AWSCDK::S3Deployment

14 types

AWS S3 Deployment Construct Library

This library allows populating an S3 bucket with the contents of .zip files from other S3 buckets or from local disk.

The following example defines a publicly accessible S3 bucket with web hosting enabled and populates it from a local directory on disk.

website_bucket = AWSCDK::S3::Bucket.new(self, "WebsiteBucket", {
    website_index_document: "index.html",
    public_read_access: true,
})

AWSCDK::S3Deployment::BucketDeployment.new(self, "DeployWebsite", {
    sources: [AWSCDK::S3Deployment::Source.asset("./website-dist")],
    destination_bucket: website_bucket,
    destination_key_prefix: "web/static",
})

This is what happens under the hood:

  1. When this stack is deployed (either via cdk deploy or via CI/CD), the contents of the local website-dist directory will be archived and uploaded to an intermediary assets bucket (the StagingBucket of the CDK bootstrap stack). If there is more than one source, they will be individually uploaded.
  2. The BucketDeployment construct synthesizes a Lambda-backed custom CloudFormation resource of type Custom::CDKBucketDeployment into the template. The source bucket/key is set to point to the assets bucket.
  3. The custom resource invokes its associated Lambda function, which downloads the .zip archive, extracts it and issues aws s3 sync --delete against the destination bucket (in this case website_bucket). If there is more than one source, the sources will be downloaded and merged pre-deployment at this step.

If you are referencing the filled bucket in another construct that depends on the files already be there, be sure to use deployment.deployedBucket. This will ensure the bucket deployment has finished before the resource that uses the bucket is created:

website_bucket = nil # AWSCDK::S3::Bucket


deployment = AWSCDK::S3Deployment::BucketDeployment.new(self, "DeployWebsite", {
    sources: [AWSCDK::S3Deployment::Source.asset(path.join(__dirname, "my-website"))],
    destination_bucket: website_bucket,
})

ConstructThatReadsFromTheBucket.new(self, "Consumer", {
    # Use 'deployment.deployedBucket' instead of 'websiteBucket' here
    bucket: deployment.deployed_bucket,
})

It is also possible to add additional sources using the add_source method.

website_bucket = nil # AWSCDK::S3::IBucket


deployment = AWSCDK::S3Deployment::BucketDeployment.new(self, "DeployWebsite", {
    sources: [AWSCDK::S3Deployment::Source.asset("./website-dist")],
    destination_bucket: website_bucket,
    destination_key_prefix: "web/static",
})

deployment.add_source(AWSCDK::S3Deployment::Source.asset("./another-asset"))

For the Lambda function to download object(s) from the source bucket, besides the obvious s3:GetObject* permissions, the Lambda's execution role needs the kms:Decrypt and kms:DescribeKey permissions on the KMS key that is used to encrypt the bucket. By default, when the source bucket is encrypted with the S3 managed key of the account, these permissions are granted by the key's resource-based policy, so they do not need to be on the Lambda's execution role policy explicitly. However, if the encryption key is not the s3 managed one, its resource-based policy is quite likely to NOT grant such KMS permissions. In this situation, the Lambda execution will fail with an error message like below:

download failed: ...
An error occurred (AccessDenied) when calling the GetObject operation:
User: *** is not authorized to perform: kms:Decrypt on the resource associated with this ciphertext
because no identity-based policy allows the kms:Decrypt action

When this happens, users can use the public handler_role property of BucketDeployment to manually add the KMS permissions:

destination_bucket = nil # AWSCDK::S3::Bucket


deployment = AWSCDK::S3Deployment::BucketDeployment.new(self, "DeployFiles", {
    sources: [AWSCDK::S3Deployment::Source.asset(path.join(__dirname, "source-files"))],
    destination_bucket: destination_bucket,
})

deployment.handler_role.add_to_policy(
AWSCDK::IAM::PolicyStatement.new({
    actions: ["kms:Decrypt", "kms:DescribeKey"],
    effect: AWSCDK::IAM::Effect::ALLOW,
    resources: ["<encryption key ARN>"],
}))

