** Please open a PR for any issue you open in which you suggest a patch - I do not have time to maintain this by myself anymore, and it generally works as advertised so any improvements requested are incumbent upon the requester to help implement, thanks / with love <3 **
A simple flow for generating CloudFormation Lambda-Backed Custom Resource handlers in node.js. The scope of this module is to structure the way developers author simple Lambda-Backed resources into simple functional definitions of Create
, Update
, Delete
.
Also supports:
- Extremely simple deployments
- Automatic creation of CloudFormation Quick Launch links, so you can easily share your open-source Custom Resources in any region!
- Automatic expansion of
__default__
Properties
values into any tree or subtree of any Custom Resource utilizingcfn-lambda
for implementation - Validation of
'ResourceProperties'
- Using inline JSONSchema objects as
Schema
- Using a
SchemaPath
to JSONSchema file - Using a custom
Validate
callback
- Using inline JSONSchema objects as
- Optional
NoUpdate
callback, which runs as a READ function for whenUpdate
should be made due to all parameters being identical - because some resources still need to return attributes forFn::GetAtt
calls. - Convenience
Environment
values- Lambda ARN
- Lambda Name
- AWS Account ID for the Lambda
- Region for the Lambda
- Array of String
TriggersReplacement
forResource.Properties
key strings that force delegation to resourceCreate
for seamless full replacement without downtime in many cases, and forcingUPDATE_COMPLETE_CLEANUP_IN_PROGRESS
. - An
SDKAlias
function generator that structures and greatly simplifies the development of custom resources that are supported by the Node.jsaws-sdk
but not supported by CloudFormation.
This package on NPM
This package on GitHub
Since version 2.0.0 of this tool, this supports Launch Pages, which are a simple way to share your resources. These are some you can try.
- Add Amazon Lex Slot Types as a supported CloudFormation resource here
- More added soon! Just click a Launch link to install into your account.
Once you build a resource with this tool, if you use the --public
setting, you can share these resources by sharing your Launch Pages. These are HTML pages hosted in S3 that the tool automatically creates during deployments. They're only accessible to the public if you specifically set --public
during a deployment. You should only do this for open source custom resource types.
These pages are accessible in all regions in which you deploy resource types to. They generally follow this link pattern: <s3 host>/<your package name>-<your AWS account ID>-<region>/<dash-delimited-version>.html
.
If you're confused, check for the HTML pages that are inserted into the S3 buckets deployed by this tool during deploys.
The "old" resources below work, they just are now supported by built-in CloudFormation resource types. They still are good examples of how to implement.
- Stable
Custom::LexSlotType
(GitHub / NPM) - Stable, old
Custom::ApiGatewayRestApi
(GitHub / NPM) - Stable, old
Custom::ApiGatewayMethod
(GitHub / NPM) - Stable, old
Custom::ApiGatewayMethodResponse
(GitHub / NPM) - Stable, old, uses LongRunning configurations correctly
Custom::ElasticSearchServiceDomain
(GitHub / NPM)
Hey you! Are you an AWS automation engineer? I'd love if you'd author open-source resources with this tool. Just submit a PR to this page for a specific tag on your repository, and I'll review it and add it to the page.
Furthermore, if you want to help style the generated HTML in the launcher pages, I'd love help with that too :)
Feel free to tweet me about involvement too: @ayetempleton thanks!
Any custom resource using this tool as a dependency can run deploy scripts from the root of the custom resource project to deploy Lambdas to all regions.
To do this most simply, add this line to the "scripts"
section of your package.json
inside your repository using this module as a direct dependency:
"deploy": "node ./node_modules/cfn-lambda/deploy.js --allregions --logs"
This will deploy your custom resource to all regions. If you want to customize this behavior, use the options below. These options also apply to using the deploy.js
script, as well.
You can also deploy the Lambdas programmatically from JS by importing the module: require('cfn-lambda')
. The same options that work on the command line below work as values on an option hash: require('cfn-lambda')(options, callback)
.
