DEV Community

hayao-k for AWS Community Builders

Posted on • Originally published at hayao-k.dev

Configure actions-runner-controller with proxy in private EKS cluster

Goals of this post

  • Run actions-runner-controller on Private EKS cluster
  • Communication with GitHub API is through on-premises proxy server

This post is validated with the following configuration.

  • Amazon EKS: Kubernetes 1.22
  • actions-runner-controller: v0.23.0
  • cert-manager v1.8.0
  • eksctl: 0.93.0
  • kubectl: v1.21.2
  • AWS CLI: 2.5.7

Image description

What is actions-runner-controller?

Controller for operating GitHub Actions self-hosted runners on Kubernetes cluster.

GitHub logo actions / actions-runner-controller

Kubernetes controller for GitHub Actions self-hosted runners

Actions Runner Controller (ARC)

CII Best Practices awesome-runners Artifact Hub

About

Actions Runner Controller (ARC) is a Kubernetes operator that orchestrates and scales self-hosted runners for GitHub Actions.

With ARC, you can create runner scale sets that automatically scale based on the number of workflows running in your repository, organization, or enterprise. Because controlled runners can be ephemeral and based on containers, new runner instances can scale up or down rapidly and cleanly. For more information about autoscaling, see "Autoscaling with self-hosted runners."

You can set up ARC on Kubernetes using Helm, then create and run a workflow that uses runner scale sets. For more information about runner scale sets, see "Deploying runner scale sets with Actions Runner Controller."

People

Actions Runner Controller (ARC) is an open-source project currently developed and maintained in collaboration with the GitHub Actions team, external maintainers @mumoshu and @toast-gear, various contributors, and the awesome community.

If you think the…

The following describes the flow of building actions-runner-controller on private EKS cluster.

Create private EKS Cluster

Set up kubectl, eksctl, and AWS CLI in your working environment in advance. See below for detailed instructions.

kubectl: https://kubernetes.io/docs/tasks/tools/

eksctl: https://docs.aws.amazon.com/eks/latest/userguide/eksctl.html

AWS CLI: https://docs.aws.amazon.com/cli/latest/userguide/getting-started-install.html

Please refer to the following document to create a private cluster with eksctl.

https://eksctl.io/usage/eks-private-cluster/

In this case, created an EKS cluster using the following config file.

apiVersion: eksctl.io/v1alpha5
kind: ClusterConfig

metadata:
  name: actions-runner
  region: ap-northeast-1
  version: "1.22"

privateCluster:
  enabled: true
  skipEndpointCreation: true

vpc:
  subnets:
    private:
      ap-northeast-1a:
        id: subnet-xxxxxxxxxxxxxxxxx
      ap-northeast-1c:
        id: subnet-yyyyyyyyyyyyyyyyy
      ap-northeast-1d:
        id: subnet-zzzzzzzzzzzzzzzzz
Enter fullscreen mode Exit fullscreen mode

For a private EKS cluster, the following VPC endpoints are required.

  • ECR (ecr.api, ecr.dkr)
  • S3 (Gateway)
  • EC2
  • STS
  • CloudWatch Logs

If you enable a private cluster with eksctl (privateCluster.enabled), these endpoints will be created automatically.

In this case, I had a requirement to use a VPC already connected via Direct Connect, so skip endpoint creation with privateCluster.skipEndpointCreation to use the existing VPC and endpoints.

NOTE: if you specify an existing VPC, eksctl will edit the route table.

However, if the subnet is associated with the main route table, eksctl will not edit the route table and will fail to create the cluster. Therefore, it is necessary to create and associate a route table explicitly.

eksctl communicates with the AWS API via a proxy server. However, communication to private EKS cluster endpoints does not use a proxy server, so the following environment variable should be set.

export https_proxy=http://xxx.xxx.xxx.xxx:yyyy
export no_proxy=.eks.amazonaws.com
Enter fullscreen mode Exit fullscreen mode

The following command creates a cluster. Creating a private cluster takes longer than creating a regular cluster. This configuration took about 22 minutes, but if you are creating VPC or VPC endpoints, it will take even longer, so adjust the timeout value.

$ eksctl create cluster --timeout 30m -f private-cluster.yaml
Enter fullscreen mode Exit fullscreen mode

Register Managed Node Groups

Add UserData to the launch template used by the managed node group and configure the Docker daemon to use a proxy.

