AWS Lambda Construct Library
This construct library allows you to define AWS Lambda Functions.
fn = lambda_.Function(self, "MyFunction",
runtime=lambda_.Runtime.NODEJS_18_X,
handler="index.handler",
code=lambda_.Code.from_asset(path.join(__dirname, "lambda-handler"))
)
Handler Code
The lambda.Code
class includes static convenience methods for various types of
runtime code.
lambda.Code.fromBucket(bucket, key[, objectVersion])
- specify an S3 object that contains the archive of your runtime code.lambda.Code.fromInline(code)
- inline the handle code as a string. This is limited to supported runtimes.lambda.Code.fromAsset(path)
- specify a directory or a .zip file in the local filesystem which will be zipped and uploaded to S3 before deployment. See also bundling asset code.lambda.Code.fromDockerBuild(path, options)
- use the result of a Docker build as code. The runtime code is expected to be located at/asset
in the image and will be zipped and uploaded to S3 as an asset.lambda.Code.fromCustomCommand(output, command, customCommandOptions)
- supply a command that is invoked during cdk synth. That command is meant to direct the generated code to output (a zip file or a directory), which is then used as the code for the created AWS Lambda.
The following example shows how to define a Python function and deploy the code
from the local directory my-lambda-handler
to it:
lambda_.Function(self, "MyLambda",
code=lambda_.Code.from_asset(path.join(__dirname, "my-lambda-handler")),
handler="index.main",
runtime=lambda_.Runtime.PYTHON_3_9
)
When deploying a stack that contains this code, the directory will be zip archived and then uploaded to an S3 bucket, then the exact location of the S3 objects will be passed when the stack is deployed.
During synthesis, the CDK expects to find a directory on disk at the asset directory specified. Note that we are referencing the asset directory relatively to our CDK project directory. This is especially important when we want to share this construct through a library. Different programming languages will have different techniques for bundling resources into libraries.
Docker Images
Lambda functions allow specifying their handlers within docker images. The docker image can be an image from ECR or a local asset that the CDK will package and load into ECR.
The following DockerImageFunction
construct uses a local folder with a
Dockerfile as the asset that will be used as the function handler.
lambda_.DockerImageFunction(self, "AssetFunction",
code=lambda_.DockerImageCode.from_image_asset(path.join(__dirname, "docker-handler"))
)
You can also specify an image that already exists in ECR as the function handler.
import aws_cdk.aws_ecr as ecr
repo = ecr.Repository(self, "Repository")
lambda_.DockerImageFunction(self, "ECRFunction",
code=lambda_.DockerImageCode.from_ecr(repo)
)
The props for these docker image resources allow overriding the image’s CMD
, ENTRYPOINT
, and WORKDIR
configurations as well as choosing a specific tag or digest. See their docs for more information.
To deploy a DockerImageFunction
on Lambda arm64
architecture, specify Architecture.ARM_64
in architecture
.
This will bundle docker image assets for arm64
architecture with --platform linux/arm64
even if build within an x86_64
host.
With that being said, if you are bundling DockerImageFunction
for Lambda amd64
architecture from a arm64
machine like a Macbook with arm64
CPU, you would
need to specify architecture: lambda.Architecture.X86_64
as well. This ensures the --platform
argument is passed to the image assets
bundling process so you can bundle up X86_64
images from the arm64
machine.
lambda_.DockerImageFunction(self, "AssetFunction",
code=lambda_.DockerImageCode.from_image_asset(path.join(__dirname, "docker-arm64-handler")),
architecture=lambda_.Architecture.ARM_64
)
Execution Role
Lambda functions assume an IAM role during execution. In CDK by default, Lambda functions will use an autogenerated Role if one is not provided.
The autogenerated Role is automatically given permissions to execute the Lambda function. To reference the autogenerated Role:
fn = lambda_.Function(self, "MyFunction",
runtime=lambda_.Runtime.NODEJS_18_X,
handler="index.handler",
code=lambda_.Code.from_asset(path.join(__dirname, "lambda-handler"))
)
role = fn.role
You can also provide your own IAM role. Provided IAM roles will not automatically be given permissions to execute the Lambda function. To provide a role and grant it appropriate permissions:
my_role = iam.Role(self, "My Role",
assumed_by=iam.ServicePrincipal("lambda.amazonaws.com")
)
fn = lambda_.Function(self, "MyFunction",
runtime=lambda_.Runtime.NODEJS_18_X,
handler="index.handler",
code=lambda_.Code.from_asset(path.join(__dirname, "lambda-handler")),
role=my_role
)
my_role.add_managed_policy(iam.ManagedPolicy.from_aws_managed_policy_name("service-role/AWSLambdaBasicExecutionRole"))
my_role.add_managed_policy(iam.ManagedPolicy.from_aws_managed_policy_name("service-role/AWSLambdaVPCAccessExecutionRole"))
Function Timeout
AWS Lambda functions have a default timeout of 3 seconds, but this can be increased
up to 15 minutes. The timeout is available as a property of Function
so that
you can reference it elsewhere in your stack. For instance, you could use it to create
a CloudWatch alarm to report when your function timed out:
import aws_cdk as cdk
import aws_cdk.aws_cloudwatch as cloudwatch
fn = lambda_.Function(self, "MyFunction",
runtime=lambda_.Runtime.NODEJS_18_X,
handler="index.handler",
code=lambda_.Code.from_asset(path.join(__dirname, "lambda-handler")),
timeout=Duration.minutes(5)
)
if fn.timeout:
cloudwatch.Alarm(self, "MyAlarm",
metric=fn.metric_duration().with(
statistic="Maximum"
),
evaluation_periods=1,
datapoints_to_alarm=1,
threshold=fn.timeout.to_milliseconds(),
treat_missing_data=cloudwatch.TreatMissingData.IGNORE,
alarm_name="My Lambda Timeout"
)
Advanced Logging
You can have more control over your function logs, by specifying the log format (Json or plain text), the system log level, the application log level, as well as choosing the log group:
from aws_cdk.aws_logs import ILogGroup
# log_group: ILogGroup
lambda_.Function(self, "Lambda",
code=lambda_.InlineCode("foo"),
handler="index.handler",
runtime=lambda_.Runtime.NODEJS_18_X,
logging_format=lambda_.LoggingFormat.JSON,
system_log_level_v2=lambda_.SystemLogLevel.INFO,
application_log_level_v2=lambda_.ApplicationLogLevel.INFO,
log_group=log_group
)
To use applicationLogLevelV2
and/or systemLogLevelV2
you must set loggingFormat
to LoggingFormat.JSON
.