You must also set up:
- Add
<reporoot>/execution-policy.json
to define the abilities the Lambda should have. - Have AWS credentials configured in your environment, via one of:
$AWS_PROFILE
in your environment- a credentials file
$AWS_ACCESS_KEY_ID
and$AWS_SECRET_ACCESS_KEY
in your environment.
You then run this from within the repository directly depending on cfn-lambda
(your custom resource implementation using this package):
$ npm run deploy
Again, this will, if you used the suggested package.json
edit to use the deploy.js
inside this repo, deploy your custom resource implementation to all regions using some default settings. Please read the below options, as you may want to restrict deployment to only a couple regions. You might want to do this if your custom resource uses AWS services only available in a smaller subset of regions than Lambda is available in.
Since this uses CloudFormation to install, you can get the ServiceToken usable in the region to begin creating resources from the Outputs and Exports of the generated stack at the ServiceToken
key. The stack by default launches with name <your-package-name>-<your package version>
. You can manually grab the value from the Outputs
in that template, and use it later, or you can use Fn::ImportValue
to directly get it in any stack.
"Fn::ImportValue": "<resource name>-<resource version>-ServiceToken"
This value may change if you use the --alias
or --version
flags below, since the Stack name will be different.
When using the module in JS, a simple hash is passed in as the first argument. When on the command line, boolean parameters are set to true
with --<argname>
. Parameters needing a value are set with --<argname> <argvalue>
.
Used to specify the AWS Account ID to launch the systems into. By default, is the account associated with the Role or User currently invoking the script. Useful when a cross-account role is being used.
Instead of deploying the systems with this naming pattern: <your package name>-<your package version>
, it replaces the package name: <specified alias>-<your package version>
.
Causes your custom resource to be deployed on all regions supporting AWS Lambda. false
by default.
Makes the deployment system log to STDOUT. Defaults to false when used via JS as a module, and defaults to true
on CLI. You can turn CLI logging off with the --quiet
option.
Sets the tool to deploy the custom service in the module you provide, relative to the current working directory. For example, if this package, and your package using cfn-lambda
are both dependencies of a project, from that project's root, set --module <custom resource to deploy>
. The --path
argument takes precedence if both --module
and --path
are defined.
The tool will deploy the cfn-lambda
-based custom resource you have defined to use the provided path. Best used when cfn-lambda
is not in your project's node_modules
directory. If you do not provide --path
or --module
, the system assumes that cfn-lambda
is in the node_modules
directory of your project, and thus uses this --path
: cfn-lambda/../../
(assumes you're using cfn-lambda
as a normal node_modules
dependency of your custom resource project directory).
Sets the Quick Launch Page to be publicly accessible, as well as the Lambda function code zip bundle for your custom resource type. Defaults to false
.
Setting this to true
if you are an open source software author will allow anyone to install your custom resources without needing to run this installation script, just by clicking a link on the browser, in the HTML page this tool generates for you in your S3 buckets in each region you have this script run on. Bear in mind, that you will be responsible for the AWS fees associated with others accessing your bucket.
Forces logs off, on both CLI and with JS module usage. Takes precedence over --logs
, so use this to make your CLI invocations stop producing logs. Defaults to false
.
Sets the AWS regions to deploy your custom resource type to. Defaults to the value of $AWS_REGION
in your environment, or none, if you do not set $AWS_REGION
.
This value is ignored if you set --allregions
.
On the CLI, values are passed in comma-delimited, with no spaces, like us-east-1,us-east-2
. When using the module via JS, pass this value in as a plain JavaScript array.
Setting this value to false
prevents the CloudFormation stacks this tool uses to deploy from rolling back when any errors occur during initial creation. On the CLI, the value must be exactly false
. With module-style usage in JS, any falsey value will achieve the same effect. Defaults to true
, thus allowing any stacks with failures during creation to roll back.
