Amazon API Gateway Construct Library

---

> AWS CDK v1 has reached End-of-Support on 2023-06-01. > This package is no longer being updated, and users should migrate to AWS CDK v2. > > For more information on how to migrate, see the Migrating to AWS CDK v2 guide.

---

Amazon API Gateway is a fully managed service that makes it easy for developers to publish, maintain, monitor, and secure APIs at any scale. Create an API to access data, business logic, or functionality from your back-end services, such as applications running on Amazon Elastic Compute Cloud (Amazon EC2), code running on AWS Lambda, or any web application.

Table of Contents

• Defining APIs

  • Breaking up Methods and Resources across Stacks

• AWS Lambda-backed APIs

• AWS StepFunctions backed APIs

• Integration Targets

• Usage Plan & API Keys

• Working with models

• Default Integration and Method Options

• Proxy Routes

• Authorizers

  • IAM-based authorizer
  • Lambda-based token authorizer
  • Lambda-based request authorizer
  • Cognito User Pools authorizer

• Mutual TLS

• Deployments

  • Deep dive: Invalidation of deployments

• Custom Domains

• Access Logging

• Cross Origin Resource Sharing (CORS)

• Endpoint Configuration

• Private Integrations

• Gateway Response

• OpenAPI Definition

  • Endpoint configuration

• Metrics

• APIGateway v2

Defining APIs

APIs are defined as a hierarchy of resources and methods. addResource and addMethod can be used to build this hierarchy. The root resource is api.root.

For example, the following code defines an API that includes the following HTTP endpoints: ANY /, GET /books, POST /books, GET /books/{book_id}, DELETE /books/{book_id}.

api = apigateway.RestApi(self, "books-api")

api.root.add_method("ANY")

books = api.root.add_resource("books")
books.add_method("GET")
books.add_method("POST")

book = books.add_resource("{book_id}")
book.add_method("GET")
book.add_method("DELETE")

AWS Lambda-backed APIs

A very common practice is to use Amazon API Gateway with AWS Lambda as the backend integration. The LambdaRestApi construct makes it easy:

The following code defines a REST API that routes all requests to the specified AWS Lambda function:

# backend: lambda.Function

apigateway.LambdaRestApi(self, "myapi",
    handler=backend
)

You can also supply proxy: false, in which case you will have to explicitly define the API model:

# backend: lambda.Function

api = apigateway.LambdaRestApi(self, "myapi",
    handler=backend,
    proxy=False
)

items = api.root.add_resource("items")
items.add_method("GET") # GET /items
items.add_method("POST") # POST /items

item = items.add_resource("{item}")
item.add_method("GET") # GET /items/{item}

# the default integration for methods is "handler", but one can
# customize this behavior per method or even a sub path.
item.add_method("DELETE", apigateway.HttpIntegration("http://amazon.com"))

AWS StepFunctions backed APIs

You can use Amazon API Gateway with AWS Step Functions as the backend integration, specifically Synchronous Express Workflows.

The StepFunctionsRestApi only supports integration with Synchronous Express state machine. The StepFunctionsRestApi construct makes this easy by setting up input, output and error mapping.

The construct sets up an API endpoint and maps the ANY HTTP method and any calls to the API endpoint starts an express workflow execution for the underlying state machine.

Invoking the endpoint with any HTTP method (GET, POST, PUT, DELETE, ...) in the example below will send the request to the state machine as a new execution. On success, an HTTP code 200 is returned with the execution output as the Response Body.

If the execution fails, an HTTP 500 response is returned with the error and cause from the execution output as the Response Body. If the request is invalid (ex. bad execution input) HTTP code 400 is returned.

The response from the invocation contains only the output field from the StartSyncExecution API. In case of failures, the fields error and cause are returned as part of the response. Other metadata such as billing details, AWS account ID and resource ARNs are not returned in the API response.

By default, a prod stage is provisioned.

In order to reduce the payload size sent to AWS Step Functions, headers are not forwarded to the Step Functions execution input. It is possible to choose whether headers, requestContext, path, querystring, and authorizer are included or not. By default, headers are excluded in all requests.

More details about AWS Step Functions payload limit can be found at https://docs.aws.amazon.com/step-functions/latest/dg/limits-overview.html#service-limits-task-executions.

The following code defines a REST API that routes all requests to the specified AWS StepFunctions state machine:

state_machine_definition = stepfunctions.Pass(self, "PassState")

state_machine = stepfunctions.StateMachine(self, "StateMachine",
    definition=state_machine_definition,
    state_machine_type=stepfunctions.StateMachineType.EXPRESS
)

apigateway.StepFunctionsRestApi(self, "StepFunctionsRestApi",
    deploy=True,
    state_machine=state_machine
)

When the REST API endpoint configuration above is invoked using POST, as follows -

curl -X POST -d '{ "customerId": 1 }' https://example.com/

AWS Step Functions will receive the request body in its input as follows:

{
  "body": {
    "customerId": 1
  },
  "path": "/",
  "querystring": {}
}

When the endpoint is invoked at path '/users/5' using the HTTP GET method as below:

curl -X GET https://example.com/users/5?foo=bar

AWS Step Functions will receive the following execution input:

{
  "body": {},
  "path": {
     "users": "5"
  },
  "querystring": {
    "foo": "bar"
  }
}

Additional information around the request such as the request context, authorizer context, and headers can be included as part of the input forwarded to the state machine. The following example enables headers to be included in the input but not query string.

apigateway.StepFunctionsRestApi(self, "StepFunctionsRestApi",
    state_machine=machine,
    headers=True,
    path=False,
    querystring=False,
    authorizer=False,
    request_context=apigateway.RequestContext(
        caller=True,
        user=True
    )
)

In such a case, when the endpoint is invoked as below:

curl -X GET https://example.com/

AWS Step Functions will receive the following execution input:

{
  "headers": {
    "Accept": "...",
    "CloudFront-Forwarded-Proto": "...",
  },
  "requestContext": {
     "accountId": "...",
     "apiKey": "...",
  },
  "body": {}
}

Breaking up Methods and Resources across Stacks

It is fairly common for REST APIs with a large number of Resources and Methods to hit the CloudFormation limit of 500 resources per stack.