The situation above could arise from the following scenarios:

Supported sources

The following source types are supported for bucket deployments:

To create a source from a single file, you can pass AssetOptions to exclude all but a single file:

IMPORTANT The aws-s3-deployment module is only intended to be used with zip files from trusted sources. Directories bundled by the CDK CLI (by using Source.asset() on a directory) are safe. If you are using Source.asset() or Source.bucket() to reference an existing zip file, make sure you trust the file you are referencing. Zips from untrusted sources might be able to execute arbitrary code in the Lambda Function used by this module, and use its permissions to read or write unexpected files in the S3 bucket.

Retain on Delete

By default, the contents of the destination bucket will not be deleted when the BucketDeployment resource is removed from the stack or when the destination is changed. You can use the option retainOnDelete: false to disable this behavior, in which case the contents will be deleted.

Configuring this has a few implications you should be aware of:

Changing the logical ID of the BucketDeployment construct, without changing the destination (for example due to refactoring, or intentional ID change) will result in the deletion of the objects. This is because CloudFormation will first create the new resource, which will have no affect, followed by a deletion of the old resource, which will cause a deletion of the objects, since the destination hasn't changed, and retain_on_delete is false.

When the destination bucket or prefix is changed, all files in the previous destination will first be deleted and then uploaded to the new destination location. This could have availability implications on your users.

General Recommendations

Shared Bucket

If the destination bucket is not dedicated to the specific BucketDeployment construct (i.e shared by other entities), we recommend to always configure the destination_key_prefix property. This will prevent the deployment from accidentally deleting data that wasn't uploaded by it.

Dedicated Bucket

If the destination bucket is dedicated, it might be reasonable to skip the prefix configuration, in which case, we recommend to remove retainOnDelete: false, and instead, configure the auto_delete_objects property on the destination bucket. This will avoid the logical ID problem mentioned above.

Prune

By default, files in the destination bucket that don't exist in the source will be deleted when the BucketDeployment resource is created or updated. You can use the option prune: false to disable this behavior, in which case the files will not be deleted.

destination_bucket = nil # AWSCDK::S3::Bucket

AWSCDK::S3Deployment::BucketDeployment.new(self, "DeployMeWithoutDeletingFilesOnDestination", {
    sources: [AWSCDK::S3Deployment::Source.asset(path.join(__dirname, "my-website"))],
    destination_bucket: destination_bucket,
    prune: false,
})

This option also enables you to multiple bucket deployments for the same destination bucket & prefix, each with its own characteristics. For example, you can set different cache-control headers based on file extensions:

destination_bucket = nil # AWSCDK::S3::Bucket

AWSCDK::S3Deployment::BucketDeployment.new(self, "BucketDeployment", {
    sources: [AWSCDK::S3Deployment::Source.asset("./website", {exclude: ["index.html"]})],
    destination_bucket: destination_bucket,
    cache_control: [
        AWSCDK::S3Deployment::CacheControl.max_age(AWSCDK::Duration.days(365)),
        AWSCDK::S3Deployment::CacheControl.immutable,
    ],
    prune: false,
})

AWSCDK::S3Deployment::BucketDeployment.new(self, "HTMLBucketDeployment", {
    sources: [AWSCDK::S3Deployment::Source.asset("./website", {exclude: ["*", "!index.html"]})],
    destination_bucket: destination_bucket,
    cache_control: [
        AWSCDK::S3Deployment::CacheControl.max_age(AWSCDK::Duration.seconds(0)),
    ],
    prune: false,
})