Instead of deploying the systems with this naming pattern: <your package name>-<your package version>
, it replaces the package version: <your package name>-<specified version>
. Technically, it does not have to be a number, but using the format x-y-z
is strongly suggested.
This is a contrived example call to fully demonstrate the way to interface with the creation API.
You can manually define these properties, or use SDKAlias
for Create
, Update
and/or Delete
.
var CfnLambda = require('cfn-lambda');
exports.handler = CfnLambda({
Create: Create, // Required function
Update: Update, // Required function
Delete: Delete, // Required function
// Any of following to validate resource Properties
// If you do not include any, the Lambda assumes any Properties are valid.
// If you define more than one, the system uses all of them in this order.
Validate: Validate, // Function
Schema: Schema, // JSONSchema v4 Object
SchemaPath: SchemaPath, // Array path to JSONSchema v4 JSON file
// end list
NoUpdate: NoUpdate, // Optional
TriggersReplacement: TriggersReplacement, // Array<String> of properties forcing Replacement
LongRunning: <see Long Running below> // Optional. Configure a lambda to last beyond 5 minutes.
});
Provides convenience Environment
values.:
var CfnLambda = require('cfn-lambda');
// After receiving `event` and `context`...
console.log(CfnLambda.Environment);
/*
{
`LambdaArn`: 'foo bar', // Full ARN for the current Lambda
`Region`: 'us-east-1', // Region in which current Lambda resides
`AccountId`: '012345678910', // The account associated with the Lambda
`LambdaName`: 'LambdaName' // Name for the current Lambda
}
*/
Only works after the generated CfnLambda
function has been called by Lambda.
Called when CloudFormation issues a 'CREATE'
command.
Accepts the CfnRequestParams
Properties object, and the reply
callback.
function Create(CfnRequestParams, reply) {
// code...
if (err) {
// Will fail the create.
// err should be informative for Cfn template developer.
return reply(err);
}
// Will pass the create.
// physicalResourceId defaults to the request's `[StackId, LogicalResourceId, RequestId].join('/')`.
// FnGetAttrsDataObj is optional.
reply(null, physicalResourceId, FnGetAttrsDataObj);
}
Called when CloudFormation issues an 'UPDATE'
command.
Accepts the RequestPhysicalId
String
, CfnRequestParams
Properties object, the OldCfnRequestParams
Properties object, and the reply
callback.
function Update(RequestPhysicalID, CfnRequestParams, OldCfnRequestParams, reply) {
// code...
if (err) {
// Will fail the update.
// err should be informative for Cfn template developer.
return reply(err);
}
// Will pass the update.
// physicalResourceId defaults to pre-update value.
// FnGetAttrsDataObj is optional.
reply(null, physicalResourceId, FnGetAttrsDataObj);
}
Called when CloudFormation issues a 'DELETE'
command.
Accepts the RequestPhysicalId
String
, CfnRequestParams
Properties object, and the reply
callback.
function Delete(RequestPhysicalID, CfnRequestParams, reply) {
// code...
if (err) {
// Will fail the delete (or rollback).
// USE CAUTION - failing aggressively will lock template,
// because DELETE is used during ROLLBACK phases.
// err should be informative for Cfn template developer.
return reply(err);
}
// Will pass the delete.
// physicalResourceId defaults to pre-delete value.
