This sample deploys a Cloud Run service with VPC ingress controls that only allows traffic from Cloud HTTPS load balancer that has IAP (Identity Aware Proxy) enabled.
IAP authenticates users with a Google account (or other external IdP) and checks if the user is allowed to access the deployed service.
You must have a domain name registered, for which you can modify certain DNS settings. You can use a subdomain as well.
Create a new project, if you don't want to re-use an existing project. For example:
gcloud projects create cloud-run-iap-terraform-demo-1 --name="IAP Cloud Run Demo" --set-as-default
If this project isn't set as your current default project, use:
gcloud config set project cloud-run-iap-terraform-demo-1
Enable billing, if you haven't done this yet, for the project. First get your billing account ID:
gcloud beta billing accounts list
This returns something like:
ACCOUNT_ID NAME OPEN MASTER_ACCOUNT_ID
0X0X0X-0X0X0X-0X0X0X My Billing Account True
Now enable billing for your new project:
gcloud beta billing projects link cloud-run-iap-terraform-demo-1 --billing-account 0X0X0X-0X0X0X-0X0X0X
Enable the following APIs in your (currently selected) project:
gcloud services enable \
run.googleapis.com \
containerregistry.googleapis.com \
compute.googleapis.com \
iap.googleapis.com \
vpcaccess.googleapis.com
It might take a minute to complete. Optionally verify that these services are enabled for your current project:
gcloud services list
Create OAuth2 web application (and note its client_id and client_secret) from https://console.cloud.google.com/apis/credentials.
If it asks you to Configure a consent screen first, then do so:
- Choose 'External' for User Type
- Fill in 'App name', e.g.
IAP Cloud Run Demo
- Select your email address from the dropdown, for 'User support email'
- You can leave the 'App domain' section blank
- Fill in your email address at the 'Developer contact information'
- Click on 'save and continue' on the next screens and then the 'Credentials' menu link in the left menu bar
First initialize the Terraform:
terraform init
(Optional) preview what resources will be created:
terraform apply -var project_id=<project-id> \
-var region=<region> \
-var domain=<domain> \
-var lb_name=<lb_name> \
-var iap_client_id=<client_id> \
-var iap_client_secret=<client_secret>
Deploy using Terraform, this will take several minutes to complete:
terraform apply -var project_id=<project-id> \
-var region=<region> \
-var domain=<domain> \
-var lb_name=<lb_name> \
-var iap_client_id=<client_id> \
-var iap_client_secret=<client_secret> \
-auto-approve
In this command fill in:
- project-id: the project id, e.g.
cloud-run-iap-terraform-demo-1
- region: the region, this is optional, will default to
us-central1
- domain: the domain name you want to use, e.g.
corpapp.ahmet.dev
- lb_name: name for the load balancer, this is optional, will default to
iap-lb
- iap_client_id: the client id for the oAuth client you created earlier
- iap_client_secret: the client secret for the oAuth client you created earlier. You can view it by editing the oAuth client.
Note: if you get an error saying Output refers to sensitive values
, add sensitive = true
to backend_services
in .terraform/modules/lb-http/modules/serverless_negs/outputs.tf
.
If you get an error like Error: Error creating Connector: googleapi: Error 403: The caller does not have permission
,
it means the account which is used to execute the terraform, does not have enough permissions. You can fix this by running:
gcloud auth application-default login
If that still doesn't solve the issue, you can view the service account of your project in the GCP console under
IAM & Admin
-> 'Service Accounts' named e.g. Default compute service account
. From there you can troubleshoot which
permissions might be missing.
After the deployment is complete:
- Configure DNS records of the used domain name with the given IP address. In other words, add a record named
www
or your subdomain name, of typeA
pointing to your ipv4 IP address, which was the output of your terraform command. It will take 10-30 minutes until the load balancer starts functioning. - Go back to Credentials page and add a the
oauth2_redirect_uri
output to the Authorized Redirect URIs of the web application you created here.
Now, you should be able to authenticate to your web application based on
the users/group specified in main.tf
by visiting your domain.
.tfstate
file created in this directory.
Run the previous terraform apply
command as terraform destroy
to clean up
the created resources.
The following diagram shows the infrastructure created.
The Serverless VPC Access is the connection between your Cloud Run service and resources inside your VPC.
The Cloud Run Service is a compute service which actually runs your website. In this example a simple hello world Docker container.
The Serverless Network Endpoint Group is the connection between your Load Balancer and your Cloud Run service. It allows requests to the load balancer, to be routed to the serverless app backend.
The HTTP/S Load Balancer listens to incoming traffic on an IP address and directs this to a backend service. A terraform module is used which creates several resources, such as:
- Global Address
- HTTP Proxy
- Forwarding rules
- Managed SSL certificate
- Serverless NEG as backend
In this case the load balancer doesn't act traditionally, where the load is balanced between a group of backend services, rather, all load is forwarded to the configured backend.
The IAM Policy defines which members will be granted the role iap.httpsResourceAccessor
. With it, you
manage who will have access to your website and who doesn't.
The IAM Policy for Identity-Aware Proxy WebBackendService configures the Identity-Aware Proxy, and binds the
IAM Policy you defined to the Load Balancer. From here you can select the HTTPS resource, click on the info panel and
add or remove members from the IAP-secured Web App User
role.
This is not an official tutorial and can go out of date. Please read the documentation for more up-to-date info.