Exclude and Include Filters

There are two points at which filters are evaluated in a deployment: asset bundling and the actual deployment. If you simply want to exclude files in the asset bundling process, you should leverage the exclude property of AssetOptions when defining your source:

destination_bucket = nil # AWSCDK::S3::Bucket

AWSCDK::S3Deployment::BucketDeployment.new(self, "HTMLBucketDeployment", {
    sources: [AWSCDK::S3Deployment::Source.asset("./website", {exclude: ["*", "!index.html"]})],
    destination_bucket: destination_bucket,
})

If you want to specify filters to be used in the deployment process, you can use the exclude and include filters on BucketDeployment. If excluded, these files will not be deployed to the destination bucket. In addition, if the file already exists in the destination bucket, it will not be deleted if you are using the prune option:

destination_bucket = nil # AWSCDK::S3::Bucket

AWSCDK::S3Deployment::BucketDeployment.new(self, "DeployButExcludeSpecificFiles", {
    sources: [AWSCDK::S3Deployment::Source.asset(path.join(__dirname, "my-website"))],
    destination_bucket: destination_bucket,
    exclude: ["*.txt"],
})

These filters follow the same format that is used for the AWS CLI. See the CLI documentation for information on Using Include and Exclude Filters.

Objects metadata

You can specify metadata to be set on all the objects in your deployment. There are 2 types of metadata in S3: system-defined metadata and user-defined metadata. System-defined metadata have a special purpose, for example cache-control defines how long to keep an object cached. User-defined metadata are not used by S3 and keys always begin with x-amz-meta- (this prefix is added automatically).

System defined metadata keys include the following:

You can find more information about system defined metadata keys in S3 PutObject documentation and aws s3 sync documentation.

website_bucket = AWSCDK::S3::Bucket.new(self, "WebsiteBucket", {
    website_index_document: "index.html",
    public_read_access: true,
})

AWSCDK::S3Deployment::BucketDeployment.new(self, "DeployWebsite", {
    sources: [AWSCDK::S3Deployment::Source.asset("./website-dist")],
    destination_bucket: website_bucket,
    destination_key_prefix: "web/static",
     # optional prefix in destination bucket
    metadata: {A: "1", b: "2"},
     # user-defined metadata

    # system-defined metadata
    content_type: "text/html",
    content_language: "en",
    storage_class: AWSCDK::S3Deployment::StorageClass::INTELLIGENT_TIERING,
    server_side_encryption: AWSCDK::S3Deployment::ServerSideEncryption::AES_256,
    cache_control: [
        AWSCDK::S3Deployment::CacheControl.set_public,
        AWSCDK::S3Deployment::CacheControl.max_age(AWSCDK::Duration.hours(1)),
    ],
    access_control: AWSCDK::S3::BucketAccessControl::BUCKET_OWNER_FULL_CONTROL,
})

CloudFront Invalidation

You can provide a CloudFront distribution and optional paths to invalidate after the bucket deployment finishes.

require 'aws-cdk-lib'


bucket = AWSCDK::S3::Bucket.new(self, "Destination")

# Handles buckets whether or not they are configured for website hosting.
distribution = AWSCDK::CloudFront::Distribution.new(self, "Distribution", {
    default_behavior: {origin: AWSCDK::CloudFrontOrigins::S3Origin.new(bucket)},
})

AWSCDK::S3Deployment::BucketDeployment.new(self, "DeployWithInvalidation", {
    sources: [AWSCDK::S3Deployment::Source.asset("./website-dist")],
    destination_bucket: bucket,
    distribution: distribution,
    distribution_paths: ["/images/*.png"],
})

By default, the deployment will wait for invalidation to succeed to complete. This will poll Cloudfront for a maximum of 13 minutes to check for a successful invalidation. The drawback to this is that the deployment will fail if invalidation fails or if it takes longer than 13 minutes. As a workaround, there is the option wait_for_distribution_invalidation, which can be set to false to skip waiting for the invalidation, but this can be risky as invalidation errors will not be reported.

