Amazon EKS Networking Bootcamp

This repository serves as a guide for the Amazon EKS Networking Bootcamp. Throughout this bootcamp, which focuses on EKS networking, you'll delve into different networking solutions and tackle IP exhaustion concerns. Additionally, you'll gain proficiency in crafting and enforcing network policies at the workload level, safeguarding your applications with enhanced security measures.

Networking and Policy options for Amazon EKS with Calico

Amazon EKS runs upstream Kubernetes, so you can install alternate compatible CNI plugins to Amazon EC2 nodes in your cluster.

Important

If you have Fargate nodes in your cluster, the Amazon VPC CNI plugin for Kubernetes is already on your Fargate nodes. It's the only CNI plugin you can use with Fargate nodes. An attempt to install an alternate CNI plugin on Fargate nodes fails.

Amazon EKS maintains relationships with a network of partners that offer support for alternate compatible CNI plugins. For this workshop we will focus on Calico options for EKS.

Amazon VPC networking with Calico Policy Enforcement

The geeky details of what you get:

Policy	IPAM	CNI	Overlay	Routing	Datastore
Calico	AWS	AWS	No	VPC Native	Kubernetes

Calico network policy enforcement on an EKS cluster using the AWS VPC CNI plugin.

EKS with Calico networking

The geeky details of what you get:

Policy	IPAM	CNI	Overlay	Routing	Datastore
Calico	Calico	Calico	VXLAN	Calico	Kubernetes

Deploying Calico as the Container Network Interface (CNI), IP Address Management (IPAM), and for network policy enforcement brings numerous advantages to EKS. It not only resolves issues related to IP exhaustion but also enables additional features like the Calico Egress Gateway.

Calico networking options for EKS

Calico CNI is designed to offer users the flexibility to select the most suitable dataplane option according to their needs. Calico supports several dataplanes, including:

• Standard Linux (iptables)
• eBPF (Extended Berkeley Packet Filter)
• Windows HNS (Host Networking Service)
• VPP (Vector Packet Processing )

Module 1: Calico Cloud on EKS - Workshop Environment Preparation

Before diving into the installation of Calico in EKS and configuring various dataplane options, let's first set up our AWS CloudShell environment for seamless execution of commands and configurations.

Getting Started with AWS CloudShell

To begin the bootcamp, you'll need to meet the following basic requirements:

• AWS Account AWS Console
• AWS CloudShell https://portal.aws.amazon.com/cloudshell
• Amazon EKS Cluster - to be created later in this bootcamp.

Instructions

1. Login to AWS Portal at https://portal.aws.amazon.com.

2. Open the AWS CloudShell.

   

3. Install the bash-completion on the AWS CloudShell.

   sudo yum -y install bash-completion

4. Configure the kubectl autocomplete.

   # set up autocomplete in bash into the current shell, bash-completion package should be installed first.
   source <(kubectl completion bash) 
   # add autocomplete permanently to your bash shell.
   echo "source <(kubectl completion bash)" >> ~/.bashrc 

   You can also use a shorthand alias for kubectl that also works with completion:

   alias k=kubectl
   complete -o default -F __start_kubectl k
   echo "alias k=kubectl"  >> ~/.bashrc
   echo "complete -o default -F __start_kubectl k" >> ~/.bashrc
   /bin/bash

5. Install the eksctl - Installation instructions

   mkdir ~/.local/bin
   curl --silent --location "https://github.com/weaveworks/eksctl/releases/latest/download/eksctl_$(uname -s)_amd64.tar.gz" | tar xz -C /tmp
   sudo mv /tmp/eksctl ~/.local/bin && eksctl version 

6. Install the K9S, if you like it.

   curl --silent --location "https://github.com/derailed/k9s/releases/download/v0.32.5/k9s_Linux_amd64.tar.gz" | tar xz -C /tmp
   sudo mv /tmp/k9s ~/.local/bin && k9s version

7. Clone this repository in your Azure Cloud Shell.

   git clone https://github.com/tigera-solutions/eks-networking-bootcamp.git && \
   cd eks-networking-bootcamp

Module 2: EKS cluster with AWS VPC CNI and Calico for Network Policy

This module provisions an EKS cluster with AWS VPC CNI and connects it to Calico Cloud.

