Phase 1 GCP - Prerequisites

Server v4.7

Server Admin

CircleCI server v4.7 is installed as a Helm chart. The installation process is broken down into four phases. There is a validation step at the end of each phase, allowing you to confirm success before moving to the next phase. Depending on your requirements, phases 3 and 4 may include multiple steps. This installation guide assumes you have already read the CircleCI server v4.7 overview.

In the following sections, replace any sections indicated by < > with your details.

1. Install required software

Download and install the following software before continuing:

Tool	Version	Used for	Notes
kubectl	1.19 or greater	Kubernetes CLI
Helm	3.9.2 or greater	Kubernetes Package Management
Helm Diff	3.5.0 or greater	Helping with `values.yaml` changes and updates	Optional, but may help with troubleshooting between releases
`gcloud` and `gsutil` as part of the Google Cloud SDK.	Latest	Managing GCP resources	GCP installs only
Terraform	0.15.4 or greater	Infrastructure Management	Required for installations within GCP or AWS
Install and configure MinIO CLI for your storage provider if you choose not to use AWS or GCP object storage.	TBC	TBC	Required for installations outside AWS and GCP, for example, local installation.

2. Create a VPC

Installing server locally? you can skip this step, but you should ensure you follow the resourcing recommendation of using /18 CIDR blocks to ensure you have adequate capacity for both your Kubernetes Cluster and the provisioning of Nomad clients to run your jobs.

If you are installing server in the cloud, either AWS or GCP, you will need to create a new virtual private cloud (VPC).

Refer to the GCP guide to creating a VPC.

VPC and cluster sizing recommendations

While there are no strict requirements for VPC (Virtual Private Cloud) setup or disk size, the following practices are recommended for optimal performance and stability in production environments.

VPC Selection

We recommend you create a new VPC specifically for CircleCI server and its components, rather than using an existing VPC.
Subnet Provisioning

For high availability, and to avoid potential outages, you should provision subnets using /18 CIDR blocks across multiple Availability Zones.

2. Create a Kubernetes cluster

CircleCI server installs into an existing Kubernetes cluster. If you have not already created a cluster, you should do so next. There are instructions for creating a cluster below, or if you are installing locally, first consider the sizing, version, and permissions requirements.

Cluster requirements

Compute resources

The application uses a large number of resources. Depending on your usage, your Kubernetes cluster should meet the following requirements:

Number of daily active CircleCI users	Minimum Nodes	Total CPU	Total RAM	NIC speed
< 500	4	24 cores	90 GB	1 Gbps
500+	6	48 cores	240 GB	10 Gbps

Supported Kubernetes versions

CircleCI Version	Kubernetes Version
4.7.x	1.26 - 1.29

Minimum permissions requirements

The installing user must have at least admin permissions for the namespace into which CircleCI is to be installed.

GKE

You can learn more about creating a GKE cluster in the GKE docs.

Do not use an Autopilot cluster. CircleCI requires functionality that is not supported by GKE Autopilot.

Install and configure the GCP CLI for your GCP account. This includes creating a Google Project, which will be required to create a cluster within your project.

When you create your project, make sure you also enable API access. If you do not enable API access, the command we will run next (to create your cluster) will fail.

Setting the default project id, compute zone and region will make running subsequent commands easier:
```
gcloud config set project <PROJECT_ID>
gcloud config set compute/zone <ZONE>
gcloud config set compute/region <REGION>
```

Create your cluster

CircleCI recommends using Workload Identity to allow workloads/pods in your GKE clusters to impersonate Identity and Access Management (IAM) service accounts to access Google Cloud services. Use the following command to provision a simple cluster:

gcloud container clusters create circleci-server \
  --num-nodes 5 \
  --machine-type n1-standard-8 \
  --workload-pool=<PROJECT_ID>.svc.id.goog \
  --network <NETWORK> \
  --subnetwork <SUBNETWORK>

Your kube-context should get updated with the new cluster credentials automatically.

If you need to update your kube-context manually, you can by running the following:

gcloud container clusters get-credentials circleci-server

Install the GKE authentication plugin for kubectl:

gcloud components install gke-gcloud-auth-plugin

Verify your cluster:
```
kubectl cluster-info
```

Enable Workload Identity in GKE (optional)

Follow these steps if you already have a GKE cluster and need to enable Workload Identity on the cluster and the node pools.