Resource-based Policies
AWS Lambda supports resource-based policies for controlling access to Lambda functions and layers on a per-resource basis. In particular, this allows you to give permission to AWS services, AWS Organizations, or other AWS accounts to modify and invoke your functions.
Grant function access to AWS services
# Grant permissions to a service
# fn: lambda.Function
principal = iam.ServicePrincipal("my-service")
fn.grant_invoke(principal)
# Equivalent to:
fn.add_permission("my-service Invocation",
principal=principal
)
You can also restrict permissions given to AWS services by providing a source account or ARN (representing the account and identifier of the resource that accesses the function or layer).
Important:
By default
fn.grantInvoke()
grants permission to the principal to invoke any version of the function, including all past ones. If you only want the principal to be granted permission to invoke the latest version or the unqualified Lambda ARN, usegrantInvokeLatestVersion(grantee)
.
# fn: lambda.Function
principal = iam.ServicePrincipal("my-service")
# Grant invoke only to latest version and unqualified lambda arn
fn.grant_invoke_latest_version(principal)
If you want to grant access for invoking a specific version of Lambda function, you can use fn.grantInvokeVersion(grantee, version)
# fn: lambda.Function
# version: lambda.IVersion
principal = iam.ServicePrincipal("my-service")
# Grant invoke only to the specific version
fn.grant_invoke_version(principal, version)
For more information, see Granting function access to AWS services in the AWS Lambda Developer Guide.
Grant function access to an AWS Organization
# Grant permissions to an entire AWS organization
# fn: lambda.Function
org = iam.OrganizationPrincipal("o-xxxxxxxxxx")
fn.grant_invoke(org)
In the above example, the principal
will be *
and all users in the
organization o-xxxxxxxxxx
will get function invocation permissions.
You can restrict permissions given to the organization by specifying an
AWS account or role as the principal
:
# Grant permission to an account ONLY IF they are part of the organization
# fn: lambda.Function
account = iam.AccountPrincipal("123456789012")
fn.grant_invoke(account.in_organization("o-xxxxxxxxxx"))
For more information, see Granting function access to an organization in the AWS Lambda Developer Guide.
Grant function access to other AWS accounts
# Grant permission to other AWS account
# fn: lambda.Function
account = iam.AccountPrincipal("123456789012")
fn.grant_invoke(account)
For more information, see Granting function access to other accounts in the AWS Lambda Developer Guide.
Grant function access to unowned principals
Providing an unowned principal (such as account principals, generic ARN
principals, service principals, and principals in other accounts) to a call to
fn.grantInvoke
will result in a resource-based policy being created. If the
principal in question has conditions limiting the source account or ARN of the
operation (see above), these conditions will be automatically added to the
resource policy.
# fn: lambda.Function
service_principal = iam.ServicePrincipal("my-service")
source_arn = "arn:aws:s3:::amzn-s3-demo-bucket"
source_account = "111122223333"
service_principal_with_conditions = service_principal.with_conditions({
"ArnLike": {
"aws:SourceArn": source_arn
},
"StringEquals": {
"aws:SourceAccount": source_account
}
})
fn.grant_invoke(service_principal_with_conditions)
Grant function access to a CompositePrincipal
To grant invoke permissions to a CompositePrincipal
use the grantInvokeCompositePrincipal
method:
# fn: lambda.Function
composite_principal = iam.CompositePrincipal(
iam.OrganizationPrincipal("o-zzzzzzzzzz"),
iam.ServicePrincipal("apigateway.amazonaws.com"))
fn.grant_invoke_composite_principal(composite_principal)
Versions
You can use versions to manage the deployment of your AWS Lambda functions. For example, you can publish a new version of a function for beta testing without affecting users of the stable production version.
The function version includes the following information:
The function code and all associated dependencies.
The Lambda runtime that executes the function.
All of the function settings, including the environment variables.
A unique Amazon Resource Name (ARN) to identify this version of the function.
You could create a version to your lambda function using the Version
construct.
# fn: lambda.Function
version = lambda_.Version(self, "MyVersion",
lambda_=fn
)
The major caveat to know here is that a function version must always point to a specific ‘version’ of the function. When the function is modified, the version will continue to point to the ‘then version’ of the function.
One way to ensure that the lambda.Version
always points to the latest version
of your lambda.Function
is to set an environment variable which changes at
least as often as your code does. This makes sure the function always has the
latest code. For instance -
code_version = "stringOrMethodToGetCodeVersion"
fn = lambda_.Function(self, "MyFunction",
runtime=lambda_.Runtime.NODEJS_18_X,
handler="index.handler",
code=lambda_.Code.from_asset(path.join(__dirname, "lambda-handler")),
environment={
"CodeVersionString": code_version
}
)
The fn.latestVersion
property returns a lambda.IVersion
which represents
the $LATEST
pseudo-version.
However, most AWS services require a specific AWS Lambda version,
and won’t allow you to use $LATEST
. Therefore, you would normally want
to use lambda.currentVersion
.
The fn.currentVersion
property can be used to obtain a lambda.Version
resource that represents the AWS Lambda function defined in your application.
Any change to your function’s code or configuration will result in the creation
of a new version resource. You can specify options for this version through the
currentVersionOptions
property.
NOTE: The currentVersion
property is only supported when your AWS Lambda function
uses either lambda.Code.fromAsset
or lambda.Code.fromInline
. Other types
of code providers (such as lambda.Code.fromBucket
) require that you define a
lambda.Version
resource directly since the CDK is unable to determine if
their contents had changed.
currentVersion
: Updated hashing logic
To produce a new lambda version each time the lambda function is modified, the
currentVersion
property under the hood, computes a new logical id based on the
properties of the function. This informs CloudFormation that a new
AWS::Lambda::Version
resource should be created pointing to the updated Lambda
function.
However, a bug was introduced in this calculation that caused the logical id to
change when it was not required (ex: when the Function’s Tags
property, or
when the DependsOn
clause was modified). This caused the deployment to fail
since the Lambda service does not allow creating duplicate versions.
