Troubleshooting common AWS CDK issues

This topic describes how to troubleshoot the following issues with the AWS CDK.

After updating the AWS CDK, code that used to work fine now results in errors

Errors in code that used to work is typically a symptom of having mismatched versions of AWS Construct Library modules. Make sure all library modules are the same version and up-to-date.

An error message commonly seen in this situation is unable to determine cloud assembly output directory. Assets must be defined indirectly within a "Stage" or an "App" scope.

The modules that make up the AWS Construct Library are a matched set. They are released together and are intended to be used together. Interfaces between modules are considered private; we may change them when necessary to implement new features in the library.

We also update the libraries that are used by the AWS Construct Library from time to time, and different versions of the library modules may have incompatible dependencies. Synchronizing the versions of the library modules will also address this issue.

JSII is an important AWS CDK dependency, especially if you are using the AWS CDK in a language other than TypeScript or JavaScript. You do not ordinarily have to concern yourself with the JSII versions, since it is a declared dependency of all AWS CDK modules. If a compatible version is not installed, however, you can see unexpected type-related errors, such as 'undefined' is not a valid TargetType. Making sure all AWS CDK modules are the same version will resolve JSII compatibility issues, since they will all depend on the same JSII version.

Below, you'll find details on managing the versions of your installed AWS Construct Library modules in TypeScript, JavaScript, Python, Java, and C#.

(back to list)

After updating the AWS CDK, the AWS CDK Toolkit (CLI) reports a mismatch with the AWS Construct Library

The version of the AWS CDK Toolkit (which provides the cdk command) must be at least equal to the version of the AWS Construct Library. The Toolkit is intended to be backward compatible within the same major version; the latest 1.x version of the toolkit can be used with any 1.x release of the library. For this reason, we recommend you install this component globally and keep it up-to-date.

npm update -g aws-cdk

If, for some reason, you need to work with multiple versions of the AWS CDK Toolkit, you can install a specific version of the toolkit locally in your project folder.

If you are using a language other than TypeScript or JavaScript, first create a node_modules folder in your project directory. Then, regardless of language, use npm to install the AWS CDK Toolkit, omitting the -g flag and specifying the desired version. For example:

npm install aws-cdk@1.50.0

To run a locally-installed AWS CDK Toolkit, use the command npx aws-cdk rather than just cdk. For example:

npx aws-cdk deploy MyStack

npx aws-cdk runs the local version of the AWS CDK Toolkit if one exists, and falls back to the global version when a project doesn't have a local installation. You may find it convenient to set up a shell alias or batch file to make sure cdk is always invoked this way. For example, Linux users might add the following statement to their .bash_profile file.

alias cdk="npx aws-cdk"

(back to list)

When deploying my AWS CDK stack, I receive a NoSuchBucket error

Your AWS environment does not have a staging bucket, which the AWS CDK uses to hold resources during deployment. Stacks require this bucket if they contain Assets or synthesize to AWS CloudFormation templates larger than 50 kilobytes. You can create the staging bucket with the following command:

cdk bootstrap aws://ACCOUNT-NUMBER/REGION

To avoid generating unexpected AWS charges, the AWS CDK does not automatically create a staging bucket. You must bootstrap your environment explicitly.

By default, the staging bucket is created in the region(s) used by stacks in the current AWS CDK application, or the region specified in your local AWS profile (set by aws configure), using that profile's account. You can specify a different account and region on the command line as follows. (You must specify the account and region if you are not in an app's directory.)

cdk bootstrap aws://ACCOUNT-NUMBER/REGION

You must bootstrap in every region where you will deploy stacks that require a staging bucket.

For more information, see Bootstrapping

(back to list)

When deploying my AWS CDK stack, I receive a forbidden: null message

You are deploying a stack that requires the use of a staging bucket, but are using an IAM role or account that lacks permission to write to it. (The staging bucket is used when deploying stacks that contain assets or that synthesize an AWS CloudFormation template larger than 50K.) Use an account or role that has permission to perform the action s3:* against the resource arn:aws:s3:::cdktoolkit-stagingbucket-*.

