to do that is to add options of:

--service-proxy=enabled

To setup such we can follow the documentation Traffic Director setup for Compute Engine VMs with automatic Envoy deployment

[One] Create a VM template which I am using as an example of a simple VM for LoadBalancer test and enable the service proxy.

This is only an example of my generated gcloud command for other purposes but adding the service proxy option at the end. You may change and create your own configuration.

instance template created

[Two] Create a Managed Instance Group using the created template. I set off the autoscaler for simplicity and set the qty to 2. Go to Compute Engine > Instance Groups > Create Instance Group.

set the instance group name, location and other configuration as needed

MIG is created and ready to be used as backend

[Three] Create Traffic Director Service and configure the Routing Rule Maps. Go to Network Services > Traffic Director > Create Service. (Ensure that we already enable the Traffic Director API or please go and enable it prior to continuing the steps).

Get Started

Continue

Configure Service

Configure the service adding the name + network (VPC) + select the backend type of Instance Group and add the MIG that we have created earlier. Add the health check and ensure that the firewall is not blocking the health check ( I already configured it; if you are wondering the steps to do that please go to the documentation). Then press continue

Traffic Director Service is created and backend is healthy !

Now we need to attached routing rule maps by click the Routing Rule Maps tab > Create Routing Rule Map.

Add forwarding rule with name, network (VPC), and set the Internal IP that we can address the rule map for Traffic Director and last but not least the port

Then we need to add the routing rules, for simplicity I just redirect everything to the traffic director service we created earlier

And now we test to call the ip of the forwarding rule

It works !

We can see that the VM is able to call the traffic director and the traffic director will load balance to the VMs inside the MIG. Let’s continue our test with creating the GKE cluster and set up pods with envoy to communicate later on. This will be continued on the next post (Part 2)