This has been fixed in the AWS CDK but existing users need to opt-in via a
feature flag. Users who have run cdk init
since this fix will be opted in,
by default.
Otherwise, you will need to enable the feature flag
@aws-cdk/aws-lambda:recognizeVersionProps
. Since CloudFormation does not
allow duplicate versions, you will also need to make some modification to
your function so that a new version can be created. To efficiently and trivially
modify all your lambda functions at once, you can attach the
FunctionVersionUpgrade
aspect to the stack, which slightly alters the
function description. This aspect is intended for one-time use to upgrade the
version of all your functions at the same time, and can safely be removed after
deploying once.
stack = Stack()
Aspects.of(stack).add(lambda_.FunctionVersionUpgrade(LAMBDA_RECOGNIZE_VERSION_PROPS))
When the new logic is in effect, you may rarely come across the following error:
The following properties are not recognized as version properties
. This will
occur, typically when property overrides are used, when a new property
introduced in AWS::Lambda::Function
is used that CDK is still unaware of.
To overcome this error, use the API Function.classifyVersionProperty()
to
record whether a new version should be generated when this property is changed.
This can be typically determined by checking whether the property can be
modified using the UpdateFunctionConfiguration API or not.
currentVersion
: Updated hashing logic for layer versions
An additional update to the hashing logic fixes two issues surrounding layers. Prior to this change, updating the lambda layer version would have no effect on the function version. Also, the order of lambda layers provided to the function was unnecessarily baked into the hash.
This has been fixed in the AWS CDK starting with version 2.27. If you ran
cdk init
with an earlier version, you will need to opt-in via a feature flag.
If you run cdk init
with v2.27 or later, this fix will be opted in, by default.
Existing users will need to enable the feature flag
@aws-cdk/aws-lambda:recognizeLayerVersion
. Since CloudFormation does not
allow duplicate versions, they will also need to make some modification to
their function so that a new version can be created. To efficiently and trivially
modify all your lambda functions at once, users can attach the
FunctionVersionUpgrade
aspect to the stack, which slightly alters the
function description. This aspect is intended for one-time use to upgrade the
version of all your functions at the same time, and can safely be removed after
deploying once.
stack = Stack()
Aspects.of(stack).add(lambda_.FunctionVersionUpgrade(LAMBDA_RECOGNIZE_LAYER_VERSION))
Aliases
You can define one or more aliases for your AWS Lambda function. A Lambda alias is like a pointer to a specific Lambda function version. Users can access the function version using the alias ARN.
The version.addAlias()
method can be used to define an AWS Lambda alias that
points to a specific version.
The following example defines an alias named live
which will always point to a
version that represents the function as defined in your CDK app. When you change
your lambda code or configuration, a new resource will be created. You can
specify options for the current version through the currentVersionOptions
property.
fn = lambda_.Function(self, "MyFunction",
current_version_options=lambda.VersionOptions(
removal_policy=RemovalPolicy.RETAIN, # retain old versions
retry_attempts=1
),
runtime=lambda_.Runtime.NODEJS_18_X,
handler="index.handler",
code=lambda_.Code.from_asset(path.join(__dirname, "lambda-handler"))
)
fn.add_alias("live")
Function URL
A function URL is a dedicated HTTP(S) endpoint for your Lambda function. When you create a function URL, Lambda automatically generates a unique URL endpoint for you. Function URLs can be created for the latest version Lambda Functions, or Function Aliases (but not for Versions).
Function URLs are dual stack-enabled, supporting IPv4 and IPv6, and cross-origin resource sharing (CORS) configuration. After you configure a function URL for your function, you can invoke your function through its HTTP(S) endpoint via a web browser, curl, Postman, or any HTTP client. To invoke a function using IAM authentication your HTTP client must support SigV4 signing.
See the Invoking Function URLs section of the AWS Lambda Developer Guide for more information on the input and output payloads of Functions invoked in this way.
IAM-authenticated Function URLs
To create a Function URL which can be called by an IAM identity, call addFunctionUrl()
, followed by grantInvokeFunctionUrl()
:
# Can be a Function or an Alias
# fn: lambda.Function
# my_role: iam.Role
fn_url = fn.add_function_url()
fn_url.grant_invoke_url(my_role)
CfnOutput(self, "TheUrl",
# The .url attributes will return the unique Function URL
value=fn_url.url
)
Calls to this URL need to be signed with SigV4.
Anonymous Function URLs
To create a Function URL which can be called anonymously, pass authType: FunctionUrlAuthType.NONE
to addFunctionUrl()
:
# Can be a Function or an Alias
# fn: lambda.Function
fn_url = fn.add_function_url(
auth_type=lambda_.FunctionUrlAuthType.NONE
)
CfnOutput(self, "TheUrl",
value=fn_url.url
)
CORS configuration for Function URLs
If you want your Function URLs to be invokable from a web page in browser, you will need to configure cross-origin resource sharing to allow the call (if you do not do this, your browser will refuse to make the call):
# fn: lambda.Function
fn.add_function_url(
auth_type=lambda_.FunctionUrlAuthType.NONE,
cors=lambda.FunctionUrlCorsOptions(
# Allow this to be called from websites on https://example.com.
# Can also be ['*'] to allow all domain.
allowed_origins=["https://example.com"]
)
)
Invoke Mode for Function URLs
Invoke mode determines how AWS Lambda invokes your function. You can configure the invoke mode when creating a Function URL using the invokeMode property
# fn: lambda.Function
fn.add_function_url(
auth_type=lambda_.FunctionUrlAuthType.NONE,
invoke_mode=lambda_.InvokeMode.RESPONSE_STREAM
)
If the invokeMode property is not specified, the default BUFFERED mode will be used.
Layers
The lambda.LayerVersion
class can be used to define Lambda layers and manage
granting permissions to other AWS accounts or organizations.
layer = lambda_.LayerVersion(stack, "MyLayer",
code=lambda_.Code.from_asset(path.join(__dirname, "layer-code")),
compatible_runtimes=[lambda_.Runtime.NODEJS_LATEST],
license="Apache-2.0",
description="A layer to test the L2 construct"
)
# To grant usage by other AWS accounts
layer.add_permission("remote-account-grant", account_id=aws_account_id)
# To grant usage to all accounts in some AWS Ogranization
# layer.grantUsage({ accountId: '*', organizationId });
lambda_.Function(stack, "MyLayeredLambda",
code=lambda_.InlineCode("foo"),
handler="index.handler",
runtime=lambda_.Runtime.NODEJS_LATEST,
layers=[layer]
)
By default, updating a layer creates a new layer version, and CloudFormation will delete the old version as part of the stack update.