(back to list)

When synthesizing an AWS CDK stack, I get the message --app is required either in command-line, in cdk.json or in ~/.cdk.json

This message usually means that you aren't in the main directory of your AWS CDK project when you issue cdk synth. The file cdk.json in this directory, created by the cdk init command, contains the command line needed to run (and thereby synthesize) your AWS CDK app. For a TypeScript app, for example, the default cdk.json looks something like this:

{
  "app": "npx ts-node bin/my-cdk-app.ts"
}

We recommend issuing cdk commands only in your project's main directory, so the AWS CDK toolkit can find cdk.json there and successfully run your app.

If this isn't practical for some reason, the AWS CDK Toolkit looks for the app's command line in two other locations:

For example, you might synthesize a stack from a TypeScript app as follows.

cdk synth --app "npx ts-node my-cdk-app.ts" MyStack

(back to list)

When synthesizing an AWS CDK stack, I receive an error because the AWS CloudFormation template contains too many resources

The AWS CDK generates and deploys AWS CloudFormation templates. AWS CloudFormation has a hard limit on the number of resources a stack can contain. With the AWS CDK, you can run up against this limit more quickly than you might expect.

The AWS Construct Library's higher-level, intent-based constructs automatically provision any auxiliary resources that are needed for logging, key management, authorization, and other purposes. For example, granting one resource access to another generates any IAM objects needed for the relevant services to communicate.

In our experience, real-world use of intent-based constructs results in 1–5 AWS CloudFormation resources per construct, though this can vary. For serverless applications, 5–8 AWS resources per API endpoint is typical.

Patterns, which represent a higher level of abstraction, let you define even more AWS resources with even less code. The AWS CDK code in Creating an AWS Fargate service using the AWS CDK, for example, generates more than fifty AWS CloudFormation resources while defining only three constructs!

Exceeding the AWS CloudFormation resource limit is an error during AWS CloudFormation synthesis. The AWS CDK issues a warning if your stack exceeds 80% of the limit. You can use a different limit by setting the maxResources property on your stack, or disable validation by setting maxResources to 0.

// rescount.js - count the resources defined in a stack
// invoke with: node rescount.js <path-to-stack-json>
// e.g. node rescount.js cdk.out/MyStack.template.json
  
import * as fs from 'fs';
const path = process.argv[2];
  
if (path) fs.readFile(path, 'utf8', function(err, contents) {
  console.log(err ? `${err}` : 
  `${Object.keys(JSON.parse(contents).Resources).length} resources defined in ${path}`);
}); else console.log("Please specify the path to the stack's output .json file");

As your stack's resource count approaches the limit, consider re-architecting to reduce the number of resources your stack contains: for example, by combining some Lambda functions, or by breaking your stack into multiple stacks. The CDK supports references between stacks, so it is straightforward to separate your app's functionality into different stacks in whatever way makes the most sense to you.

(back to list)

I specified three (or more) Availability Zones for my EC2 Auto-Scaling Group or Virtual Private Cloud, but it was only deployed in two

To get the number of Availability Zones you requested, specify the account and region in the stack's env property. If you do not specify both, the AWS CDK, by default, synthesizes the stack as environment-agnostic, such that it can be deployed to any region. You can then deploy the stack to a specific region using AWS CloudFormation. Because some regions have only two availability zones, an environment-agnostic template never uses more than two.

You can change this behavior by overriding your stack's availablilityZones (Python: availability_zones) property to explicitly specify the zones you want to use.

For more information about specifying a stack's account and region at synthesis time, while retaining the flexibility to deploy to any region, see Environments.

(back to list)

My S3 bucket, DynamoDB table, or other resource is not deleted when I issue cdk destroy

By default, resources that can contain user data have a removalPolicy (Python: removal_policy) property of RETAIN, and the resource is not deleted when the stack is destroyed. Instead, the resource is orphaned from the stack. You must then delete the resource manually after the stack is destroyed. Until you do, redeploying the stack fails, because the name of the new resource being created during deployment conflicts with the name of the orphaned resource.

If you set a resource's removal policy to DESTROY, that resource will be deleted when the stack is destroyed.

(back to list)