To help with this, Resources and Methods for the same REST API can be re-organized across multiple stacks. A common way to do this is to have a stack per Resource or groups of Resources, but this is not the only possible way. The following example uses sets up two Resources '/pets' and '/books' in separate stacks using nested stacks:

from aws_cdk.aws_apigateway import IntegrationResponse, MethodResponse, IntegrationResponse, MethodResponse
from aws_cdk.core import App, CfnOutput, NestedStack, NestedStackProps, Stack
from constructs import Construct
from aws_cdk.aws_apigateway import Deployment, Method, MockIntegration, PassthroughBehavior, RestApi, Stage

#
# This file showcases how to split up a RestApi's Resources and Methods across nested stacks.
#
# The root stack 'RootStack' first defines a RestApi.
# Two nested stacks BooksStack and PetsStack, create corresponding Resources '/books' and '/pets'.
# They are then deployed to a 'prod' Stage via a third nested stack - DeployStack.
#
# To verify this worked, go to the APIGateway
#

class RootStack(Stack):
    def __init__(self, scope):
        super().__init__(scope, "integ-restapi-import-RootStack")

        rest_api = RestApi(self, "RestApi",
            deploy=False
        )
        rest_api.root.add_method("ANY")

        pets_stack = PetsStack(self,
            rest_api_id=rest_api.rest_api_id,
            root_resource_id=rest_api.rest_api_root_resource_id
        )
        books_stack = BooksStack(self,
            rest_api_id=rest_api.rest_api_id,
            root_resource_id=rest_api.rest_api_root_resource_id
        )
        DeployStack(self,
            rest_api_id=rest_api.rest_api_id,
            methods=pets_stack.methods.concat(books_stack.methods)
        )

        CfnOutput(self, "PetsURL",
            value=f"https://{restApi.restApiId}.execute-api.{this.region}.amazonaws.com/prod/pets"
        )

        CfnOutput(self, "BooksURL",
            value=f"https://{restApi.restApiId}.execute-api.{this.region}.amazonaws.com/prod/books"
        )

class PetsStack(NestedStack):

    def __init__(self, scope, *, restApiId, rootResourceId, parameters=None, timeout=None, notificationArns=None, removalPolicy=None):
        super().__init__(scope, "integ-restapi-import-PetsStack", restApiId=restApiId, rootResourceId=rootResourceId, parameters=parameters, timeout=timeout, notificationArns=notificationArns, removalPolicy=removalPolicy)

        api = RestApi.from_rest_api_attributes(self, "RestApi",
            rest_api_id=rest_api_id,
            root_resource_id=root_resource_id
        )

        method = api.root.add_resource("pets").add_method("GET", MockIntegration(
            integration_responses=[IntegrationResponse(
                status_code="200"
            )],
            passthrough_behavior=PassthroughBehavior.NEVER,
            request_templates={
                "application/json": "{ \"statusCode\": 200 }"
            }
        ),
            method_responses=[MethodResponse(status_code="200")]
        )

        self.methods.push(method)

class BooksStack(NestedStack):

    def __init__(self, scope, *, restApiId, rootResourceId, parameters=None, timeout=None, notificationArns=None, removalPolicy=None):
        super().__init__(scope, "integ-restapi-import-BooksStack", restApiId=restApiId, rootResourceId=rootResourceId, parameters=parameters, timeout=timeout, notificationArns=notificationArns, removalPolicy=removalPolicy)

        api = RestApi.from_rest_api_attributes(self, "RestApi",
            rest_api_id=rest_api_id,
            root_resource_id=root_resource_id
        )

        method = api.root.add_resource("books").add_method("GET", MockIntegration(
            integration_responses=[IntegrationResponse(
                status_code="200"
            )],
            passthrough_behavior=PassthroughBehavior.NEVER,
            request_templates={
                "application/json": "{ \"statusCode\": 200 }"
            }
        ),
            method_responses=[MethodResponse(status_code="200")]
        )

        self.methods.push(method)

class DeployStack(NestedStack):
    def __init__(self, scope, *, restApiId, methods=None, parameters=None, timeout=None, notificationArns=None, removalPolicy=None):
        super().__init__(scope, "integ-restapi-import-DeployStack", restApiId=restApiId, methods=methods, parameters=parameters, timeout=timeout, notificationArns=notificationArns, removalPolicy=removalPolicy)

        deployment = Deployment(self, "Deployment",
            api=RestApi.from_rest_api_id(self, "RestApi", rest_api_id)
        )
        if methods:
            for method in methods:
                deployment.node.add_dependency(method)
        Stage(self, "Stage", deployment=deployment)

RootStack(App())

Integration Targets

Methods are associated with backend integrations, which are invoked when this method is called. API Gateway supports the following integrations:

• MockIntegration - can be used to test APIs. This is the default integration if one is not specified.
• LambdaIntegration - can be used to invoke an AWS Lambda function.
• AwsIntegration - can be used to invoke arbitrary AWS service APIs.
• HttpIntegration - can be used to invoke HTTP endpoints.