require 'aws-cdk-lib'

bucket = nil # AWSCDK::S3::IBucket
distribution = nil # AWSCDK::CloudFront::IDistribution


AWSCDK::S3Deployment::BucketDeployment.new(self, "DeployWithInvalidation", {
    sources: [AWSCDK::S3Deployment::Source.asset("./website-dist")],
    destination_bucket: bucket,
    distribution: distribution,
    distribution_paths: ["/images/*.png"],
    # Invalidate cache but don't wait or verify that invalidation has completed successfully.
    wait_for_distribution_invalidation: false,
})

Signed Content Payloads

By default, deployment uses streaming uploads which set the x-amz-content-sha256 request header to UNSIGNED-PAYLOAD (matching default behavior of the AWS CLI tool). In cases where bucket policy restrictions require signed content payloads, you can enable generation of a signed x-amz-content-sha256 request header with signContent: true.

bucket = nil # AWSCDK::S3::IBucket


AWSCDK::S3Deployment::BucketDeployment.new(self, "DeployWithSignedPayloads", {
    sources: [AWSCDK::S3Deployment::Source.asset("./website-dist")],
    destination_bucket: bucket,
    sign_content: true,
})

Size Limits

The default memory limit for the deployment resource is 128MiB. If you need to copy larger files, you can use the memory_limit configuration to increase the size of the AWS Lambda resource handler.

The default ephemeral storage size for the deployment resource is 512MiB. If you need to upload larger files, you may hit this limit. You can use the ephemeral_storage_size configuration to increase the storage size of the AWS Lambda resource handler.

NOTE: a new AWS Lambda handler will be created in your stack for each combination of memory and storage size.

JSON-Aware Source Processing

When using Source.jsonData with CDK Tokens (references to construct properties), you may need to enable the escaping option. This is particularly important when the referenced properties might contain special characters that require proper JSON escaping (like double quotes, line breaks, etc.).

bucket = nil # AWSCDK::S3::Bucket
param = nil # AWSCDK::SSM::StringParameter


# Example with a secret value that contains double quotes
deployment = AWSCDK::S3Deployment::BucketDeployment.new(self, "JsonDeployment", {
    sources: [
        AWSCDK::S3Deployment::Source.json_data("config.json", {
            api_endpoint: "https://api.example.com",
            secret_value: param.string_value,
             # value with double quotes
            config: {
                enabled: true,
                features: ["feature1", "feature2"],
            },
        }, {escape: true}),
    ],
    destination_bucket: bucket,
})

EFS Support

If your workflow needs more disk space than default (512 MB) disk space, you may attach an EFS storage to underlying lambda function. To Enable EFS support set efs and vpc props for BucketDeployment.

Check sample usage below. Please note that creating VPC inline may cause stack deletion failures. It is shown as below for simplicity. To avoid such condition, keep your network infra (VPC) in a separate stack and pass as props.

destination_bucket = nil # AWSCDK::S3::Bucket
vpc = nil # AWSCDK::EC2::VPC


AWSCDK::S3Deployment::BucketDeployment.new(self, "DeployMeWithEfsStorage", {
    sources: [AWSCDK::S3Deployment::Source.asset(path.join(__dirname, "my-website"))],
    destination_bucket: destination_bucket,
    destination_key_prefix: "efs/",
    use_efs: true,
    vpc: vpc,
    retain_on_delete: false,
})

Data with deploy-time values

The content passed to Source.data(), Source.jsonData(), or Source.yamlData() can include references that will get resolved only during deployment. Only a subset of CloudFormation functions are supported however, namely: Ref, Fn::GetAtt, Fn::Join, and Fn::Select (Fn::Split may be nested under Fn::Select).