Alternatively, a removal policy can be used to retain the old version:
lambda_.LayerVersion(self, "MyLayer",
removal_policy=RemovalPolicy.RETAIN,
code=lambda_.Code.from_asset(path.join(__dirname, "lambda-handler"))
)
Architecture
Lambda functions, by default, run on compute systems that have the 64 bit x86 architecture.
The AWS Lambda service also runs compute on the ARM architecture, which can reduce cost for some workloads.
A lambda function can be configured to be run on one of these platforms:
lambda_.Function(self, "MyFunction",
runtime=lambda_.Runtime.NODEJS_18_X,
handler="index.handler",
code=lambda_.Code.from_asset(path.join(__dirname, "lambda-handler")),
architecture=lambda_.Architecture.ARM_64
)
Similarly, lambda layer versions can also be tagged with architectures it is compatible with.
lambda_.LayerVersion(self, "MyLayer",
removal_policy=RemovalPolicy.RETAIN,
code=lambda_.Code.from_asset(path.join(__dirname, "lambda-handler")),
compatible_architectures=[lambda_.Architecture.X86_64, lambda_.Architecture.ARM_64]
)
Lambda Insights
Lambda functions can be configured to use CloudWatch Lambda Insights which provides low-level runtime metrics for a Lambda functions.
lambda_.Function(self, "MyFunction",
runtime=lambda_.Runtime.NODEJS_18_X,
handler="index.handler",
code=lambda_.Code.from_asset(path.join(__dirname, "lambda-handler")),
insights_version=lambda_.LambdaInsightsVersion.VERSION_1_0_98_0
)
If the version of insights is not yet available in the CDK, you can also provide the ARN directly as so -
layer_arn = "arn:aws:lambda:us-east-1:580247275435:layer:LambdaInsightsExtension:14"
lambda_.Function(self, "MyFunction",
runtime=lambda_.Runtime.NODEJS_18_X,
handler="index.handler",
code=lambda_.Code.from_asset(path.join(__dirname, "lambda-handler")),
insights_version=lambda_.LambdaInsightsVersion.from_insight_version_arn(layer_arn)
)
If you are deploying an ARM_64 Lambda Function, you must specify a
Lambda Insights Version >= 1_0_119_0
.
lambda_.Function(self, "MyFunction",
runtime=lambda_.Runtime.NODEJS_18_X,
handler="index.handler",
architecture=lambda_.Architecture.ARM_64,
code=lambda_.Code.from_asset(path.join(__dirname, "lambda-handler")),
insights_version=lambda_.LambdaInsightsVersion.VERSION_1_0_119_0
)
Parameters and Secrets Extension
Lambda functions can be configured to use the Parameters and Secrets Extension. The Parameters and Secrets Extension can be used to retrieve and cache secrets from Secrets Manager or parameters from Parameter Store in Lambda functions without using an SDK.
import aws_cdk.aws_secretsmanager as sm
import aws_cdk.aws_ssm as ssm
secret = sm.Secret(self, "Secret")
parameter = ssm.StringParameter(self, "Parameter",
parameter_name="mySsmParameterName",
string_value="mySsmParameterValue"
)
params_and_secrets = lambda_.ParamsAndSecretsLayerVersion.from_version(lambda_.ParamsAndSecretsVersions.V1_0_103,
cache_size=500,
log_level=lambda_.ParamsAndSecretsLogLevel.DEBUG
)
lambda_function = lambda_.Function(self, "MyFunction",
runtime=lambda_.Runtime.NODEJS_18_X,
handler="index.handler",
architecture=lambda_.Architecture.ARM_64,
code=lambda_.Code.from_asset(path.join(__dirname, "lambda-handler")),
params_and_secrets=params_and_secrets
)
secret.grant_read(lambda_function)
parameter.grant_read(lambda_function)
If the version of Parameters and Secrets Extension is not yet available in the CDK, you can also provide the ARN directly as so:
import aws_cdk.aws_secretsmanager as sm
import aws_cdk.aws_ssm as ssm
secret = sm.Secret(self, "Secret")
parameter = ssm.StringParameter(self, "Parameter",
parameter_name="mySsmParameterName",
string_value="mySsmParameterValue"
)
layer_arn = "arn:aws:lambda:us-east-1:177933569100:layer:AWS-Parameters-and-Secrets-Lambda-Extension:4"
params_and_secrets = lambda_.ParamsAndSecretsLayerVersion.from_version_arn(layer_arn,
cache_size=500
)
lambda_function = lambda_.Function(self, "MyFunction",
runtime=lambda_.Runtime.NODEJS_18_X,
handler="index.handler",
architecture=lambda_.Architecture.ARM_64,
code=lambda_.Code.from_asset(path.join(__dirname, "lambda-handler")),
params_and_secrets=params_and_secrets
)
secret.grant_read(lambda_function)
parameter.grant_read(lambda_function)
Event Rule Target
You can use an AWS Lambda function as a target for an Amazon CloudWatch event rule:
import aws_cdk.aws_events as events
import aws_cdk.aws_events_targets as targets
# fn: lambda.Function
rule = events.Rule(self, "Schedule Rule",
schedule=events.Schedule.cron(minute="0", hour="4")
)
rule.add_target(targets.LambdaFunction(fn))
Event Sources
AWS Lambda supports a variety of event sources.
In most cases, it is possible to trigger a function as a result of an event by
using one of the add<Event>Notification
methods on the source construct. For
example, the s3.Bucket
construct has an onEvent
method which can be used to
trigger a Lambda when an event, such as PutObject occurs on an S3 bucket.
An alternative way to add event sources to a function is to use function.addEventSource(source)
.
This method accepts an IEventSource
object. The module @aws-cdk/aws-lambda-event-sources
includes classes for the various event sources supported by AWS Lambda.