This template is based on the following blog.

https://yomon.hatenablog.com/entry/2020/10/eks_docker_inside_proxy#%E3%83%9E%E3%83%8D%E3%83%BC%E3%82%B8%E3%83%89%E5%9E%8B%E3%83%8E%E3%83%BC%E3%83%89%E3%82%B0%E3%83%AB%E3%83%BC%E3%83%97%E3%81%ABProxy%E8%A8%AD%E5%AE%9A

AWSTemplateFormatVersion: "2010-09-09"
Parameters:
  ClusterIp:
    Type: String
  ProxyIP:
    Type: String
  ProxyPort:
    Type: Number
  SshKeyPairName:
    Type: String

Resources:
  EKSManagedNodeInPrivateNetworkLaunchTemplate:
    Type: AWS::EC2::LaunchTemplate
    Properties: 
      LaunchTemplateName: eks-managednodes-in-private-network
      LaunchTemplateData: 
        InstanceType: t3.small
        KeyName: !Ref SshKeyPairName
        TagSpecifications: 
          - ResourceType: instance
            Tags:
              - Key: Name 
                Value: actions-runner-nodegroup
          - ResourceType: volume
            Tags:
              - Key: Name
                Value: actions-runner-nodegroup

        UserData: !Base64 
          "Fn::Sub": |
            Content-Type: multipart/mixed; boundary="==BOUNDARY=="
            MIME-Version:  1.0

            --==BOUNDARY==
            Content-Type: text/cloud-boothook; charset="us-ascii"

            #Set the proxy hostname and port
            PROXY=${ProxyIP}:${ProxyPort}
            MAC=$(curl -s http://169.254.169.254/latest/meta-data/mac/)
            VPC_CIDR=$(curl -s http://169.254.169.254/latest/meta-data/network/interfaces/macs/$MAC/vpc-ipv4-cidr-blocks | xargs | tr ' ' ',')

            #Create the docker systemd directory
            mkdir -p /etc/systemd/system/docker.service.d

            #Configure docker with the proxy
            cloud-init-per instance docker_proxy_config tee <<EOF /etc/systemd/system/docker.service.d/http-proxy.conf >/dev/null
            [Service]
            Environment="HTTP_PROXY=http://$PROXY"
            Environment="HTTPS_PROXY=http://$PROXY"
            Environment="NO_PROXY=${ClusterIp},$VPC_CIDR,localhost,127.0.0.1,169.254.169.254,.internal,s3.amazonaws.com,.s3.ap-northeast-1.amazonaws.com,api.ecr.ap-northeast-1.amazonaws.com,dkr.ecr.ap-northeast-1.amazonaws.com,ec2.ap-northeast-1.amazonaws.com,ap-northeast-1.eks.amazonaws.com"
            EOF

            #Reload the daemon and restart docker to reflect proxy configuration at launch of instance
            cloud-init-per instance reload_daemon systemctl daemon-reload 
            cloud-init-per instance enable_docker systemctl enable --now --no-block docker
            --==BOUNDARY==
Enter fullscreen mode Exit fullscreen mode

After creating the launch template, create a Managed Node Group with the following config file.

apiVersion: eksctl.io/v1alpha5
kind: ClusterConfig

metadata:
  name: actions-runner
  region: ap-northeast-1

managedNodeGroups:
- name: t3s-proxy
  desiredCapacity: 1
  privateNetworking: true
  launchTemplate:
    id: lt-xxxxxxxxxxxxxxxxx
    version: "1"
Enter fullscreen mode Exit fullscreen mode

Run create nodegroup.

$ eksctl create nodegroup -f nodegroup.yaml
Enter fullscreen mode Exit fullscreen mode

Installing actions-runner-controller

actions-runner-controller uses cert-manager to manage certificates for Admission Webhooks, so it must be installed in advance.

$ kubectl apply -f https://github.com/cert-manager/cert-manager/releases/download/v1.8.0/cert-manager.yaml
Enter fullscreen mode Exit fullscreen mode

To use actions-runner-controller in a proxy environment, it is necessary to configure the proxy server information in the following three locations.

  1. manager container of controller-manager
  2. runner container of RunnerDeployment
  3. sidecar (dind) container of RunnerDeployment

I deployed actions-runner-controller with kubectl this time.

1 sets the https_proxy environment variable to the Deployment definition of controller-manager.
2 and 3 set the https_proxy environment variable to the definition of RunnerDeployment.

Edit the manifest file of actions-runner-controller as follows.

