Deploy an API gateway with Kubernetes
You've developed an API server, deployed it to your Kuberenetes cluster, and now you want to make it available online in a way that lets you:
- Use Kubernetes-native resources and networking
- Consistently apply security and traffic management policy in one place
- View observability data in one place
What you'll learn
In this guide, you'll learn how to implement ngrok as an API gateway, using the ngrok Kubernetes Operator.
What you'll need
kubectl
and Helm installed on your local workstation.- A Kubernetes cluster—see our [local cluster options](/docs/k8s/guides/local-cluster] doc for some options.
- A reserved domain, which you can get in the ngrok
dashboard or with the ngrok
API.
- You can choose from an ngrok subdomain or bring your own custom branded
domain, like
https://api.example.com
. - We'll refer to this domain as
<YOUR_NGROK_DOMAIN>
.
- You can choose from an ngrok subdomain or bring your own custom branded
domain, like
- An Auth0 account (free is fine) for creating an API and generating JWTs.
Deploy demo API services
If you want to quickly wire up a POC using ngrok, we recommend using our demo API service, which responds with details about your requests.
You can also bring your own services to this guide—in that case, make sure to change service names, target ports, and namespaces accordingly in the resources you'll define in the next few steps.
The following manifests create two different but similar deployments that we will direct traffic to. Using the container environment variables section, they are configured slightly differently to demonstrate that requests are actually being routed to different destinations.
- Deployment+service: abc
- Deployment+service: xyz
Loading…
Loading…
Apply these with kubectl apply -f ...
into your cluster.
1. Deploy the ngrok Kubernetes Operator
You can check out our installation doc for more details and options on this process, but as a quickstart, add the ngrok Kubernetes Operator repo to Helm.
Loading…
Set up a few environment variables to help apply the ngrok Kubernetes Operator.
NGROK_AUTHTOKEN
: An ngrok authtoken—you can use either your default authtoken or create a new authtoken for this deployment.NGROK_API_KEY
: An API key created in the ngrok dashboard to associate with your deployment.
Loading…
Install the ngrok Kubernetes Operator into a new ngrok-operator
namespace.
Loading…
2. Create internal agent endpoints
These endpoints are private to your account and can only receive traffic
forwarded with the forward-internal
Traffic Policy action, which means they're
never publicly accessible.
The ngrok Kubernetes Operator comes with an AgentEndpoint
CRD that helps you
quickly map specific upstream services to ngrok endponits.
If you're bringing your own API services instead of using the demo API, you'll
need to change the url
and upstream.url
fields based on your architecture.
- abc
- xyz
Loading…
Loading…
Apply these with kubectl apply -f ...
to your cluster.
3. Create a cloud endpoint
Cloud endpoints are persistent, always-on endpoints that you can manage with the ngrok dashboard or API.
You centrally control your traffic management and security policy on your cloud endpoint, which operates as the "front door" of your API gateway, then forward traffic your API services in your cluster. That's much easier than synchronizing policies using cloud-specific tools, since they're all configured and managed in different ways.
- Dashboard
- API
Hop over to the Endpoints section of the ngrok dashboard and select + New.
Leave the Binding value Public, then enter the domain name you reserved earlier. Select Create Cloud Endpoint.
The ngrok
CLI provides a helpful wrapper around the ngrok
API, which you can use to create a cloud endpoint and
apply a file containing Traffic Policy rules.
Because every cloud endpoint must contain a Traffic Policy rule, create a
new file named policy.yaml
on your local workstation with the following
YAML, which is temporary until you add proper routing.
Loading…
Create a cloud endpoint on <YOUR_NGROK_DOMAIN>
, passing your
policy.yaml
file as an option.
Loading…
You'll get a 201
response—save the value of id
, as you'll need it
again later to continue configuring the Traffic Policy applied to your
cloud endpoint. We'll refer to it as <CLOUD_ENDPOINT_ID>
.
4. Route traffic to your services
Your front door is ready, but it currently doesn't have any logic for routing traffic to your API services.
Enter our Traffic Policy system, which lets you filter traffic based on its properties and take action as it passes through ngrok's global network. Two important concepts of Traffic Policy to note:
- Phases are the distinct points in the lifecycle of
a request where you can filter and take action. For this use case, we're using
on_http_request
, which activates when ngrok receives an HTTP request over an established connection. - Expressions define when to run
your actions. They're written in Common Expression
Language, and must evaluate to
true
to run the corresponding action.
The rules below:
- Filter for requests arriving on only
https://<YOUR_NGROK_DOMAIN>/abc
and forward them to your internal agent endpoint athttps://abc.internal
. - Filter for requests arriving on only
https://<YOUR_NGROK_DOMAIN>/xyz
and forward them to your internal agent endpoint inhttps://xyz.internal
.
You can also route by other properties, like subdomains and headers.