The following example shows how to integrate the GET /book/{book_id} method to an AWS Lambda function:

# get_book_handler: lambda.Function
# book: apigateway.Resource


get_book_integration = apigateway.LambdaIntegration(get_book_handler)
book.add_method("GET", get_book_integration)

Integration options can be optionally be specified:

# get_book_handler: lambda.Function
# get_book_integration: apigateway.LambdaIntegration


get_book_integration = apigateway.LambdaIntegration(get_book_handler,
    content_handling=apigateway.ContentHandling.CONVERT_TO_TEXT,  # convert to base64
    credentials_passthrough=True
)

Method options can optionally be specified when adding methods:

# book: apigateway.Resource
# get_book_integration: apigateway.LambdaIntegration


book.add_method("GET", get_book_integration,
    authorization_type=apigateway.AuthorizationType.IAM,
    api_key_required=True
)

It is possible to also integrate with AWS services in a different region. The following code integrates with Amazon SQS in the eu-west-1 region.

get_message_integration = apigateway.AwsIntegration(
    service="sqs",
    path="queueName",
    region="eu-west-1"
)

Usage Plan & API Keys

A usage plan specifies who can access one or more deployed API stages and methods, and the rate at which they can be accessed. The plan uses API keys to identify API clients and meters access to the associated API stages for each key. Usage plans also allow configuring throttling limits and quota limits that are enforced on individual client API keys.

The following example shows how to create and asscociate a usage plan and an API key:

# integration: apigateway.LambdaIntegration


api = apigateway.RestApi(self, "hello-api")

v1 = api.root.add_resource("v1")
echo = v1.add_resource("echo")
echo_method = echo.add_method("GET", integration, api_key_required=True)

plan = api.add_usage_plan("UsagePlan",
    name="Easy",
    throttle=apigateway.ThrottleSettings(
        rate_limit=10,
        burst_limit=2
    )
)

key = api.add_api_key("ApiKey")
plan.add_api_key(key)

To associate a plan to a given RestAPI stage:

# plan: apigateway.UsagePlan
# api: apigateway.RestApi
# echo_method: apigateway.Method


plan.add_api_stage(
    stage=api.deployment_stage,
    throttle=[apigateway.ThrottlingPerMethod(
        method=echo_method,
        throttle=apigateway.ThrottleSettings(
            rate_limit=10,
            burst_limit=2
        )
    )
    ]
)

Existing usage plans can be imported into a CDK app using its id.

imported_usage_plan = apigateway.UsagePlan.from_usage_plan_id(self, "imported-usage-plan", "")

The name and value of the API Key can be specified at creation; if not provided, a name and value will be automatically generated by API Gateway.

# api: apigateway.RestApi

key = api.add_api_key("ApiKey",
    api_key_name="myApiKey1",
    value="MyApiKeyThatIsAtLeast20Characters"
)

Existing API keys can also be imported into a CDK app using its id.

imported_key = apigateway.ApiKey.from_api_key_id(self, "imported-key", "")

The "grant" methods can be used to give prepackaged sets of permissions to other resources. The following code provides read permission to an API key.

# imported_key: apigateway.ApiKey
# lambda_fn: lambda.Function

imported_key.grant_read(lambda_fn)

⚠️ Multiple API Keys

It is possible to specify multiple API keys for a given Usage Plan, by calling usagePlan.addApiKey().

When using multiple API keys, a past bug of the CDK prevents API key associations to a Usage Plan to be deleted. If the CDK app had the feature flag - @aws-cdk/aws-apigateway:usagePlanKeyOrderInsensitiveId - enabled when the API keys were created, then the app will not be affected by this bug.

If this is not the case, you will need to ensure that the CloudFormation logical ids of the API keys that are not being deleted remain unchanged. Make note of the logical ids of these API keys before removing any, and set it as part of the addApiKey() method:

# usageplan: apigateway.UsagePlan
# api_key: apigateway.ApiKey


usageplan.add_api_key(api_key,
    override_logical_id="..."
)

Rate Limited API Key

In scenarios where you need to create a single api key and configure rate limiting for it, you can use RateLimitedApiKey. This construct lets you specify rate limiting properties which should be applied only to the api key being created. The API key created has the specified rate limits, such as quota and throttles, applied.

The following example shows how to use a rate limited api key :

# api: apigateway.RestApi


key = apigateway.RateLimitedApiKey(self, "rate-limited-api-key",
    customer_id="hello-customer",
    resources=[api],
    quota=apigateway.QuotaSettings(
        limit=10000,
        period=apigateway.Period.MONTH
    )
)

Working with models

When you work with Lambda integrations that are not Proxy integrations, you have to define your models and mappings for the request, response, and integration.

hello = lambda_.Function(self, "hello",
    runtime=lambda_.Runtime.NODEJS_14_X,
    handler="hello.handler",
    code=lambda_.Code.from_asset("lambda")
)

api = apigateway.RestApi(self, "hello-api")
resource = api.root.add_resource("v1")

You can define more parameters on the integration to tune the behavior of API Gateway

# hello: lambda.Function


integration = apigateway.LambdaIntegration(hello,
    proxy=False,
    request_parameters={
        # You can define mapping parameters from your method to your integration
        # - Destination parameters (the key) are the integration parameters (used in mappings)
        # - Source parameters (the value) are the source request parameters or expressions
        # @see: https://docs.aws.amazon.com/apigateway/latest/developerguide/request-response-data-mappings.html
        "integration.request.querystring.who": "method.request.querystring.who"
    },
    allow_test_invoke=True,
    request_templates={
        # You can define a mapping that will build a payload for your integration, based
        #  on the integration parameters that you have specified
        # Check: https://docs.aws.amazon.com/apigateway/latest/developerguide/api-gateway-mapping-template-reference.html
        "application/json": JSON.stringify({"action": "sayHello", "poll_id": "$util.escapeJavaScript($input.params('who'))"})
    },
    # This parameter defines the behavior of the engine is no suitable response template is found
    passthrough_behavior=apigateway.PassthroughBehavior.NEVER,
    integration_responses=[apigateway.IntegrationResponse(
        # Successful response from the Lambda function, no filter defined
        #  - the selectionPattern filter only tests the error message
        # We will set the response status code to 200
        status_code="200",
        response_templates={
            # This template takes the "message" result from the Lambda function, and embeds it in a JSON response
            # Check https://docs.aws.amazon.com/apigateway/latest/developerguide/api-gateway-mapping-template-reference.html
            "application/json": JSON.stringify({"state": "ok", "greeting": "$util.escapeJavaScript($input.body)"})
        },
        response_parameters={
            # We can map response parameters
            # - Destination parameters (the key) are the response parameters (used in mappings)
            # - Source parameters (the value) are the integration response parameters or expressions
            "method.response.header.Content-Type": "'application/json'",
            "method.response.header.Access-Control-Allow-Origin": "'*'",
            "method.response.header.Access-Control-Allow-Credentials": "'true'"
        }
    ), apigateway.IntegrationResponse(
        # For errors, we check if the error message is not empty, get the error data
        selection_pattern="(\n|.)+",
        # We will set the response status code to 200
        status_code="400",
        response_templates={
            "application/json": JSON.stringify({"state": "error", "message": "$util.escapeJavaScript($input.path('$.errorMessage'))"})
        },
        response_parameters={
            "method.response.header.Content-Type": "'application/json'",
            "method.response.header.Access-Control-Allow-Origin": "'*'",
            "method.response.header.Access-Control-Allow-Credentials": "'true'"
        }
    )
    ]
)