Create an EKS cluster with Amazon VPC CNI

1. Define the environment variables to be used by the resources definition.

Note

In this section, we'll create some environment variables. If your terminal session restarts, you may need to reset these variables. You can do that using the following command:

source ~/labVars.env

# Feel free to use the cluster name and the region that better suits you.
export CLUSTERNAME1=eks-calico
export REGION=us-west-2
export VERSION=1.29
export INSTANCETYPE=m5.xlarge

# Persist for later sessions in case of disconnection.
echo "# Start EKS Bootcamp Lab Params" > ~/labVars.env
echo export CLUSTERNAME1=$CLUSTERNAME1 >> ~/labVars.env
echo export REGION=$REGION >> ~/labVars.env
echo export VERSION=$VERSION >> ~/labVars.env
echo export INSTANCETYPE=$INSTANCETYPE >> ~/labVars.env

2. Create the EKS cluster.

   eksctl create cluster \
     --name $CLUSTERNAME1 \
     --region $REGION \
     --version $VERSION \
     --node-type $INSTANCETYPE

3. Verify you have API access to your new EKS cluster

   kubectl get nodes

Tip

If you loose access to your EKS cluster, you can update the kubeconfig to restore the kubectl access to it by running the following command:

source ~/labVars.env

aws eks update-kubeconfig --name $CLUSTERNAME1 --region $REGION

After provisioning the EKS cluster, the PODs will be networked using Amazon VPC CNI, enabling routable IPs. For enhanced security and observability features offered by Calico, you have the option to connect your cluster to Calico Cloud or install Calico Enterprise on it.

Connect your cluster to Calico Cloud

Follow instructions in the official Calico Cloud documentation to connect your EKS cluster to Calico Cloud management plane.

Module 3: Enforce Workload-level Network Policy

During this bootcamp, we will use the Cat Facts Application as an example to work on. In this module, we will learn how to use Calico to implement the workload-level network policies for each of the workloads.

1. Install the example application stack:

   From the cloned directory, execute:

   kubectl apply -f pre

Included in the pre folder, there is the application that will be used in the exercises during the workshop. The diagram below shows how the applications' microservices communicate between themselves.

There are also other objects that will be created. We will learn about them later.

Important

Wait until all the pods are up and running to move to the next step.

Service Graph

Connect to Calico Cloud GUI. From the menu, select Service Graph > Default. Explore the options.

Implement the Default Deny Principle

A global default deny policy ensures that unwanted traffic (ingress and egress) is denied by default. Pods without policy (or incorrect policy) are not allowed traffic until the appropriate network policy is defined. Although the staging policy tool will help you find the incorrect or the missing policy, a global deny policy helps mitigate other lateral malicious attacks.

By default, all traffic is allowed between the pods in a cluster. Let's start by testing connectivity between application components and across application stacks. All of these tests should succeed as there are no policies in place.

We recommend creating a global default deny policy after you complete reviewing the policy for the traffic you want to allow. Use the stage policy feature to get your allowed traffic working as expected, then lock down the cluster to block unwanted traffic.

1. Create a staged global default-deny policy. It will show all the traffic that would be blocked if it were enforced.

   • Go to the Policies Board
   • On the bottom of the tier box default click on Add Policy
     • In the Create Policy page enter the policy name: default-deny
     • On the Applies To session, click Add Namespace Seletor First, lets apply only to the catfacts namespace
       • Select Key... kubernetes.io/metadata.name
       • =
       • Select Value... catfacts
     • On the field Type select both checkboxes: Ingress and Egress.
     • You are done. Click Stage on the top-right of your page.

   The staged policy does not affect the traffic directly but allows you to view the policy impact if it were to be enforced. You can see the denied traffic in the staged policy.

Enforce Workload-level Network Policy

Calico Security Policies provide a richer set of policy capabilities than the native Kubernetes network policies, including:

• Policies that can be applied to any endpoint: pods/containers, VMs, and/or to host interfaces
• Policies that can define rules that apply to ingress, egress, or both
• Policy rules support:
  • Actions: allow, deny, log, pass
  • Source and destination match criteria:
    • Ports: numbered, ports in a range, and Kubernetes named ports
    • Protocols: TCP, UDP, ICMP, SCTP, UDPlite, ICMPv6, protocol numbers (1-255)
    • HTTP attributes
    • ICMP attributes
    • IP version (IPv4, IPv6)
    • IP or CIDR
    • Endpoint selectors (using label expression to select pods, VMs, host interfaces, and/or network sets)
    • Namespace selectors
    • Service account selectors

Creating the policies

1. Based on the application design, the db lists on port 3306 and receive connections from the worker and the facts microservices. Let's use the Calico Cloud UI to create a policy to microsegment this traffic.