For example, the following code adds an SQS queue as an event source for a function:
import aws_cdk.aws_lambda_event_sources as eventsources
import aws_cdk.aws_sqs as sqs
# fn: lambda.Function
queue = sqs.Queue(self, "Queue")
fn.add_event_source(eventsources.SqsEventSource(queue))
The following code adds an S3 bucket notification as an event source:
import aws_cdk.aws_lambda_event_sources as eventsources
import aws_cdk.aws_s3 as s3
# fn: lambda.Function
bucket = s3.Bucket(self, "Bucket")
fn.add_event_source(eventsources.S3EventSource(bucket,
events=[s3.EventType.OBJECT_CREATED, s3.EventType.OBJECT_REMOVED],
filters=[s3.NotificationKeyFilter(prefix="subdir/")]
))
The following code adds an DynamoDB notification as an event source filtering insert events:
import aws_cdk.aws_lambda_event_sources as eventsources
import aws_cdk.aws_dynamodb as dynamodb
# fn: lambda.Function
table = dynamodb.Table(self, "Table",
partition_key=dynamodb.Attribute(
name="id",
type=dynamodb.AttributeType.STRING
),
stream=dynamodb.StreamViewType.NEW_IMAGE
)
fn.add_event_source(eventsources.DynamoEventSource(table,
starting_position=lambda_.StartingPosition.LATEST,
filters=[lambda_.FilterCriteria.filter({"event_name": lambda_.FilterRule.is_equal("INSERT")})]
))
By default, Lambda will encrypt Filter Criteria using AWS managed keys. But if you want to use a self managed KMS key to encrypt the filters, You can specify the self managed key using the filterEncryption
property.
import aws_cdk.aws_lambda_event_sources as eventsources
import aws_cdk.aws_dynamodb as dynamodb
from aws_cdk.aws_kms import Key
# fn: lambda.Function
table = dynamodb.Table(self, "Table",
partition_key=dynamodb.Attribute(
name="id",
type=dynamodb.AttributeType.STRING
),
stream=dynamodb.StreamViewType.NEW_IMAGE
)
# Your self managed KMS key
my_key = Key.from_key_arn(self, "SourceBucketEncryptionKey", "arn:aws:kms:us-east-1:123456789012:key/<key-id>")
fn.add_event_source(eventsources.DynamoEventSource(table,
starting_position=lambda_.StartingPosition.LATEST,
filters=[lambda_.FilterCriteria.filter({"event_name": lambda_.FilterRule.is_equal("INSERT")})],
filter_encryption=my_key
))
Lambda requires allow
kms:Decrypt
on Lambda principallambda.amazonaws.com
to use the key for Filter Criteria Encryption. If you create the KMS key in the stack, CDK will automatically add this permission to the Key when you creates eventSourceMapping. However, if you import the key using function likeKey.fromKeyArn
then you need to add the following permission to the KMS key before using it to encrypt Filter Criteria
{
"Version": "2012-10-17",
"Statement": [
{
"Effect": "Allow",
"Principal": {
"Service": "lambda.amazonaws.com"
},
"Action": "kms:Decrypt",
"Resource": "*"
}
]
}
See the documentation for the @aws-cdk/aws-lambda-event-sources module for more details.
Imported Lambdas
When referencing an imported lambda in the CDK, use fromFunctionArn()
for most use cases:
fn = lambda_.Function.from_function_arn(self, "Function", "arn:aws:lambda:us-east-1:123456789012:function:MyFn")
The fromFunctionAttributes()
API is available for more specific use cases:
fn = lambda_.Function.from_function_attributes(self, "Function",
function_arn="arn:aws:lambda:us-east-1:123456789012:function:MyFn",
# The following are optional properties for specific use cases and should be used with caution:
# Use Case: imported function is in the same account as the stack. This tells the CDK that it
# can modify the function's permissions.
same_environment=True,
# Use Case: imported function is in a different account and user commits to ensuring that the
# imported function has the correct permissions outside the CDK.
skip_permissions=True
)
Function.fromFunctionArn()
and Function.fromFunctionAttributes()
will attempt to parse the Function’s Region and Account ID from the ARN. addPermissions
will only work on the Function
object if the Region and Account ID are deterministically the same as the scope of the Stack the referenced Function
object is created in.
If the containing Stack is environment-agnostic or the Function ARN is a Token, this comparison will fail, and calls to Function.addPermission
will do nothing.
If you know Function permissions can safely be added, you can use Function.fromFunctionName()
instead, or pass sameEnvironment: true
to Function.fromFunctionAttributes()
.
fn = lambda_.Function.from_function_name(self, "Function", "MyFn")
Lambda with DLQ
A dead-letter queue can be automatically created for a Lambda function by
setting the deadLetterQueueEnabled: true
configuration. In such case CDK creates
a sqs.Queue
as deadLetterQueue
.
fn = lambda_.Function(self, "MyFunction",
runtime=lambda_.Runtime.NODEJS_18_X,
handler="index.handler",
code=lambda_.Code.from_inline("exports.handler = function(event, ctx, cb) { return cb(null, \"hi\"); }"),
dead_letter_queue_enabled=True
)
It is also possible to provide a dead-letter queue instead of getting a new queue created:
import aws_cdk.aws_sqs as sqs
dlq = sqs.Queue(self, "DLQ")
fn = lambda_.Function(self, "MyFunction",
runtime=lambda_.Runtime.NODEJS_18_X,
handler="index.handler",
code=lambda_.Code.from_inline("exports.handler = function(event, ctx, cb) { return cb(null, \"hi\"); }"),
dead_letter_queue=dlq
)
You can also use a sns.Topic
instead of an sqs.Queue
as dead-letter queue:
import aws_cdk.aws_sns as sns
dlt = sns.Topic(self, "DLQ")
fn = lambda_.Function(self, "MyFunction",
runtime=lambda_.Runtime.NODEJS_18_X,
handler="index.handler",
code=lambda_.Code.from_inline("// your code here"),
dead_letter_topic=dlt
)
See the AWS documentation to learn more about AWS Lambdas and DLQs.
Lambda with X-Ray Tracing
fn = lambda_.Function(self, "MyFunction",
runtime=lambda_.Runtime.NODEJS_18_X,
handler="index.handler",
code=lambda_.Code.from_inline("exports.handler = function(event, ctx, cb) { return cb(null, \"hi\"); }"),
tracing=lambda_.Tracing.ACTIVE
)
See the AWS documentation to learn more about AWS Lambda’s X-Ray support.