You can define models for your responses (and requests)

# api: apigateway.RestApi


# We define the JSON Schema for the transformed valid response
response_model = api.add_model("ResponseModel",
    content_type="application/json",
    model_name="ResponseModel",
    schema=apigateway.JsonSchema(
        schema=apigateway.JsonSchemaVersion.DRAFT4,
        title="pollResponse",
        type=apigateway.JsonSchemaType.OBJECT,
        properties={
            "state": apigateway.JsonSchema(type=apigateway.JsonSchemaType.STRING),
            "greeting": apigateway.JsonSchema(type=apigateway.JsonSchemaType.STRING)
        }
    )
)

# We define the JSON Schema for the transformed error response
error_response_model = api.add_model("ErrorResponseModel",
    content_type="application/json",
    model_name="ErrorResponseModel",
    schema=apigateway.JsonSchema(
        schema=apigateway.JsonSchemaVersion.DRAFT4,
        title="errorResponse",
        type=apigateway.JsonSchemaType.OBJECT,
        properties={
            "state": apigateway.JsonSchema(type=apigateway.JsonSchemaType.STRING),
            "message": apigateway.JsonSchema(type=apigateway.JsonSchemaType.STRING)
        }
    )
)

And reference all on your method definition.

# integration: apigateway.LambdaIntegration
# resource: apigateway.Resource
# response_model: apigateway.Model
# error_response_model: apigateway.Model


resource.add_method("GET", integration,
    # We can mark the parameters as required
    request_parameters={
        "method.request.querystring.who": True
    },
    # we can set request validator options like below
    request_validator_options=apigateway.RequestValidatorOptions(
        request_validator_name="test-validator",
        validate_request_body=True,
        validate_request_parameters=False
    ),
    method_responses=[apigateway.MethodResponse(
        # Successful response from the integration
        status_code="200",
        # Define what parameters are allowed or not
        response_parameters={
            "method.response.header.Content-Type": True,
            "method.response.header.Access-Control-Allow-Origin": True,
            "method.response.header.Access-Control-Allow-Credentials": True
        },
        # Validate the schema on the response
        response_models={
            "application/json": response_model
        }
    ), apigateway.MethodResponse(
        # Same thing for the error responses
        status_code="400",
        response_parameters={
            "method.response.header.Content-Type": True,
            "method.response.header.Access-Control-Allow-Origin": True,
            "method.response.header.Access-Control-Allow-Credentials": True
        },
        response_models={
            "application/json": error_response_model
        }
    )
    ]
)

Specifying requestValidatorOptions automatically creates the RequestValidator construct with the given options. However, if you have your RequestValidator already initialized or imported, use the requestValidator option instead.

Default Integration and Method Options

The defaultIntegration and defaultMethodOptions properties can be used to configure a default integration at any resource level. These options will be used when defining method under this resource (recursively) with undefined integration or options.

> If not defined, the default integration is MockIntegration. See reference > documentation for default method options.

The following example defines the booksBackend integration as a default integration. This means that all API methods that do not explicitly define an integration will be routed to this AWS Lambda function.

# books_backend: apigateway.LambdaIntegration

api = apigateway.RestApi(self, "books",
    default_integration=books_backend
)

books = api.root.add_resource("books")
books.add_method("GET") # integrated with `booksBackend`
books.add_method("POST") # integrated with `booksBackend`

book = books.add_resource("{book_id}")
book.add_method("GET")

A Method can be configured with authorization scopes. Authorization scopes are used in conjunction with an authorizer that uses Amazon Cognito user pools. Read more about authorization scopes here.

Authorization scopes for a Method can be configured using the authorizationScopes property as shown below -

# books: apigateway.Resource


books.add_method("GET", apigateway.HttpIntegration("http://amazon.com"),
    authorization_type=apigateway.AuthorizationType.COGNITO,
    authorization_scopes=["Scope1", "Scope2"]
)

Proxy Routes

The addProxy method can be used to install a greedy {proxy+} resource on a path. By default, this also installs an "ANY" method:

# resource: apigateway.Resource
# handler: lambda.Function

proxy = resource.add_proxy(
    default_integration=apigateway.LambdaIntegration(handler),

    # "false" will require explicitly adding methods on the `proxy` resource
    any_method=True
)

Authorizers

API Gateway supports several different authorization types that can be used for controlling access to your REST APIs.

IAM-based authorizer

The following CDK code provides 'execute-api' permission to an IAM user, via IAM policies, for the 'GET' method on the books resource:

# books: apigateway.Resource
# iam_user: iam.User


get_books = books.add_method("GET", apigateway.HttpIntegration("http://amazon.com"),
    authorization_type=apigateway.AuthorizationType.IAM
)

iam_user.attach_inline_policy(iam.Policy(self, "AllowBooks",
    statements=[
        iam.PolicyStatement(
            actions=["execute-api:Invoke"],
            effect=iam.Effect.ALLOW,
            resources=[get_books.method_arn]
        )
    ]
))

Lambda-based token authorizer

API Gateway also allows lambda functions to be used as authorizers.