   • Go to the Policies Board
   • On the bottom of the tier box platform click on Add Policy
     • In the Create Policy page enter the policy name: db
     • Change the Scope from Global to Namespace and select the namespace catfacts
     • On the Applies To session, click Add Label Seletor
       • Select Key... app
       • =
       • Select Value... db
       • Click on SAVE LABEL SELECTOR
     • On the field Type select the checkbox for Ingress.
     • Click on Add Ingress Rule
       • On the Create New Policy Rule window,
         • Click on the dropdown with Any Protocol
         • Change the radio button to Protocol is and select TCP
         • In the field To: click on Add Port
         • Port is 3306 - Save
         • In the field From:, click Add Label Seletor
           • Select Key... app
           • =
           • Select Value... worker
           • Click on SAVE LABEL SELECTOR
           • On OR, click Add Label Seletor
           • Select Key... app
           • =
           • Select Value... facts
           • Click on SAVE LABEL SELECTOR
         • Click on the button Save Rule
     • You are done. Click Enforce on the top-right of your page.

2. Now, let's use the Recommend a Policy feature to create the policies for the other workloads.

   Let's start with the facts workload.

   

3. Now that you have learned how to create policies using Calico Cloud UI, go ahead and create microsegmentation policies for the worker workload.

Tip

• Look into the Cat Facts application diagram to figure out all the communications to and from the worker microservice.
• You can use the Recommend a Policy feature as well. Don't forget to edit the recommendations to make it as granular as possible, like restricting the domain names of the API's that the worker needs to communicate with (dog.ceo and catfact.ninja).

If you create all the policies correctly, at some point, you will start seeing zero traffic being denied by your default-deny staged policy. At that point, you can go ahead and enforce the default-deny policy. Voilà! The catfacts namespace is now secure.

Tip

• After enforcing the default-deny policy, if you need to troubleshoot, use the Service Graph or the Flow Visualizations tools to see what traffic is being blocked.

Clean up

First, we should delete the application to release the load balancer resource associated with it. Once the application is removed, we can proceed with deleting the EKS cluster.

1. Delete the applications stack to clean up any loadbalancer services.

   kubectl delete -f pre/40-catfacts-app.yaml 

2. Delete EKS cluster.

   source ~/labVars.env
   eksctl delete cluster --name $CLUSTERNAME1 --region $REGION

---

🎉 Congratulations! You've completed the Amazon EKS Networking Bootcamp.

If you have the time and are still eager to learn, consider exploring the following bonus modules!

• Bonus Module 1: EKS with Calico CNI
• Bonus Module 2: EKS with Windows nodegroup and Calico for Policy

---

Bonus Module 1: EKS with Calico CNI in eBPF mode

In this module, the EKS cluster is provisioned and configured to use Calico CNI in eBPF mode. With Calico in place, it will handle Policy, IPAM, and CNI responsibilities.

1. Define the environment variables to be used by the resources definition.

Note

In this section, we'll create some environment variables. If your terminal session restarts, you may need to reset these variables. You can do that using the following command:

source ~/labVars.env

# Feel free to use the cluster name and the region that better suits you.
export CLUSTERNAME2=eks-calico-ebpf

# Persist for later sessions in case of disconnection.
echo export CLUSTERNAME2=$CLUSTERNAME2 >> ~/labVars.env

2. Create the EKS cluster.

Note

If you intend to use Wireguard based encryption to security pod-to-pod communications, the EKS cluster must be provisioned to use AMI that includes Wireguard bits. One of such AMI's is Bottlerocket that is used in this example.

eksctl create cluster \
  --name $CLUSTERNAME2 \
  --region $REGION \
  --version $VERSION \
  --without-nodegroup