Lambda with AWS Distro for OpenTelemetry layer
To have automatic integration with XRay without having to add dependencies or change your code, you can use the AWS Distro for OpenTelemetry Lambda (ADOT) layer. Consuming the latest ADOT layer can be done with the following snippet:
from aws_cdk.aws_lambda import AdotLambdaExecWrapper, AdotLayerVersion, AdotLambdaLayerJavaScriptSdkVersion
fn = lambda_.Function(self, "MyFunction",
runtime=lambda_.Runtime.NODEJS_18_X,
handler="index.handler",
code=lambda_.Code.from_inline("exports.handler = function(event, ctx, cb) { return cb(null, \"hi\"); }"),
adot_instrumentation=lambda.AdotInstrumentationConfig(
layer_version=AdotLayerVersion.from_java_script_sdk_layer_version(AdotLambdaLayerJavaScriptSdkVersion.LATEST),
exec_wrapper=AdotLambdaExecWrapper.REGULAR_HANDLER
)
)
To use a different layer version, use one of the following helper functions for the layerVersion
prop:
AdotLayerVersion.fromJavaScriptSdkLayerVersion
AdotLayerVersion.fromPythonSdkLayerVersion
AdotLayerVersion.fromJavaSdkLayerVersion
AdotLayerVersion.fromJavaAutoInstrumentationSdkLayerVersion
AdotLayerVersion.fromGenericSdkLayerVersion
Each helper function expects a version value from a corresponding enum-like class as below:
AdotLambdaLayerJavaScriptSdkVersion
AdotLambdaLayerPythonSdkVersion
AdotLambdaLayerJavaSdkVersion
AdotLambdaLayerJavaAutoInstrumentationSdkVersion
AdotLambdaLayerGenericSdkVersion
For more examples, see our the integration test.
If you want to retrieve the ARN of the ADOT Lambda layer without enabling ADOT in a Lambda function:
# fn: lambda.Function
layer_arn = lambda_.AdotLambdaLayerJavaSdkVersion.V1_19_0.layer_arn(fn.stack, fn.architecture)
When using the AdotLambdaLayerPythonSdkVersion
the AdotLambdaExecWrapper
needs to be AdotLambdaExecWrapper.INSTRUMENT_HANDLER
as per AWS Distro for OpenTelemetry Lambda Support For Python
Lambda with Profiling
The following code configures the lambda function with CodeGuru profiling. By default, this creates a new CodeGuru profiling group -
fn = lambda_.Function(self, "MyFunction",
runtime=lambda_.Runtime.PYTHON_3_9,
handler="index.handler",
code=lambda_.Code.from_asset("lambda-handler"),
profiling=True
)
The profilingGroup
property can be used to configure an existing CodeGuru profiler group.
CodeGuru profiling is supported for all Java runtimes and Python3.6+ runtimes.
See the AWS documentation to learn more about AWS Lambda’s Profiling support.
Lambda with Reserved Concurrent Executions
fn = lambda_.Function(self, "MyFunction",
runtime=lambda_.Runtime.NODEJS_18_X,
handler="index.handler",
code=lambda_.Code.from_inline("exports.handler = function(event, ctx, cb) { return cb(null, \"hi\"); }"),
reserved_concurrent_executions=100
)
https://docs.aws.amazon.com/lambda/latest/dg/invocation-recursion.html
Lambda with SnapStart
SnapStart is currently supported only on Java 11 and later Java managed runtimes. SnapStart does not support provisioned concurrency, Amazon Elastic File System (Amazon EFS), or ephemeral storage greater than 512 MB. After you enable Lambda SnapStart for a particular Lambda function, publishing a new version of the function will trigger an optimization process.
See the AWS documentation to learn more about AWS Lambda SnapStart
fn = lambda_.Function(self, "MyFunction",
code=lambda_.Code.from_asset(path.join(__dirname, "handler.zip")),
runtime=lambda_.Runtime.JAVA_11,
handler="example.Handler::handleRequest",
snap_start=lambda_.SnapStartConf.ON_PUBLISHED_VERSIONS
)
version = fn.current_version
AutoScaling
You can use Application AutoScaling to automatically configure the provisioned concurrency for your functions. AutoScaling can be set to track utilization or be based on a schedule. To configure AutoScaling on a function alias:
import aws_cdk.aws_autoscaling as autoscaling
# fn: lambda.Function
alias = fn.add_alias("prod")
# Create AutoScaling target
as = alias.add_auto_scaling(max_capacity=50)
# Configure Target Tracking
as.scale_on_utilization(
utilization_target=0.5
)
# Configure Scheduled Scaling
as.scale_on_schedule("ScaleUpInTheMorning",
schedule=autoscaling.Schedule.cron(hour="8", minute="0"),
min_capacity=20
)
import aws_cdk.aws_applicationautoscaling as appscaling
import aws_cdk as cdk
from ......cx_api import LAMBDA_RECOGNIZE_LAYER_VERSION
import aws_cdk as lambda_
#
# Stack verification steps:
# aws application-autoscaling describe-scalable-targets --service-namespace lambda --resource-ids function:<function name>:prod
# has a minCapacity of 3 and maxCapacity of 50
#
class TestStack(cdk.Stack):
def __init__(self, scope, id):
super().__init__(scope, id)
fn = lambda_.Function(self, "MyLambda",
code=lambda_.InlineCode("exports.handler = async () => { console.log('hello world'); };"),
handler="index.handler",
runtime=lambda_.Runtime.NODEJS_LATEST
)
version = fn.current_version
alias = lambda_.Alias(self, "Alias",
alias_name="prod",
version=version
)
scaling_target = alias.add_auto_scaling(min_capacity=3, max_capacity=50)
scaling_target.scale_on_utilization(
utilization_target=0.5
)
scaling_target.scale_on_schedule("ScaleUpInTheMorning",
schedule=appscaling.Schedule.cron(hour="8", minute="0"),
min_capacity=20
)
scaling_target.scale_on_schedule("ScaleDownAtNight",
schedule=appscaling.Schedule.cron(hour="20", minute="0"),
max_capacity=20
)
cdk.CfnOutput(self, "FunctionName",
value=fn.function_name
)
app = cdk.App()
stack = TestStack(app, "aws-lambda-autoscaling")
# Changes the function description when the feature flag is present
# to validate the changed function hash.
cdk.Aspects.of(stack).add(lambda_.FunctionVersionUpgrade(LAMBDA_RECOGNIZE_LAYER_VERSION))
app.synth()
See the AWS documentation on autoscaling lambda functions.