This module provides support for token-based Lambda authorizers. When a client makes a request to an API's methods configured with such an authorizer, API Gateway calls the Lambda authorizer, which takes the caller's identity as input and returns an IAM policy as output. A token-based Lambda authorizer (also called a token authorizer) receives the caller's identity in a bearer token, such as a JSON Web Token (JWT) or an OAuth token.

API Gateway interacts with the authorizer Lambda function handler by passing input and expecting the output in a specific format. The event object that the handler is called with contains the authorizationToken and the methodArn from the request to the API Gateway endpoint. The handler is expected to return the principalId (i.e. the client identifier) and a policyDocument stating what the client is authorizer to perform. See here for a detailed specification on inputs and outputs of the Lambda handler.

The following code attaches a token-based Lambda authorizer to the 'GET' Method of the Book resource:

# auth_fn: lambda.Function
# books: apigateway.Resource


auth = apigateway.TokenAuthorizer(self, "booksAuthorizer",
    handler=auth_fn
)

books.add_method("GET", apigateway.HttpIntegration("http://amazon.com"),
    authorizer=auth
)

A full working example is shown below.

!cdk-integ pragma:ignore-assets

from aws_cdk.aws_apigateway import IntegrationResponse, MethodResponse
import path as path
import aws_cdk.aws_lambda as lambda_
from aws_cdk.core import App, Stack
from aws_cdk.aws_apigateway import MockIntegration, PassthroughBehavior, RestApi, TokenAuthorizer

#
# Stack verification steps:
# * `curl -s -o /dev/null -w "%{http_code}" ` should return 401
# * `curl -s -o /dev/null -w "%{http_code}" -H 'Authorization: deny' ` should return 403
# * `curl -s -o /dev/null -w "%{http_code}" -H 'Authorization: allow' ` should return 200
#

app = App()
stack = Stack(app, "TokenAuthorizerInteg")

authorizer_fn = lambda_.Function(stack, "MyAuthorizerFunction",
    runtime=lambda_.Runtime.NODEJS_14_X,
    handler="index.handler",
    code=lambda_.AssetCode.from_asset(path.join(__dirname, "integ.token-authorizer.handler"))
)

restapi = RestApi(stack, "MyRestApi")

authorizer = TokenAuthorizer(stack, "MyAuthorizer",
    handler=authorizer_fn
)

restapi.root.add_method("ANY", MockIntegration(
    integration_responses=[IntegrationResponse(status_code="200")
    ],
    passthrough_behavior=PassthroughBehavior.NEVER,
    request_templates={
        "application/json": "{ \"statusCode\": 200 }"
    }
),
    method_responses=[MethodResponse(status_code="200")
    ],
    authorizer=authorizer
)

By default, the TokenAuthorizer looks for the authorization token in the request header with the key 'Authorization'. This can, however, be modified by changing the identitySource property.

Authorizers can also be passed via the defaultMethodOptions property within the RestApi construct or the Method construct. Unless explicitly overridden, the specified defaults will be applied across all Methods across the RestApi or across all Resources, depending on where the defaults were specified.

Lambda-based request authorizer

This module provides support for request-based Lambda authorizers. When a client makes a request to an API's methods configured with such an authorizer, API Gateway calls the Lambda authorizer, which takes specified parts of the request, known as identity sources, as input and returns an IAM policy as output. A request-based Lambda authorizer (also called a request authorizer) receives the identity sources in a series of values pulled from the request, from the headers, stage variables, query strings, and the context.

API Gateway interacts with the authorizer Lambda function handler by passing input and expecting the output in a specific format. The event object that the handler is called with contains the body of the request and the methodArn from the request to the API Gateway endpoint. The handler is expected to return the principalId (i.e. the client identifier) and a policyDocument stating what the client is authorizer to perform. See here for a detailed specification on inputs and outputs of the Lambda handler.

The following code attaches a request-based Lambda authorizer to the 'GET' Method of the Book resource:

# auth_fn: lambda.Function
# books: apigateway.Resource


auth = apigateway.RequestAuthorizer(self, "booksAuthorizer",
    handler=auth_fn,
    identity_sources=[apigateway.IdentitySource.header("Authorization")]
)

books.add_method("GET", apigateway.HttpIntegration("http://amazon.com"),
    authorizer=auth
)

A full working example is shown below.

!cdk-integ pragma:ignore-assets

from aws_cdk.aws_apigateway import IntegrationResponse, MethodResponse
import path as path
import aws_cdk.aws_lambda as lambda_
from aws_cdk.core import App, Stack
from aws_cdk.aws_apigateway import MockIntegration, PassthroughBehavior, RestApi
from aws_cdk.aws_apigateway import RequestAuthorizer
from aws_cdk.aws_apigateway import IdentitySource

# Against the RestApi endpoint from the stack output, run
# `curl -s -o /dev/null -w "%{http_code}" ` should return 401
# `curl -s -o /dev/null -w "%{http_code}" -H 'Authorization: deny' ?allow=yes` should return 403
# `curl -s -o /dev/null -w "%{http_code}" -H 'Authorization: allow' ?allow=yes` should return 200

app = App()
stack = Stack(app, "RequestAuthorizerInteg")

authorizer_fn = lambda_.Function(stack, "MyAuthorizerFunction",
    runtime=lambda_.Runtime.NODEJS_14_X,
    handler="index.handler",
    code=lambda_.AssetCode.from_asset(path.join(__dirname, "integ.request-authorizer.handler"))
)

restapi = RestApi(stack, "MyRestApi")

authorizer = RequestAuthorizer(stack, "MyAuthorizer",
    handler=authorizer_fn,
    identity_sources=[IdentitySource.header("Authorization"), IdentitySource.query_string("allow")]
)

restapi.root.add_method("ANY", MockIntegration(
    integration_responses=[IntegrationResponse(status_code="200")
    ],
    passthrough_behavior=PassthroughBehavior.NEVER,
    request_templates={
        "application/json": "{ \"statusCode\": 200 }"
    }
),
    method_responses=[MethodResponse(status_code="200")
    ],
    authorizer=authorizer
)

By default, the RequestAuthorizer does not pass any kind of information from the request. This can, however, be modified by changing the identitySource property, and is required when specifying a value for caching.