// FnGetAttrsDataObj is optional.
reply(null, physicalResourceId, FnGetAttrsDataObj);
}
If your handler function is async, then your custom lambda resource may unexpectedly exit before your Promise(s) have resolved. The result of this is your CloudFormation stack hanging for up to an hour which is very undesirable behavior. You can read more about the underlying issue in this great medium article. To fix this, you can use the async counterparts of the given handler type:
Create -> AsyncCreate
Update -> AsyncUpdate
Delete -> AsyncDelete
NoUpdate -> AsyncNoUpdate
This difference with the async counterparts is that there is no reply() callback. Instead, your function should return a Promise (or be marked async
) which resolves with an object containing the PhysicalResourceId
and FnGetAttrsDataObj
outputs. See below example:
const wait = () => {
return new Promise((resolve, reject) => {
setTimeout(() => {
resolve();
}, 2000);
});
};
const createHandler = async (cfnRequestParams) => {
await wait();
return {
PhysicalResourceId: "yopadope",
FnGetAttrsDataObj: {
MyObj: "dopeayope"
}
};
};
exports.handler = cfnLambda({
AsyncCreate: createHandler,
...
});
Used before 'CREATE'
, 'UPDATE'
, or 'DELETE'
method handlers. The CloudFormation request will automatically fail if any truthy values are returned, and any String
values returned are displayed to the template developer, to assist with resource Properties
object correction.
Important: To prevent ROLLBACK
lockage, the 'DELETE'
will be short circuited if this check fails. If this check fails, CloudFormation will be told that everything went fine, but no actual further actions will occur. This is because CloudFormation will immediately issue a 'DELETE'
after a failure in a 'CREATE'
or an 'UPDATE'
. Since these failures themselves will have resulted from a validation method failure if the subsequent 'DELETE'
fails, this is safe.
May be a:
- Custom validation function as
Validate
callback - JSONSchema v4
Schema
- JSONSchema v4 file path as
SchemaPath
The truthy String
return value will cause a 'FAILURE'
, and the String
value is used as the CloudFormation 'REASON'
.
// Example using a custom function
// CfnRequestParams are all resource `Properties`,
// except for the required system `ServiceToken`.
function Validate(CfnRequestParams) {
// code...
if (unmetParamCondition) {
return 'You must blah blah include a parameter... etc'
}
if (someOtherCondition) {
return 'Informative message to CFN template developer goes here.'
}
// Returning a falsey value will allow the action to proceed.
// DO NOT return truthy if the request params are valid.
}
Using a JSONSchema Schema
property value will automatically generate the String
invalidation return values for you when validating against the parameters - simply provide the template and the validation and error messging is taken care of for you.
If you choose to use a JSONSchema template, the service will also use the JSONSchema metaschema to ensure the provided JSONSchema is a valid schema itself.
// Example using a custom JSONSchema Version 4 template
// This might be in a file you manually load like `schema.json`, or a JS object.
var Schema = {
type: 'object',
required: [
'foo'
],
properties: {
foo: {
type: 'string'
},
selectable: {
type: 'string',
enum: ['list', 'of', 'valid', 'values']
}
},
additionalProperties: false
};
A convenient way to get the benefits of Schema
object validation, but keeping your code clean and segregated nicely.
The path is defined as an Array so that we can use the path
module.
var SchemaPath = [__dirname, 'src', 'mytemplate.json'];
Optional. Triggered by deep JSON object equality of the old and new parameters, if defined.
Even when short-circuiting an Update
is a good idea, a resource provider may still need to return a set of properties for use with Fn::GetAtt
in CloudFormation templates. This NoUpdate
handler triggers in the special case where no settings on the resource change, allowing the developer to simultaneously skip manipulation logic while doing read operations on resources to generate the attribute sets Fn::GetAtt
will need.
// Using a custom NoUpdate for READ to supply properties
// for Fn::GetAtt to access in CloudFormation templates
function NoUpdate(PhysicalResourceId, CfnResourceProperties, reply) {
// code that should be read-only if you're sane...
if (errorAccessingInformation) {
return reply('with an informative message');
}
// Can have many keys on the object, though I only show one here
reply(null, PhysicalResourceId, {Accessible: 'Attrs object in CFN template'});
}
This is very advanced Lambda self replication.