Enable Workload Identity on existing cluster:

  gcloud container clusters update "<CLUSTER_NAME>" \
    --workload-pool="<PROJECT_ID>.svc.id.goog"

Get node pools of existing GKE cluster:

  gcloud container node-pools list --cluster "<CLUSTER_NAME>"

Update existing node pools:

  gcloud container node-pools update "<NODEPOOL_NAME>" \
    --cluster="<CLUSTER_NAME>" \
    --workload-metadata="GKE_METADATA"

You must repeat Step 3 for all the existing node pools. Follow these links for steps to enable Workload Identity for your Kubernetes service accounts:

3. Create a new GitHub OAuth app

If GitHub Enterprise and CircleCI server are not on the same domain, then images and icons from GHE will fail to load in the CircleCI web app.

Registering and setting up a new GitHub OAuth app for CircleCI server allows for authorization control to your server installation using GitHub OAuth and for updates to GitHub projects/repos using build status information. The following steps apply for both GitHub.com and GitHub Enterprise.

In your browser, navigate to menu:Your-GitHub-instance[User Settings>Developer Settings>OAuth Apps] and select New OAuth App.

Figure 1. New GitHub OAuth App
Complete the following fields, based on your planned installation:
- Homepage URL: The URL of your planned CircleCI installation.
- Authorization callback URL: The authorization callback URL is the URL of your planned CircleCI installation followed by /auth/github
Once completed, you will be shown the Client ID. Select Generate a new Client Secret to generate a Client Secret for your new OAuth App. You will need these values when you configure CircleCI server.

Figure 2. Client ID and Secret

GitHub Enterprise

If using GitHub Enterprise, you also need a personal access token and the domain name of your GitHub Enterprise instance.

Create the defaultToken by navigating to User Settings > Developer Settings > Personal access tokens. The default token requires no scopes. You will need this value when you configure CircleCI server.

4. Static IP addresses

It is recommended to provision a static IP address to assign to the load balancer created by the cluster. While this is not necessary, it does eliminate the need to update DNS records if the service-created load balancer is reprovisioned.

GCP: Reserve a static external IP address

The Google Cloud docs provide information on how reserve an IP address.

Make note of the returned IPv4 address for use later in the values.yaml file.

5. Frontend TLS certificates

By default, CircleCI server creates self-signed certificates to get you started. In production, you should supply a certificate from a trusted certificate authority. The Let’s Encrypt certificate authority, for example, can issue a free certificate using their Certbot tool. The sections below cover using Google Cloud DNS and AWS Route 53.

It is important that your certificate contains both your domain and the app.* subdomain as subjects. For example, if you host your installation at server.example.com, your certificate must cover app.server.example.com and server.example.com.

Once you have created your certificates using one of the methods described below, you can use the following commands to retrieve the certificates later when you need them during this installation:

ls -l /etc/letsencrypt/live/<CIRCLECI_SERVER_DOMAIN>

cat /etc/letsencrypt/live/<CIRCLECI_SERVER_DOMAIN>/fullchain.pem

cat /etc/letsencrypt/live/<CIRCLECI_SERVER_DOMAIN>/privkey.pem

Google Cloud DNS

If you host your DNS on Google Cloud, you need the certbot-dns-google plugin installed. You can install the plugin with the following command:
```
python3 -m pip install certbot-dns-google
```

The service account used to run certbot will need to have access to Cloud DNS in order to provision the necessary records used by Let’s Encrypt for domain validation.

Create a service account:

gcloud iam service-accounts create <SERVICE_ACCOUNT_ID> --description="<DESCRIPTION>" \
  --display-name="<DISPLAY_NAME>"

Retrieve credentials for the service account:

gcloud iam service-accounts keys create <KEY_FILE> \
  --iam-account <SERVICE_ACCOUNT_ID>@<PROJECT_ID>.iam.gserviceaccount.com

Create a custom role for Certbot:

gcloud iam roles create certbot --project=<PROJECT_ID> \
    --title="<TITLE>" --description="<DESCRIPTION>" \
    --permissions="dns.changes.create,dns.changes.get,dns.changes.list,dns.managedZones.get,dns.managedZones.list,dns.resourceRecordSets.create,dns.resourceRecordSets.delete,dns.resourceRecordSets.list,dns.resourceRecordSets.update" \
    --stage=ALPHA