Authorizers can also be passed via the defaultMethodOptions property within the RestApi construct or the Method construct. Unless explicitly overridden, the specified defaults will be applied across all Methods across the RestApi or across all Resources, depending on where the defaults were specified.

Cognito User Pools authorizer

API Gateway also allows Amazon Cognito user pools as authorizer

The following snippet configures a Cognito user pool as an authorizer:

# books: apigateway.Resource
user_pool = cognito.UserPool(self, "UserPool")

auth = apigateway.CognitoUserPoolsAuthorizer(self, "booksAuthorizer",
    cognito_user_pools=[user_pool]
)
books.add_method("GET", apigateway.HttpIntegration("http://amazon.com"),
    authorizer=auth,
    authorization_type=apigateway.AuthorizationType.COGNITO
)

Mutual TLS (mTLS)

Mutual TLS can be configured to limit access to your API based by using client certificates instead of (or as an extension of) using authorization headers.

# acm: Any


apigateway.DomainName(self, "domain-name",
    domain_name="example.com",
    certificate=acm.Certificate.from_certificate_arn(self, "cert", "arn:aws:acm:us-east-1:1111111:certificate/11-3336f1-44483d-adc7-9cd375c5169d"),
    mtls=apigateway.MTLSConfig(
        bucket=s3.Bucket(self, "bucket"),
        key="truststore.pem",
        version="version"
    )
)

Instructions for configuring your trust store can be found here.

Deployments

By default, the RestApi construct will automatically create an API Gateway Deployment and a "prod" Stage which represent the API configuration you defined in your CDK app. This means that when you deploy your app, your API will be have open access from the internet via the stage URL.

The URL of your API can be obtained from the attribute restApi.url, and is also exported as an Output from your stack, so it's printed when you cdk deploy your app:

$ cdk deploy
...
books.booksapiEndpointE230E8D5 = https://6lyktd4lpk.execute-api.us-east-1.amazonaws.com/prod/

To disable this behavior, you can set { deploy: false } when creating your API. This means that the API will not be deployed and a stage will not be created for it. You will need to manually define a apigateway.Deployment and apigateway.Stage resources.

Use the deployOptions property to customize the deployment options of your API.

The following example will configure API Gateway to emit logs and data traces to AWS CloudWatch for all API calls:

> By default, an IAM role will be created and associated with API Gateway to > allow it to write logs and metrics to AWS CloudWatch unless cloudWatchRole is > set to false.

api = apigateway.RestApi(self, "books",
    deploy_options=apigateway.StageOptions(
        logging_level=apigateway.MethodLoggingLevel.INFO,
        data_trace_enabled=True
    )
)

Deep dive: Invalidation of deployments

API Gateway deployments are an immutable snapshot of the API. This means that we want to automatically create a new deployment resource every time the API model defined in our CDK app changes.

In order to achieve that, the AWS CloudFormation logical ID of the AWS::ApiGateway::Deployment resource is dynamically calculated by hashing the API configuration (resources, methods). This means that when the configuration changes (i.e. a resource or method are added, configuration is changed), a new logical ID will be assigned to the deployment resource. This will cause CloudFormation to create a new deployment resource.

By default, old deployments are deleted. You can set retainDeployments: true to allow users revert the stage to an old deployment manually.

Custom Domains

To associate an API with a custom domain, use the domainName configuration when you define your API:

# acm_certificate_for_example_com: Any


api = apigateway.RestApi(self, "MyDomain",
    domain_name=apigateway.DomainNameOptions(
        domain_name="example.com",
        certificate=acm_certificate_for_example_com
    )
)

This will define a DomainName resource for you, along with a BasePathMapping from the root of the domain to the deployment stage of the API. This is a common set up.

To route domain traffic to an API Gateway API, use Amazon Route 53 to create an alias record. An alias record is a Route 53 extension to DNS. It's similar to a CNAME record, but you can create an alias record both for the root domain, such as example.com, and for subdomains, such as www.example.com. (You can create CNAME records only for subdomains.)

import aws_cdk.aws_route53 as route53
import aws_cdk.aws_route53_targets as targets

# api: apigateway.RestApi
# hosted_zone_for_example_com: Any


route53.ARecord(self, "CustomDomainAliasRecord",
    zone=hosted_zone_for_example_com,
    target=route53.RecordTarget.from_alias(targets.ApiGateway(api))
)

You can also define a DomainName resource directly in order to customize the default behavior:

# acm_certificate_for_example_com: Any


apigateway.DomainName(self, "custom-domain",
    domain_name="example.com",
    certificate=acm_certificate_for_example_com,
    endpoint_type=apigateway.EndpointType.EDGE,  # default is REGIONAL
    security_policy=apigateway.SecurityPolicy.TLS_1_2
)

Once you have a domain, you can map base paths of the domain to APIs. The following example will map the URL https://example.com/go-to-api1 to the api1 API and https://example.com/boom to the api2 API.

# domain: apigateway.DomainName
# api1: apigateway.RestApi
# api2: apigateway.RestApi


domain.add_base_path_mapping(api1, base_path="go-to-api1")
domain.add_base_path_mapping(api2, base_path="boom")

You can specify the API Stage to which this base path URL will map to. By default, this will be the deploymentStage of the RestApi.