The inner workings of this feature are a lot to take in. I strongly suggest you just read the source code for cfn-elasticsearch-domain
to see how the index.js
file utilizes the LongRunning
feature, as the concrete example code is much more understandable than abstract definitions of parameters and options.
cfn-elasticsearch-domain/index.js
GitHub
If you have the appetite for it... Read on...
Some resources will take a considerable amount of time to complete, like an Elasticsearch Domain. In order to utilize Lambda-Backed Custom Resources within CloudFormation while avoiding the hard 300 second / 5 minute Lambda timeout for resources that will take more than 5 minutes to finish, cfn-lambda
allows resource developers to leverage bundled Lambda self-replication logic. Developers can configure the LongRunning
property on the lambda definition options object with a few settings to tell cfn-lambda
to simply run some action initialization code (such as initiating an Elasticsearch Domain Create), then periodically self-replicate to check the status of the long-running process. The majority of cases where AWS APIs or SDKs return statusCode === 202
will use this technique to avoid Lambda death at 5 minutes.
The self-replication strategy will trigger if the developer configures the following on the LongRunning property object: PingInSeconds
, MaxPings
, LambdaApi
, Methods.METHOD_NAME
.
The duration a Lambda will wait between spawning self-replication calls and triggering the next LongRunning.Methods.METHOD_NAME
call. This value should not exceed 240
(4 minutes), because we need to leave enough time before the 5 minute hard process death is triggered by AWS.
After this time, the lambda will spawn a new lambda, which will call the LongRunning.Method.METHOD_NAME
, where METHOD_NAME
is Create
, Delete
, Update
, depending on which are configured and the lifecycle phase the resource is moving through.
The maximum number of self-respawn and check cycles the Lambda will go through. After exceeding this number, the Lambda will circuit break and send a Failed to Stabilize
response to the CloudFormation stack.
cfn-lambda
uses this namespace to invoke the Lambda. Allows the Custom Resource developer using cfn-lambda
to specify a Lambda API version, or stub the value out for testing.
In most cases, just pass new AWS.Lambda({apiVersion: '2015-03-31'})
as the API namespace.
Most of the LongRunning
logic happens here. At its most configured, this subobject will have 3 properties corresponding to the normal actions: Create
, Update
, and Delete
.
When you configure one of these properties, the flow of that CloudFormation action type changes - within the reply
callback function in the corresponding normal/top-level callback you defined for the resource, reply
-ing with success just tells cfn-lambda
that you correctly initialized the Create
/Delete
/Update
for the resource, and to start using the corresponding LongRunning.Methods.METHOD
to ping to final completion. That is, the resource will not COMPLETE
the action until the function you define finalizes the SUCCESS
.
Read below to see how to define each LongRunning.Methods.METHOD
...
All three LongRunning.Methods
receive a special object as their first parameter. The LongRunningContext
object carries useful state across all spawned lambda ping cycles.
LongRunningContext.RawResponse
: carries the original intercepted call that your first Create initialization call tried to send to CloudFormation. Used internally for state manipulation. DO NOT ALTER THIS VALUE unless you really know what you're doing, as tampering can cause Lambda recursion to spiral out of control!LongRunningContext.PhysicalResourceId
: Carries the originalPhysicalResourceId
intercepted call that your first Create initialization call tried to send to CloudFormation. Useful if your check functions need this value and cannot recompute it from theResourceProperties
sent by CloudFormation.LongRunningContext.Data
: Carries the original data hash, if present, intercepted from your call toreply
within the initializer method. Useful if your check functions need these data value(s) and cannot recompute them from theResourceProperties
sent by CloudFormation. Will not be present if you did not pass a third parameter toreply
in the initializer, since theGetAtt
-usableData
hash is optional incfn-lambda
.LongRunningContext.PassedPings
: The number of ping spawns before this current run that have occurred. DO NOT ALTER THIS NUMBER! Subtracting from this number will make your Lambdas infinitely self-replicate, very very bad!