- Dashboard
- API
Copy and paste the rules below into your cloud endpoint's Traffic Policy editor
in the dashboard. If you're bringing your own API services instead of using the
demo API, you'll need to change /abc
and /xyz
to match your services' paths
and the url
for your internal agent endpoints.
Loading…
Hit Save to lock in the new policy.
Update your existing policy.yaml
file with the YAML below. If you're
bringing your own API services instead of using the demo API, you'll need to
change /abc
and /xyz
to match your services' paths and the url
for
your internal agent endpoints.
Loading…
Update your cloud endpoint.
Loading…
You can now curl
different paths to see your requests routed to the
appropriate upstream API service.
Loading…
You should get a response like:
Loading…
And when you run curl https://<YOUR_NGROK_DOMAIN>/xyz
?
Loading…
5. Add traffic management policies
Your API gateway routes traffic, but doesn't yet do essential work of an API gateway: offload all the non-functional requirements away from your services.
One great feature of ngrok's building blocks of endpoints and Traffic Policy rules is that they're composable—you can reuse them, chain them, and apply them at multiple stages in the lifecycle of an API request.
With the shape you've already created, you can centrally manage certain policies, like authentication, on your cloud endpoint, then compose additional rules onto specific services.
Validate JWTs on all APIs and requests
API authentication is too important not to apply consistently across all your
APIs and requests. That's where the always-on, front door to all your routes
quality of a cloud endpoint comes in handy—you can apply the jwt-validation
action once for dependable AuthN
no matter how many services you end up deploying behind your multicloud API
gateway.
ngrok's JWT validation action helps you:
- Give your end users many ways to access your APIs.
- Ensure only requests containing the correct access token, specified by an
Authorization: Bearer ...
header, can access any of your APIs. - Add claims to tokens for authorization and fine-grained access control where a
specific token may only have access to a certain API (
service_access: abc
) or apply RBAC (features: read
). - Use a single credential for end users who need to access multiple upstream services.
- Offload all this logic from your API services and run it in ngrok's network.
You can use any OAuth provider for JWT validation, but but let's quickly cover the process with Auth0.
- Log in to your Auth0 tenant dashboard.
- Select Applications > APIs, then + Create API.
- Name your API whatever you'd like.
- Replace the value of the Identifier field with
<YOUR_NGROK_DOMAIN>
. - Leave the default values for JSON Web Token (JWT) Profile and JSON Web Token Signing Algorithm.
- Select Create.
- Navigate to your application and select the Test tab, where you can find a signed, fully functional JWT and examples of how to programmatically generate more.
The rule below builds on top of the previous cloud endpoint policy to:
- Reject requests missing a token with a
401 Unauthorized
error. - Reject requests with an invalid token with a
403 Forbidden
error. - Forward requests with a valid token to one of your internal agent endpoints based on the pathname.
You'll need to change the variables accordingly—if you're not sure where to find this information, we have a full integration guide with more details.
Loading…
Apply in either the dashboard or ngrok API.
Rate limit specific API services
Let's say one of your services (like abc
, if you're following along with
the demo service), needs additional protection from unintentional misuse and
malicious attacks.
The rate-limit
Traffic Policy
action allows you to reject requests
with a 429
error code once a user or group have exceeded your customizable
threshold, and the AgentEndpoint
CRD allows you to define a Traffic Policy for
just that endpoint.
The rule below builds on top of your AgentEndpoint
CR from earlier to:
- Allow up to
10
requests per IP in a60s
window. - Reject requests that exceed the rate limiting capacity with a
429
error response. - Forward all other requests to the upstream API service at
http://abc-app-service.default:4000
.
Loading…
Apply the updated manifest with kubectl apply -f ...
.
Ready to test your rate limit in action? Run the below command after replacing
<YOUR_NGROK_DOMAIN>
and the path, if relevant.
Loading…
You'll see a few normal responses until you hit the rate limit, and then you'll
see 429
errors. Run the same command on the /xyz
path and you won't see the
same errors, since you've applied this policy only to the
https://abc-cloud-a.internal
agent endpoint.
If you want all your APIs to have a consistent rate limiting strategy, you can
move the rule to your cloud endpoint above the jwt-validation
action.
What's next?
You've now brought your APIs online with ngrok's API gateway and Kubernetes, which also automatically gives you features like DDoS protection and global load balancing. Plus, you've added global AuthN with JWTs and explored composing Traffic Policy rules on multiple endpoints.
There's still a lot more to explore with ngrok's API gateway. Next up, we recommend you:
- Check out your Traffic Inspector (documentation) to observe, modify, and replay requests across your API gateway.
- Explore other opportunities to manage and take action on API traffic in our Traffic Policy documentation.
- Extend your API gateway into multicloud
by deploying the Kubernetes Operator on a second (or third) cluster and
provisioning an
AgentEndpoint
resource.