For example:

require 'aws-cdk-lib'

destination_bucket = nil # AWSCDK::S3::Bucket
topic = nil # AWSCDK::SNS::Topic
tg = nil # AWSCDK::ElasticLoadBalancingv2::ApplicationTargetGroup


app_config = {
    topic_arn: topic.topic_arn,
    base_url: "https://my-endpoint",
    lb_name: tg.first_load_balancer_full_name,
}

AWSCDK::S3Deployment::BucketDeployment.new(self, "BucketDeployment", {
    sources: [AWSCDK::S3Deployment::Source.json_data("config.json", app_config)],
    destination_bucket: destination_bucket,
})

The value in topic.topicArn is a deploy-time value. It only gets resolved during deployment by placing a marker in the generated source file and substituting it when its deployed to the destination with the actual value.

Substitutions from Templated Files

The DeployTimeSubstitutedFile construct allows you to specify substitutions to make from placeholders in a local file which will be resolved during deployment. This is especially useful in situations like creating an API from a spec file, where users might want to reference other CDK resources they have created.

The syntax for template variables is {{ variableName }} in your local file. Then, you would specify the substitutions in CDK like this:

require 'aws-cdk-lib'

my_lambda_function = nil # AWSCDK::Lambda::Function
destination_bucket = nil # AWSCDK::S3::Bucket
# (Optional) if provided, the resulting processed file would be uploaded to the destinationBucket under the destinationKey name.
destination_key = nil
role = nil # AWSCDK::IAM::Role


AWSCDK::S3Deployment::DeployTimeSubstitutedFile.new(self, "MyFile", {
    source: "my-file.yaml",
    destination_key: destination_key,
    destination_bucket: destination_bucket,
    substitutions: {
        variable_name: my_lambda_function.function_name,
    },
    role: role,
})

Nested variables, like {{ {{ foo }} }} or {{ foo {{ bar }} }}, are not supported by this construct. In the first case of a single variable being is double nested {{ {{ foo }} }}, only the {{ foo }} would be replaced by the substitution, and the extra brackets would remain in the file. In the second case of two nexted variables {{ foo {{ bar }} }}, only the {{ bar }} would be replaced in the file.

Keep Files Zipped

By default, files are zipped, then extracted into the destination bucket.

You can use the option extract: false to disable this behavior, in which case, files will remain in a zip file when deployed to S3. To reference the object keys, or filenames, which will be deployed to the bucket, you can use the object_keys getter on the bucket deployment.

require 'aws-cdk-lib'

destination_bucket = nil # AWSCDK::S3::Bucket


my_bucket_deployment = AWSCDK::S3Deployment::BucketDeployment.new(self, "DeployMeWithoutExtractingFilesOnDestination", {
    sources: [AWSCDK::S3Deployment::Source.asset(path.join(__dirname, "my-website"))],
    destination_bucket: destination_bucket,
    extract: false,
})

AWSCDK::CfnOutput.new(self, "ObjectKey", {
    value: AWSCDK::Fn.select(0, my_bucket_deployment.object_keys),
})

Controlling the Output of Source Object Keys

By default, the keys of the source objects copied to the destination bucket are returned in the Data property of the custom resource. However, you can disable this behavior by setting the outputObjectKeys property to false. This is particularly useful when the number of objects is too large and might exceed the size limit of the responseData property.

require 'aws-cdk-lib'

destination_bucket = nil # AWSCDK::S3::Bucket


my_bucket_deployment = AWSCDK::S3Deployment::BucketDeployment.new(self, "DeployMeWithoutExtractingFilesOnDestination", {
    sources: [AWSCDK::S3Deployment::Source.asset(path.join(__dirname, "my-website"))],
    destination_bucket: destination_bucket,
    output_object_keys: false,
})

AWSCDK::CfnOutput.new(self, "ObjectKey", {
    value: AWSCDK::Fn.select(0, my_bucket_deployment.object_keys),
})

Specifying a Custom VPC, Subnets, and Security Groups in BucketDeployment

By default, the AWS CDK BucketDeployment construct runs in a publicly accessible environment. However, for enhanced security and compliance, you may need to deploy your assets from within a VPC while restricting network access through custom subnets and security groups.