Log Group
By default, Lambda functions automatically create a log group with the name /aws/lambda/<function-name>
upon first execution with
log data set to never expire.
This is convenient, but prevents you from changing any of the properties of this auto-created log group using the AWS CDK.
For example you cannot set log retention or assign a data protection policy.
To fully customize the logging behavior of your Lambda function, create a logs.LogGroup
ahead of time and use the logGroup
property to instruct the Lambda function to send logs to it.
This way you can use the full features set supported by Amazon CloudWatch Logs.
from aws_cdk.aws_logs import LogGroup
my_log_group = LogGroup(self, "MyLogGroupWithLogGroupName",
log_group_name="customLogGroup"
)
lambda_.Function(self, "Lambda",
code=lambda_.InlineCode("foo"),
handler="index.handler",
runtime=lambda_.Runtime.NODEJS_18_X,
log_group=my_log_group
)
Providing a user-controlled log group was rolled out to commercial regions on 2023-11-16. If you are deploying to another type of region, please check regional availability first.
Lambda with Recursive Loop protection
Recursive loop protection is to stop unintended loops. The customers are opted in by default for Lambda to detect and terminate unintended loops between Lambda and other AWS Services. The property can be assigned two values here, “Allow” and “Terminate”.
The default value is set to “Terminate”, which lets the Lambda to detect and terminate the recursive loops.
When the value is set to “Allow”, the customers opt out of recursive loop detection and Lambda does not terminate recursive loops if any.
See the AWS documentation to learn more about AWS Lambda Recusrive Loop Detection
fn = lambda_.Function(self, "MyFunction",
code=lambda_.Code.from_asset(path.join(__dirname, "handler.zip")),
runtime=lambda_.Runtime.JAVA_11,
handler="example.Handler::handleRequest",
recursive_loop=lambda_.RecursiveLoop.TERMINATE
)
Legacy Log Retention
As an alternative to providing a custom, user controlled log group, the legacy logRetention
property can be used to set a different expiration period.
This feature uses a Custom Resource to change the log retention of the automatically created log group.
By default, CDK uses the AWS SDK retry options when creating a log group. The logRetentionRetryOptions
property
allows you to customize the maximum number of retries and base backoff duration.
Note that a CloudFormation custom resource is added to the stack that pre-creates the log group as part of the stack deployment, if it already doesn’t exist, and sets the correct log retention period (never expire, by default). This Custom Resource will also create a log group to log events of the custom resource. The log retention period for this addtional log group is hard-coded to 1 day.
Further note that, if the log group already exists and the logRetention
is not set, the custom resource will reset
the log retention to never expire even if it was configured with a different value.
FileSystem Access
You can configure a function to mount an Amazon Elastic File System (Amazon EFS) to a
directory in your runtime environment with the filesystem
property. To access Amazon EFS
from lambda function, the Amazon EFS access point will be required.
The following sample allows the lambda function to mount the Amazon EFS access point to /mnt/msg
in the runtime environment and access the filesystem with the POSIX identity defined in posixUser
.
import aws_cdk.aws_ec2 as ec2
import aws_cdk.aws_efs as efs
# create a new VPC
vpc = ec2.Vpc(self, "VPC")
# create a new Amazon EFS filesystem
file_system = efs.FileSystem(self, "Efs", vpc=vpc)
# create a new access point from the filesystem
access_point = file_system.add_access_point("AccessPoint",
# set /export/lambda as the root of the access point
path="/export/lambda",
# as /export/lambda does not exist in a new efs filesystem, the efs will create the directory with the following createAcl
create_acl=efs.Acl(
owner_uid="1001",
owner_gid="1001",
permissions="750"
),
# enforce the POSIX identity so lambda function will access with this identity
posix_user=efs.PosixUser(
uid="1001",
gid="1001"
)
)
fn = lambda_.Function(self, "MyLambda",
# mount the access point to /mnt/msg in the lambda runtime environment
filesystem=lambda_.FileSystem.from_efs_access_point(access_point, "/mnt/msg"),
runtime=lambda_.Runtime.NODEJS_18_X,
handler="index.handler",
code=lambda_.Code.from_asset(path.join(__dirname, "lambda-handler")),
vpc=vpc
)
IPv6 support
You can configure IPv6 connectivity for lambda function by setting Ipv6AllowedForDualStack
to true.
It allows Lambda functions to specify whether the IPv6 traffic should be allowed when using dual-stack VPCs.
To access IPv6 network using Lambda, Dual-stack VPC is required. Using dual-stack VPC a function communicates with subnet over either of IPv4 or IPv6.
import aws_cdk.aws_ec2 as ec2
nat_provider = ec2.NatProvider.gateway()
# create dual-stack VPC
vpc = ec2.Vpc(self, "DualStackVpc",
ip_protocol=ec2.IpProtocol.DUAL_STACK,
subnet_configuration=[ec2.SubnetConfiguration(
name="Ipv6Public1",
subnet_type=ec2.SubnetType.PUBLIC
), ec2.SubnetConfiguration(
name="Ipv6Public2",
subnet_type=ec2.SubnetType.PUBLIC
), ec2.SubnetConfiguration(
name="Ipv6Private1",
subnet_type=ec2.SubnetType.PRIVATE_WITH_EGRESS
)
],
nat_gateway_provider=nat_provider
)
nat_gateway_id = nat_provider.configured_gateways[0].gateway_id
(vpc.private_subnets[0]).add_ipv6_nat64_route(nat_gateway_id)
fn = lambda_.Function(self, "Lambda_with_IPv6_VPC",
code=lambda_.InlineCode("def main(event, context): pass"),
handler="index.main",
runtime=lambda_.Runtime.PYTHON_3_9,
vpc=vpc,
ipv6_allowed_for_dual_stack=True
)
Outbound traffic
By default, when creating a Lambda function, it would add a security group outbound rule to allow sending all network traffic (except IPv6). This is controlled by allowAllOutbound
in function properties, which has a default value of true
.