# domain: apigateway.DomainName
# restapi: apigateway.RestApi


beta_deploy = apigateway.Deployment(self, "beta-deployment",
    api=restapi
)
beta_stage = apigateway.Stage(self, "beta-stage",
    deployment=beta_deploy
)
domain.add_base_path_mapping(restapi, base_path="api/beta", stage=beta_stage)

If you don't specify basePath, all URLs under this domain will be mapped to the API, and you won't be able to map another API to the same domain:

# domain: apigateway.DomainName
# api: apigateway.RestApi

domain.add_base_path_mapping(api)

This can also be achieved through the mapping configuration when defining the domain as demonstrated above.

If you wish to setup this domain with an Amazon Route53 alias, use the targets.ApiGatewayDomain:

# hosted_zone_for_example_com: Any
# domain_name: apigateway.DomainName

import aws_cdk.aws_route53 as route53
import aws_cdk.aws_route53_targets as targets


route53.ARecord(self, "CustomDomainAliasRecord",
    zone=hosted_zone_for_example_com,
    target=route53.RecordTarget.from_alias(targets.ApiGatewayDomain(domain_name))
)

Access Logging

Access logging creates logs every time an API method is accessed. Access logs can have information on who has accessed the API, how the caller accessed the API and what responses were generated. Access logs are configured on a Stage of the RestApi. Access logs can be expressed in a format of your choosing, and can contain any access details, with a minimum that it must include the 'requestId'. The list of variables that can be expressed in the access log can be found here. Read more at Setting Up CloudWatch API Logging in API Gateway

# production stage
prd_log_group = logs.LogGroup(self, "PrdLogs")
api = apigateway.RestApi(self, "books",
    deploy_options=apigateway.StageOptions(
        access_log_destination=apigateway.LogGroupLogDestination(prd_log_group),
        access_log_format=apigateway.AccessLogFormat.json_with_standard_fields()
    )
)
deployment = apigateway.Deployment(self, "Deployment", api=api)

# development stage
dev_log_group = logs.LogGroup(self, "DevLogs")
apigateway.Stage(self, "dev",
    deployment=deployment,
    access_log_destination=apigateway.LogGroupLogDestination(dev_log_group),
    access_log_format=apigateway.AccessLogFormat.json_with_standard_fields(
        caller=False,
        http_method=True,
        ip=True,
        protocol=True,
        request_time=True,
        resource_path=True,
        response_length=True,
        status=True,
        user=True
    )
)

The following code will generate the access log in the CLF format.

log_group = logs.LogGroup(self, "ApiGatewayAccessLogs")
api = apigateway.RestApi(self, "books",
    deploy_options=apigateway.StageOptions(
        access_log_destination=apigateway.LogGroupLogDestination(log_group),
        access_log_format=apigateway.AccessLogFormat.clf()
    )
)

You can also configure your own access log format by using the AccessLogFormat.custom() API. AccessLogField provides commonly used fields. The following code configures access log to contain.

log_group = logs.LogGroup(self, "ApiGatewayAccessLogs")
apigateway.RestApi(self, "books",
    deploy_options=apigateway.StageOptions(
        access_log_destination=apigateway.LogGroupLogDestination(log_group),
        access_log_format=apigateway.AccessLogFormat.custom(f"{apigateway.AccessLogField.contextRequestId()} {apigateway.AccessLogField.contextErrorMessage()} {apigateway.AccessLogField.contextErrorMessageString()}")
    )
)

You can use the methodOptions property to configure default method throttling for a stage. The following snippet configures the a stage that accepts 100 requests per minute, allowing burst up to 200 requests per minute.

api = apigateway.RestApi(self, "books")
deployment = apigateway.Deployment(self, "my-deployment", api=api)
stage = apigateway.Stage(self, "my-stage",
    deployment=deployment,
    method_options={
        "/*/*": apigateway.MethodDeploymentOptions( # This special path applies to all resource paths and all HTTP methods
            throttling_rate_limit=100,
            throttling_burst_limit=200)
    }
)

Configuring methodOptions on the deployOptions of RestApi will set the throttling behaviors on the default stage that is automatically created.

api = apigateway.RestApi(self, "books",
    deploy_options=apigateway.StageOptions(
        method_options={
            "/*/*": apigateway.MethodDeploymentOptions( # This special path applies to all resource paths and all HTTP methods
                throttling_rate_limit=100,
                throttling_burst_limit=1000)
        }
    )
)

Cross Origin Resource Sharing (CORS)

Cross-Origin Resource Sharing (CORS) is a mechanism that uses additional HTTP headers to tell browsers to give a web application running at one origin, access to selected resources from a different origin. A web application executes a cross-origin HTTP request when it requests a resource that has a different origin (domain, protocol, or port) from its own.

You can add the CORS preflight OPTIONS HTTP method to any API resource via the defaultCorsPreflightOptions option or by calling the addCorsPreflight on a specific resource.

The following example will enable CORS for all methods and all origins on all resources of the API:

apigateway.RestApi(self, "api",
    default_cors_preflight_options=apigateway.CorsOptions(
        allow_origins=apigateway.Cors.ALL_ORIGINS,
        allow_methods=apigateway.Cors.ALL_METHODS
    )
)

The following example will add an OPTIONS method to the myResource API resource, which only allows GET and PUT HTTP requests from the origin https://amazon.com.

# my_resource: apigateway.Resource


my_resource.add_cors_preflight(
    allow_origins=["https://amazon.com"],
    allow_methods=["GET", "PUT"]
)

See the CorsOptions API reference for a detailed list of supported configuration options.

You can specify defaults this at the resource level, in which case they will be applied to the entire resource sub-tree:

# resource: apigateway.Resource


subtree = resource.add_resource("subtree",
    default_cors_preflight_options=apigateway.CorsOptions(
        allow_origins=["https://amazon.com"]
    )
)

This means that all resources under subtree (inclusive) will have a preflight OPTIONS added to them.