33836     spec:
33837       containers:
33838       - args:
33839         - --metrics-addr=127.0.0.1:8080
33840         - --enable-leader-election
33841         command:
33842         - /manager
31843         env:
31844         - name: GITHUB_TOKEN
31845           valueFrom:
31846             secretKeyRef:
31847               key: github_token
31848               name: controller-manager
31849               optional: true
31850         - name: GITHUB_APP_ID
31851           valueFrom:
31852             secretKeyRef:
31853               key: github_app_id
31854               name: controller-manager
31855               optional: true
31856         - name: GITHUB_APP_INSTALLATION_ID
31857           valueFrom:
31858             secretKeyRef:
31859               key: github_app_installation_id
31860               name: controller-manager
31861               optional: true
31862         - name: GITHUB_APP_PRIVATE_KEY
31863           value: /etc/actions-runner-controller/github_app_private_key
+ 31864       - name: http_proxy
+ 31865         value: "http://xxx.xxx.xxx.xxx:xxxx"
+ 31866       - name: https_proxy
+ 31867         value: "http://xxx.xxx.xxx.xxx:xxxx"
+ 31868       - name: no_proxy
+ 31869         value: "172.20.0.1,*.eks.amazonaws.com"
31870         image: summerwind/actions-runner-controller:v0.22.3
31871         name: manager
Enter fullscreen mode Exit fullscreen mode

Run kubectl create.

$ kubectl create -f actions-runners-controller.yaml
Enter fullscreen mode Exit fullscreen mode

NOTE: Due to a known issue with Kubernetes, we must use create (or replace when updating) instead of apply.

https://github.com/actions-runner-controller/actions-runner-controller/issues/1317

GitHub API can be authenticated using the GitHub App or Personal Access Token (PAT).

In this case, we use PAT authentication. Set the appropriate scope for the token depending on the type of runner and create a secret.

kubectl create secret generic controller-manager \
    -n actions-runner-system \
    --from-literal=github_token=${GITHUB_TOKEN}
Enter fullscreen mode Exit fullscreen mode

NOTE: PAT must be pre-authorized before accessing an Organization with SAML SSO configured.

https://docs.github.com/ja/enterprise-cloud@latest/authentication/authenticating-with-saml-single-sign-on/authorizing-a-personal-access-token-for-use-with-saml-single-sign-on

RunnerDeployment is defined below.

apiVersion: actions.summerwind.dev/v1alpha1
kind: RunnerDeployment
metadata:
  name: runner-deploy
spec:
  template:
    spec:
      enterprise: <enterprise-name>
      labels:
        - runner-container
      env:
        - name: RUNNER_FEATURE_FLAG_EPHEMERAL
          value: "true"
        - name: http_proxy
          value: "http://xxx.xxx.xxx.xxx:yyyy"
        - name: https_proxy
          value: "http://xxx.xxx.xxx.xxx:yyyy"
        - name: no_proxy
          value: "172.20.0.1,10.1.2.0/24,*.eks.amazonaws.com"
      dockerEnv:
        - name: http_proxy
          value: "http://xxx.xxx.xxx.xxx:yyyy"
        - name: https_proxy
          value: "http://xxx.xxx.xxx.xxx:yyyy"
        - name: no_proxy
          value: "172.20.0.1,10.1.2.0/24,*.eks.amazonaws.com"
---
apiVersion: actions.summerwind.dev/v1alpha1
kind: HorizontalRunnerAutoscaler
metadata:
  name: runner-deployment-autoscaler
spec:
  scaleDownDelaySecondsAfterScaleOut: 300
  scaleTargetRef:
    name: runner-deploy
  minReplicas: 1
  maxReplicas: 5
  metrics:
  - type: PercentageRunnersBusy
    scaleUpThreshold: '0.75'
    scaleDownThreshold: '0.25'
    scaleUpFactor: '2'
    scaleDownFactor: '0.5'
Enter fullscreen mode Exit fullscreen mode

When a container is used in a GitHub Actions job, the docker container for dind, launched as a RunnerDeployment sidecar, is used. Specify dockerEnv so that this dind container can perform image pulls.

After deployment, confirm that the pod is successfully started and is visible from GitHub.

$ kubectl apply -f RunnerDeployment.yaml

$ kubectl get pod
NAME                            READY   STATUS    RESTARTS   AGE
pod/runner-deploy-xxxxx-xxxxx   2/2     Running   0          53s
Enter fullscreen mode Exit fullscreen mode

If controller-manager does not start properly, check the settings in 1 or GitHub; if Runner does not start properly, check the settings in 2 or 3.

Top comments (0)