Will be called during Lambda pingspawn cycles. Here, CheckCreate
is an example of a check function definition for LongRunning.Methods.Create
.
function CheckCreate(LongRunningContext, params, reply, notDone) {
// LongRunningContext is object type specified above
// params are Properties straight from CloudFomation
// reply is callback just like in normal Create,
// call it with reply(errMsg) or reply(null, physicalId, AttrHash)
// notDone takes no parameters, use this to tell
// cfn-lambda to use another ping/spawn cycle and check again later
}
Will be called during Lambda pingspawn cycles. Here, CheckUpdate
is an example of a check function definition for LongRunning.Methods.Update
.
function CheckUpdate(LongRunningContext, physcialId, params, oldParams, reply, notDone) {
// LongRunningContext is object type specified above
// physicalId is PhysicalResourceId from pre-Update resource state
// params are Properties straight from CloudFomation
// oldParams are Properties from CloudFormation for before the Update began
// reply is callback just like in normal Update,
// call it with reply(errMsg) or reply(null, physicalId, AttrHash)
// to finalize the transition and notify CloudFormation.
// notDone takes no parameters, use this to denote no errors and tell
// cfn-lambda to use another ping/spawn cycle and check again later
}
Will be called during Lambda pingspawn cycles. Here, CheckDelete
is an example of a check function definition for LongRunning.Methods.Delete
.
function CheckDelete(LongRunningContext, physcialId, params, reply, notDone) {
// LongRunningContext is object type specified above
// physicalId is PhysicalResourceId from pre-Delete resource state
// params are Properties straight from CloudFomation
// reply is callback just like in normal Delete,
// call it with reply(errMsg) or reply(null, physicalId, AttrHash)
// to finalize the transition and notify CloudFormation.
// notDone takes no parameters, use this to denote no errors and tell
// cfn-lambda to use another ping/spawn cycle and check again later
}
Optional. Tells cfn-lambda
to divert the 'Update'
call from CloudFormation to the Create
handler the developer assigns to the Lambda. This technique results in the most seamless resource replacement possible, by causing the new resource to be created before the old one is deleted. This Delete
cleanup process occurs in the UPDATE_COMPLETE_CLEANUP_IN_PROGRESS
phase after all new resources are created. This property facilitates triggering that said phase.
exports.handler = CfnLambda({
// other properties
TriggersReplacement: ['Foo', 'Bar'],
// other properties
});
Now, if the Lambda above ever detects a change in the value of Foo
or Bar
resource Properties on Update
, the Lambda will delegate to a two-phase Create
-new-then-Delete
-old resource replacement cycle. It will use the Create
handler provided to the same CfnLambda
, then subsequently the prodvided Delete
if and only if the Create
handler sends a PhysicalResourceId
different from the original to the reply
callback in the handler.
Structures and accelerates development of resources supported by the aws-sdk
(or your custom SDK) by offering declarative tools to ingest events and proxy them to AWS services.
Will automatically correctly ignore ServiceToken
from CloudFormation Properties. All settings are optional, except api
and method
.
var AWS = require('aws-sdk');
var AnAWSApi = new AWS.SomeNamespace();
var CfnLambda = require('cfn-lambda');
// Then used as the Create property as defined in Usage above
var MyAliasActionName = CfnLambda.SDKAlias({ // Like Create, Update, Delete
returnPhysicalId: 'KeyFromSDKReturn' || function(data) { return 'customValue'; },
downcase: boolean, // Downcase first letter of all top-level params from CloudFormation
api: AnAWSApi, // REQUIRED
method: 'methodNameInSDK', // REQUIRED
mapKeys: {
KeyNameInCfn: 'KeyNameForSDK'
},
forceBools: [ // CloudFormation doesn't allow Lambdas to recieve true booleans.
'PathToCfnPropertyParam', // This will coerce the parameter at this path.