See #906 for a list of CORS features which are not yet supported.

Endpoint Configuration

API gateway allows you to specify an API Endpoint Type. To define an endpoint type for the API gateway, use endpointConfiguration property:

api = apigateway.RestApi(self, "api",
    endpoint_configuration=apigateway.EndpointConfiguration(
        types=[apigateway.EndpointType.EDGE]
    )
)

You can also create an association between your Rest API and a VPC endpoint. By doing so, API Gateway will generate a new Route53 Alias DNS record which you can use to invoke your private APIs. More info can be found here.

Here is an example:

# some_endpoint: ec2.IVpcEndpoint


api = apigateway.RestApi(self, "api",
    endpoint_configuration=apigateway.EndpointConfiguration(
        types=[apigateway.EndpointType.PRIVATE],
        vpc_endpoints=[some_endpoint]
    )
)

By performing this association, we can invoke the API gateway using the following format:

https://{rest-api-id}-{vpce-id}.execute-api.{region}.amazonaws.com/{stage}

Private Integrations

A private integration makes it simple to expose HTTP/HTTPS resources behind an Amazon VPC for access by clients outside of the VPC. The private integration uses an API Gateway resource of VpcLink to encapsulate connections between API Gateway and targeted VPC resources. The VpcLink is then attached to the Integration of a specific API Gateway Method. The following code sets up a private integration with a network load balancer -

import aws_cdk.aws_elasticloadbalancingv2 as elbv2


vpc = ec2.Vpc(self, "VPC")
nlb = elbv2.NetworkLoadBalancer(self, "NLB",
    vpc=vpc
)
link = apigateway.VpcLink(self, "link",
    targets=[nlb]
)

integration = apigateway.Integration(
    type=apigateway.IntegrationType.HTTP_PROXY,
    options=apigateway.IntegrationOptions(
        connection_type=apigateway.ConnectionType.VPC_LINK,
        vpc_link=link
    )
)

The uri for the private integration, in the case of a VpcLink, will be set to the DNS name of the VPC Link's NLB. If the VPC Link has multiple NLBs or the VPC Link is imported or the DNS name cannot be determined for any other reason, the user is expected to specify the uri property.

Any existing VpcLink resource can be imported into the CDK app via the VpcLink.fromVpcLinkId().

awesome_link = apigateway.VpcLink.from_vpc_link_id(self, "awesome-vpc-link", "us-east-1_oiuR12Abd")

Gateway response

If the Rest API fails to process an incoming request, it returns to the client an error response without forwarding the request to the integration backend. API Gateway has a set of standard response messages that are sent to the client for each type of error. These error responses can be configured on the Rest API. The list of Gateway responses that can be configured can be found here. Learn more about Gateway Responses.

The following code configures a Gateway Response when the response is 'access denied':

api = apigateway.RestApi(self, "books-api")
api.add_gateway_response("test-response",
    type=apigateway.ResponseType.ACCESS_DENIED,
    status_code="500",
    response_headers={
        "Access-Control-Allow-Origin": "test.com",
        "test-key": "test-value"
    },
    templates={
        "application/json": "{ \"message\": $context.error.messageString, \"statusCode\": \"488\", \"type\": \"$context.error.responseType\" }"
    }
)

OpenAPI Definition

CDK supports creating a REST API by importing an OpenAPI definition file. It currently supports OpenAPI v2.0 and OpenAPI v3.0 definition files. Read more about Configuring a REST API using OpenAPI.

The following code creates a REST API using an external OpenAPI definition JSON file -

# integration: apigateway.Integration


api = apigateway.SpecRestApi(self, "books-api",
    api_definition=apigateway.ApiDefinition.from_asset("path-to-file.json")
)

books_resource = api.root.add_resource("books")
books_resource.add_method("GET", integration)

It is possible to use the addResource() API to define additional API Gateway Resources.

Note: Deployment will fail if a Resource of the same name is already defined in the Open API specification.

Note: Any default properties configured, such as defaultIntegration, defaultMethodOptions, etc. will only be applied to Resources and Methods defined in the CDK, and not the ones defined in the spec. Use the API Gateway extensions to OpenAPI to configure these.

There are a number of limitations in using OpenAPI definitions in API Gateway. Read the Amazon API Gateway important notes for REST APIs for more details.

Note: When starting off with an OpenAPI definition using SpecRestApi, it is not possible to configure some properties that can be configured directly in the OpenAPI specification file. This is to prevent people duplication of these properties and potential confusion.

Endpoint configuration

By default, SpecRestApi will create an edge optimized endpoint.

This can be modified as shown below:

# api_definition: apigateway.ApiDefinition


api = apigateway.SpecRestApi(self, "ExampleRestApi",
    api_definition=api_definition,
    endpoint_types=[apigateway.EndpointType.PRIVATE]
)

Note: For private endpoints you will still need to provide the x-amazon-apigateway-policy and x-amazon-apigateway-endpoint-configuration in your openApi file.

Metrics

The API Gateway service sends metrics around the performance of Rest APIs to Amazon CloudWatch. These metrics can be referred to using the metric APIs available on the RestApi construct. The APIs with the metric prefix can be used to get reference to specific metrics for this API. For example, the method below refers to the client side errors metric for this API.

api = apigateway.RestApi(self, "my-api")
client_error_metric = api.metric_client_error()

APIGateway v2

APIGateway v2 APIs are now moved to its own package named aws-apigatewayv2. For backwards compatibility, existing APIGateway v2 "CFN resources" (such as CfnApi) that were previously exported as part of this package, are still exported from here and have been marked deprecated. However, updates to these CloudFormation resources, such as new properties and new resource types will not be available.

Move to using aws-apigatewayv2 to get the latest APIs and updates.

---

This module is part of the AWS Cloud Development Kit project.