3. Verify you have API access to your new EKS cluster

   kubectl get pod -A

Tip

If you loose access to your EKS cluster, you can update the kubeconfig to restore the kubectl access to it by running the following command:

source ~/labVars.env

aws eks update-kubeconfig --name $CLUSTERNAME2 --region $REGION

4. Uninstall the AWS VPC CNI and install Calico CNI.

   To install Calico CNI, the first step is to remove the AWS VPC CNI, and then proceed with the installation of Calico. For further information about Calico CNI installation on AWS EKS, please refer to the Project Calico documentation

   Delete the aws-node daemonset resource to uninstall AWS VPN CNI

   kubectl delete daemonset -n kube-system aws-node

5. Install Calico CNI

   kubectl create -f https://raw.githubusercontent.com/projectcalico/calico/v3.28.2/manifests/tigera-operator.yaml

6. Create the installation configuration.

   kubectl create -f - <<EOF
   kind: Installation
   apiVersion: operator.tigera.io/v1
   metadata:
     name: default
   spec:
     kubernetesProvider: EKS
     cni:
       type: Calico
     calicoNetwork:
       bgp: Disabled
       linuxDataplane: BPF
       hostPorts: Disabled
   EOF

7. Check data plane configuration

   kubectl get installation.operator.tigera.io default -ojsonpath='{.spec.calicoNetwork.linuxDataplane}'

8. Create the nodegroup and the nodes. Two nodes are enough to demonstrate the concept.

   eksctl create nodegroup \
   --cluster $CLUSTERNAME2 \
   --region $REGION \
   --node-type $INSTANCETYPE \
   --node-ami-family Bottlerocket \
   --max-pods-per-node 100

   Wait for all Calico components to get deployed and only then proceed to next step.

   kubectl get tigerastatus

9. Configure Calico to talk directly to the API server

   • Get config values from EKS cluster settings

     APISERVER_ADDR=$(kubectl get configmap -n kube-system kube-proxy -o yaml | grep   server |  awk -F "/" '{print $3}')
     APISERVER_PORT=$(kubectl get endpoints kubernetes -ojsonpath='{.subsets[0].ports  [0].port}')

   • Configure kubernetes-services-endpoint configmap

     kubectl apply -f - <<-EOF
     kind: ConfigMap
     apiVersion: v1
     metadata:
       name: kubernetes-services-endpoint
       namespace: tigera-operator
     data:
       KUBERNETES_SERVICE_HOST: "${APISERVER_ADDR}"
       KUBERNETES_SERVICE_PORT: "${APISERVER_PORT}"
     EOF

10. Disable kube-proxy

In eBPF mode Calico replaces kube-proxy so it doesn't make sense to run both. Disable the kube-proxy in the cluster.

kubectl patch ds -n kube-system kube-proxy -p '{"spec":{"template":{"spec":{"nodeSelector":{"non-calico": "true"}}}}}'

11. [optional] Enable DSR mode

Direct return mode skips a hop through the network for traffic to services (such as node ports) from outside the cluster. This reduces latency and CPU overhead but it requires the underlying network to allow nodes to send traffic with each other's IPs. In AWS, this requires all your nodes to be in the same subnet and for the source/dest check to be disabled.

DSR mode is disabled by default; to enable it, set the BPFExternalServiceMode Felix configuration parameter to "DSR". This can only be done with calicoctl:

• Use the following command to download the calicoctl binary.

  curl -L https://github.com/projectcalico/calico/releases/download/v3.28.2/calicoctl-linux-amd64 -o kubectl-calico

• Set the file to be executable.

  chmod +x kubectl-calico

• Configure Calico to use DSR

  calicoctl patch felixconfiguration default --patch='{"spec": {"bpfExternalServiceMode": "DSR"}}'

If you want to learn more about eBPF dataplane, here are some links:

• Introducing the Calico eBPF dataplane
• eBPF Explained: Use Cases, Concepts, and Architecture
• eBPF XDP: The Basics and a Quick Tutorial
• Get started with XDP

With Calico CNI you avoid dependencies on AWS VPC CNI. Allocating pod IPs from the underlying VPC is problematic due to IP address range exhaustion challenges, or if the maximum number of pods supported per node by the Amazon VPC CNI plugin is not sufficient for your needs, we recommend using Calico networking in cross-subnet overlay mode. Pod IPs will not be routable outside of the cluster, but you can scale the cluster up to the limits of Kubernetes with no dependencies on the underlying cloud network.