'Also.Supports.Wildcards.*',
'But',
'only.at.path.end'
],
keys: [ // Defaults to including ALL keys from CloudFormation, minus ServiceToken
'KeysFrom',
'CloudFormationProperties',
'ToPassTo',
'TheSDKMethod',
'**UsedBeforeMapKeys**'
],
returnAttrs: [
'KeysFrom',
'SDKReturnValue',
'ToUseWithCfn',
'Fn::GetAttr',
'You.Can.Access.Nested.Properties.As.Well'
],
ignoreErrorCodes: [IntegerCodeToIgnore, ExWouldBe404ForDeleteOps],
physicalIdAs: 'UsePhysicalIdAsThisKeyInSDKCall',
// physicalIdAs: 'OrUseNested.Property.Using.Dot.Notation'
});
// Then...
exports.handler = CfnLambda({
Create: MyAliasActionName, // Doesn't have to be Create, can be Update or Delete
// ...
});
Sometimes it is advantageous to be able to reuse JSON objects or fragments of JSON objects in Properties
of Custom Resources, like when you need to build similar complex/large resources frequently that differ by only a few properties.
Any module using cfn-lambda
supports __default__
property expansion. __default__
can be added anywhere in the Properties
object for a resource, with __default__
containing an arbitrary JSON/String/Array/null/Number
value serialized using toBase64(JSON.stringify(anyObject))
. cfn-lambda
will expand these properties before hitting any validation checks, by running JSON.parse(fromBase64(encodedDefault))
recursively, and overwriting any values in the __default__
tree with those actually set on the Properties
object.
The best example of this is the cfn-variable
module's example.template.json
, wherein a very large RestApi
is created with over a large repeated subtree of Resource
objects. cfn-variable
is a custom resource that takes any value and serializes it using toBase64(JSON.stringify(anyValue))
, making it a perfect fit for this behavior.
In the example in cfn-variable
, this technique is used to create 120 Resource
objects in under 15 seconds (this example uses less):
// This is cfn-variable storing the serialized object:
"MySubtreeVariable": {
"Type": "Custom::Variable",
"Properties": {
"ServiceToken": {
"Fn::Join": [
":",
[
"arn",
"aws",
"lambda",
{
"Ref": "AWS::Region"
},
{
"Ref": "AWS::AccountId"
},
"function",
{
"Ref": "VariableCustomResourceName"
}
]
]
},
"VariableValue": {
"ChildResources": [
{
"PathPart": "a",
"ChildResources": [
{
"PathPart": "aa",
"ChildResources": [
{
"PathPart": "aaa"
},
{
"PathPart": "aab"
},
{
"PathPart": "aac"
}
]
},
{
"PathPart": "ab",
"ChildResources": [
{
"PathPart": "aba"
},
{
"PathPart": "abb"
},
{
"PathPart": "abc"
}
]
},
{
"PathPart": "ac",
"ChildResources": [
{
"PathPart": "aca"
},
{
"PathPart": "acb"
},
{
"PathPart": "acc"
}
]
}
]
}
]
}
}
},
// Then this will make the tree 3x because you used a variable with __default__
"ExpandedResourceTree": {
"DependsOn": [
"MyRestApi",
"MyVariable"
],
"Type": "Custom::ApiGatewayResourceTree",
"Properties": {
"ServiceToken": "<the token>",
"RestApiId": {
"Ref": "MyRestApi"
},
"ParentId": {
"Fn::GetAtt": [
"MyRestApi",
"RootResourceId"
]
},
"ChildResources": [
{
"PathPart": "alpha",
"__default__": {
"Fn::GetAtt": [
"MySubtreeVariable",
"Value"
]
}
},
{
"PathPart": "beta",
"__default__": {
"Fn::GetAtt": [
"MySubtreeVariable",
"Value"
]
}
},
{
"PathPart": "gamma",
"__default__": {
"Fn::GetAtt": [
"MySubtreeVariable",
"Value"
]
}
}
]
}
}