To allow outbound IPv6 traffic by default, explicitly set allowAllIpv6Outbound
to true
in function properties as shown below (the default value for allowAllIpv6Outbound
is false
):
import aws_cdk.aws_ec2 as ec2
vpc = ec2.Vpc(self, "Vpc")
fn = lambda_.Function(self, "LambdaWithIpv6Outbound",
code=lambda_.InlineCode("def main(event, context): pass"),
handler="index.main",
runtime=lambda_.Runtime.PYTHON_3_9,
vpc=vpc,
allow_all_ipv6_outbound=True
)
Do not specify allowAllOutbound
or allowAllIpv6Outbound
property if the securityGroups
or securityGroup
property is set. Instead, configure these properties directly on the security group.
Ephemeral Storage
You can configure ephemeral storage on a function to control the amount of storage it gets for reading
or writing data, allowing you to use AWS Lambda for ETL jobs, ML inference, or other data-intensive workloads.
The ephemeral storage will be accessible in the functions’ /tmp
directory.
from aws_cdk import Size
fn = lambda_.Function(self, "MyFunction",
runtime=lambda_.Runtime.NODEJS_18_X,
handler="index.handler",
code=lambda_.Code.from_asset(path.join(__dirname, "lambda-handler")),
ephemeral_storage_size=Size.mebibytes(1024)
)
Read more about using this feature in this AWS blog post.
Singleton Function
The SingletonFunction
construct is a way to guarantee that a lambda function will be guaranteed to be part of the stack,
once and only once, irrespective of how many times the construct is declared to be part of the stack. This is guaranteed
as long as the uuid
property and the optional lambdaPurpose
property stay the same whenever they’re declared into the
stack.
A typical use case of this function is when a higher level construct needs to declare a Lambda function as part of it but
needs to guarantee that the function is declared once. However, a user of this higher level construct can declare it any
number of times and with different properties. Using SingletonFunction
here with a fixed uuid
will guarantee this.
For example, the AwsCustomResource
construct requires only one single lambda function for all api calls that are made.
Bundling Asset Code
When using lambda.Code.fromAsset(path)
it is possible to bundle the code by running a
command in a Docker container. The asset path will be mounted at /asset-input
. The
Docker container is responsible for putting content at /asset-output
. The content at
/asset-output
will be zipped and used as Lambda code.
Example with Python:
lambda_.Function(self, "Function",
code=lambda_.Code.from_asset(path.join(__dirname, "my-python-handler"),
bundling=BundlingOptions(
image=lambda_.Runtime.PYTHON_3_9.bundling_image,
command=["bash", "-c", "pip install -r requirements.txt -t /asset-output && cp -au . /asset-output"
]
)
),
runtime=lambda_.Runtime.PYTHON_3_9,
handler="index.handler"
)
Runtimes expose a bundlingImage
property that points to the AWS SAM build image.
Use cdk.DockerImage.fromRegistry(image)
to use an existing image or
cdk.DockerImage.fromBuild(path)
to build a specific image:
lambda_.Function(self, "Function",
code=lambda_.Code.from_asset("/path/to/handler",
bundling=BundlingOptions(
image=DockerImage.from_build("/path/to/dir/with/DockerFile",
build_args={
"ARG1": "value1"
}
),
command=["my", "cool", "command"]
)
),
runtime=lambda_.Runtime.PYTHON_3_9,
handler="index.handler"
)
Language-specific APIs
Language-specific higher level constructs are provided in separate modules:
Code Signing
Code signing for AWS Lambda helps to ensure that only trusted code runs in your Lambda functions. When enabled, AWS Lambda checks every code deployment and verifies that the code package is signed by a trusted source. For more information, see Configuring code signing for AWS Lambda. The following code configures a function with code signing.
Please note the code will not be automatically signed before deployment. To ensure your code is properly signed, you’ll need to conduct the code signing process either through the AWS CLI (Command Line Interface) start-signing-job or by accessing the AWS Signer console.
import aws_cdk.aws_signer as signer
signing_profile = signer.SigningProfile(self, "SigningProfile",
platform=signer.Platform.AWS_LAMBDA_SHA384_ECDSA
)
code_signing_config = lambda_.CodeSigningConfig(self, "CodeSigningConfig",
signing_profiles=[signing_profile]
)
lambda_.Function(self, "Function",
code_signing_config=code_signing_config,
runtime=lambda_.Runtime.NODEJS_18_X,
handler="index.handler",
code=lambda_.Code.from_asset(path.join(__dirname, "lambda-handler"))
)
Runtime updates
Lambda runtime management controls help reduce the risk of impact to your workloads in the rare event of a runtime version incompatibility. For more information, see Runtime management controls
lambda_.Function(self, "Lambda",
runtime_management_mode=lambda_.RuntimeManagementMode.AUTO,
runtime=lambda_.Runtime.NODEJS_18_X,
handler="index.handler",
code=lambda_.Code.from_asset(path.join(__dirname, "lambda-handler"))
)
If you want to set the “Manual” setting, using the ARN of the runtime version as the argument.
lambda_.Function(self, "Lambda",
runtime_management_mode=lambda_.RuntimeManagementMode.manual("runtimeVersion-arn"),
runtime=lambda_.Runtime.NODEJS_18_X,
handler="index.handler",
code=lambda_.Code.from_asset(path.join(__dirname, "lambda-handler"))
)
Exclude Patterns for Assets
When using lambda.Code.fromAsset(path)
an exclude
property allows you to ignore particular files for assets by providing patterns for file paths to exclude. Note that this has no effect on Assets
bundled using the bundling
property.
The ignoreMode
property can be used with the exclude
property to specify the file paths to ignore based on the .gitignore specification or the .dockerignore specification. The default behavior is to ignore file paths based on simple glob patterns.
lambda_.Function(self, "Function",
code=lambda_.Code.from_asset(path.join(__dirname, "my-python-handler"),
exclude=["*.ignore"],
ignore_mode=IgnoreMode.DOCKER
),
runtime=lambda_.Runtime.PYTHON_3_9,
handler="index.handler"
)
You can also write to include only certain files by using a negation.
lambda_.Function(self, "Function",
code=lambda_.Code.from_asset(path.join(__dirname, "my-python-handler"),
exclude=["*", "!index.py"]
),
runtime=lambda_.Runtime.PYTHON_3_9,
handler="index.handler"
)