Bind the new role to the service account which we created earlier:

gcloud projects add-iam-policy-binding <PROJECT_ID> \
    --member="serviceAccount:<SERVICE_ACCOUNT_ID>@<PROJECT_ID>.iam.gserviceaccount.com" \
    --role="<ROLE_NAME>"

Finally, the following commands will provision a certification for your installation:

certbot certonly --dns-google --dns-google-credentials <KEY_FILE> -d "<CIRCLECI_SERVER_DOMAIN>" -d "app.<CIRCLECI_SERVER_DOMAIN>"

Upstream TLS termination

You may have a requirement to terminate TLS for CircleCI server outside the application. This is an alternate method to using ACM or supplying the certificate chain during Helm deployment. An example would be a proxy running in front of the CircleCI installation providing TLS termination for your domain name. In this case the CircleCI application acts as the backend for your load balancer or proxy.

CircleCI server listens on the following port numbers, which need to be configured depending how you are routing the traffic:

Frontend / API Gateway [TCP 80, 443]
Nomad server [TCP 4647]

Depending on your requirements you may choose to terminate TLS for only the frontend/api-gateway or provide TLS for services listening on all the ports.

6. Encryption/signing keys

The keysets generated in this section are used to encrypt and sign artifacts generated by CircleCI. You will need these values to configure server.

Store these values securely. If they are lost, job history and artifacts will not be recoverable.

a. Artifact signing key

To generate an artifact signing key, run the following command:

docker run circleci/server-keysets:latest generate signing -a stdout

b. Encryption signing key

To generate an encryption signing key, run the following command:

docker run circleci/server-keysets:latest generate encryption -a stdout

7. Object storage and permissions

CircleCI server v4.7 hosts build artifacts, test results, and other state object storage. The following storage options are supported:

While any S3 compatible object storage may work, we test and support AWS S3 and MinIO. Follow the instructions below to create a bucket and access method for AWS S3 or GCS.

If you are installing locally rather than in AWS or GCP, follow the MinIO instructions in the Air-gapped installation docs.

If you are installing behind a proxy, object storage should be behind this proxy also. Otherwise, proxy details will need to be supplied at the job level within every project .circleci/config.yml to allow artifacts, test results, cache save and restore, and workspaces to work. For more information see the Installing server behind a proxy guide.

Google Cloud Storage

a. Create a GCP bucket

If your server installation runs in a GKE cluster, ensure that your current IAM user is cluster admin for this cluster, as RBAC (role-based access control) objects need to be created. More information can be found in the GKE documentation.

gsutil mb gs://circleci-server-bucket

b. Set up authentication

The recommended method for workload/pod authentication is to use Workload Identity. However, you may also use static credentials (JSON key file).

Create a Service Account.

gcloud iam service-accounts create circleci-storage --description="Service account for CircleCI object storage" --display-name="circleci-storage"

Bind the objectAdmin role to the service account.

gcloud projects add-iam-policy-binding <PROJECT_ID> \
    --member="serviceAccount:circleci-storage@<PROJECT_ID>.iam.gserviceaccount.com" \
    --role="roles/storage.objectAdmin" \
    --condition='expression=resource.name.startsWith("projects/_/buckets/circleci-server-bucket"),title=restrict_bucket'

Either enable Workload Identity or use static credentials.
Option 1: Workload Identity

When using Workload Identity you need to configure your account such that the workloads/pods can access the storage bucket from the cluster using the Kubernetes service account "<K8S_NAMESPACE>/object-storage".

gcloud projects add-iam-policy-binding <PROJECT_ID> \ --member serviceAccount:circleci-storage@<PROJECT_ID>.iam.gserviceaccount.com \ --role roles/iam.workloadIdentityUser \ --condition=None gcloud iam service-accounts add-iam-policy-binding circleci-storage@<PROJECT_ID>.iam.gserviceaccount.com \ --role roles/iam.workloadIdentityUser \ --member "serviceAccount:<PROJECT_ID>.svc.id.goog[<K8S_NAMESPACE>/object-storage]"

gcloud projects add-iam-policy-binding <PROJECT_ID> \ --member serviceAccount:circleci-storage@<PROJECT_ID>.iam.gserviceaccount.com \ --role roles/iam.serviceAccountTokenCreator \ --condition=None
Option 2: Static credentials

If you are not using Workload Identity, create a JSON file containing static credentials.

gcloud iam service-accounts keys create <KEY_FILE> \ --iam-account circleci-storage@<PROJECT_ID>.iam.gserviceaccount.com