Clean up Cluster 2

Important

Before deleting the EKS cluster, it's essential to uninstall any sample application and release the load balancer resources allocated to it. This ensures that no resources are left dangling and allows for a clean deletion of the cluster.

1. Delete EKS cluster.

   source ~/labVars.env
   eksctl delete cluster --name $CLUSTERNAME2 --region $REGION

Bonus Module 2: EKS with Windows nodegroup and Calico for Policy

In this module EKS cluster is provisioned with AWS VPC CNI option where each pod gets a routable IP assigned from the Amazon EKS VPC. A Windows nodegroup is also added to the cluster. Calico plugin is installed for network policy enforcement.

1. Define the environment variables to be used by the resources definition.

Note

In this section, we'll create some environment variables. If your terminal session restarts, you may need to reset these variables. You can do that using the following command:

source ~/labVars.env

# Feel free to use the cluster name and the region that better suits you.
export CLUSTERNAME3=eks-windows

# Persist for later sessions in case of disconnection.
echo export CLUSTERNAME3=$CLUSTERNAME3 >> ~/labVars.env

2. Create the EKS cluster.

   eksctl create cluster \
     --name $CLUSTERNAME3 \
     --region $REGION \
     --version $VERSION \
     --node-type $INSTANCETYPE

   • Verify you have API access to your new EKS cluster

     kubectl get nodes

Tip

If you loose access to your EKS cluster, you can update the kubeconfig to restore the kubectl access to it by running the following command:

source ~/labVars.env

aws eks update-kubeconfig --name $CLUSTERNAME3 --region $REGION

3. Enable Windows support for your Amazon EKS cluster

   kubectl apply -f - <<EOF
   apiVersion: v1
   kind: ConfigMap
   metadata:
     name: amazon-vpc-cni
     namespace: kube-system
   data:
     enable-windows-ipam: "true"
   EOF

4. Create the Windows nodegroup

   eksctl create nodegroup \
     --cluster=$CLUSTERNAME3 \
     --region=$REGION \
     --node-ami-family=WindowsServer2022CoreContainer

5. Get config values from EKS cluster settings

   APISERVER_ADDR=$(kubectl get configmap -n kube-system kube-proxy -o yaml | grep server |  awk -F "/" '{print $3}')
   APISERVER_PORT=$(kubectl get endpoints kubernetes -ojsonpath='{.subsets[0].ports[0].port}')
   SERVICE_CIDR=$(aws eks describe-cluster --name $CLUSTERNAME3 --region $REGION --query 'cluster.kubernetesNetworkConfig.serviceIpv4Cidr' --no-cli-pager | tr -d '"')

6. Configure kubernetes-services-endpoint configmap

   kubectl apply -f - << EOF
   kind: ConfigMap
   apiVersion: v1
   metadata:
     name: kubernetes-services-endpoint
     namespace: tigera-operator
   data:
     KUBERNETES_SERVICE_HOST: "${APISERVER_ADDR}"
     KUBERNETES_SERVICE_PORT: "${APISERVER_PORT}"
   EOF

7. Configure service CIDR in the Installation CR

   kubectl patch installation default --type merge --patch="{\"spec\": {\"serviceCIDRs\": [\"${SERVICE_CIDR}\"], \"calicoNetwork\": {\"windowsDataplane\": \"HNS\"}}}"

8. Install kube-proxy on Windows nodes.

   curl -L  https://raw.githubusercontent.com/kubernetes-sigs/sig-windows-tools/master/hostprocess/calico/kube-proxy/kube-proxy.yml | sed "s/KUBE_PROXY_VERSION/v1.28.2/g" | kubectl apply -f -

9. Enable firewall rules for Prometheus

   kubectl patch felixConfiguration default --type merge --patch '{"spec": {"windowsManageFirewallRules": "Enabled"}}'

Clean up Cluster 3

Important

Before deleting the EKS cluster, it's essential to uninstall any sample application and release the load balancer resources allocated to it. This ensures that no resources are left dangling and allows for a clean deletion of the cluster.

1. Delete EKS cluster.

   source ~/labVars.env
   eksctl delete cluster --name $CLUSTERNAME3 --region $REGION