Using a Custom VPC

To deploy assets within a private network, specify the vpc property in BucketDeploymentProps. This ensures that the deployment Lambda function executes within your specified VPC.

vpc = AWSCDK::EC2::VPC.from_lookup(self, "ExistingVPC", {vpc_id: "vpc-12345678"})
bucket = AWSCDK::S3::Bucket.new(self, "MyBucket")

AWSCDK::S3Deployment::BucketDeployment.new(self, "DeployToS3", {
    destination_bucket: bucket,
    vpc: vpc,
    sources: [AWSCDK::S3Deployment::Source.asset("./website")],
})

Specifying Subnets for Deployment

By default, when you specify a VPC, the BucketDeployment function is deployed in the private subnets of that VPC. However, you can customize the subnet selection using the vpcSubnets property.

vpc = AWSCDK::EC2::VPC.from_lookup(self, "ExistingVPC", {vpc_id: "vpc-12345678"})
bucket = AWSCDK::S3::Bucket.new(self, "MyBucket")

AWSCDK::S3Deployment::BucketDeployment.new(self, "DeployToS3", {
    destination_bucket: bucket,
    vpc: vpc,
    vpc_subnets: {subnet_type: AWSCDK::EC2::SubnetType::PUBLIC},
    sources: [AWSCDK::S3Deployment::Source.asset("./website")],
})

Defining Custom Security Groups

For enhanced network security, you can now specify custom security groups in BucketDeploymentProps. This allows fine-grained control over ingress and egress rules for the deployment Lambda function.

vpc = AWSCDK::EC2::VPC.from_lookup(self, "ExistingVPC", {vpc_id: "vpc-12345678"})
bucket = AWSCDK::S3::Bucket.new(self, "MyBucket")

security_group = AWSCDK::EC2::SecurityGroup.new(self, "CustomSG", {
    vpc: vpc,
    description: "Allow HTTPS outbound access",
    allow_all_outbound: false,
})

security_group.add_egress_rule(AWSCDK::EC2::Peer.any_ipv4, AWSCDK::EC2::Port.tcp(443), "Allow HTTPS traffic")

AWSCDK::S3Deployment::BucketDeployment.new(self, "DeployWithSecurityGroup", {
    destination_bucket: bucket,
    vpc: vpc,
    security_groups: [security_group],
    sources: [AWSCDK::S3Deployment::Source.asset("./website")],
})

Notes

Development

The custom resource is implemented in Python 3.9 in order to be able to leverage the AWS CLI for "aws s3 sync". The code is now in the @aws-cdk/custom-resource-handlers package under lib/aws-s3-deployment/bucket-deployment-handler and unit tests are under test/aws-s3-deployment/bucket-deployment-handler.

This package requires Python 3.9 during build time in order to create the custom resource Lambda bundle and test it. It also relies on a few bash scripts, so might be tricky to build on Windows.

Roadmap

API Reference

Classes 4

BucketDeployment`BucketDeployment` populates an S3 bucket with the contents of .zip files from other S3 bu CacheControlUsed for HTTP cache-control header, which influences downstream caches. DeployTimeSubstitutedFile`DeployTimeSubstitutedFile` is an extension of `BucketDeployment` that allows users to upl SourceSpecifies bucket deployment source.

Interfaces 8

BucketDeploymentPropsProperties for `BucketDeployment`. DeploymentSourceContextBind context for ISources. DeployTimeSubstitutedFileProps ISourceRepresents a source for bucket deployments. JsonProcessingOptionsDefine options which can be passed using the method `Source.jsonData()`. MarkersConfigA configuration for markers substitution strategy. SourceConfigSource information. UserDefinedObjectMetadataCustom user defined metadata.

Enums 2

ServerSideEncryptionIndicates whether server-side encryption is enabled for the object, and whether that encry StorageClassStorage class used for storing the object.