When to use

Common use cases for custom resources:

1. Provisioning non-AWS resources - While AWS offers huge amount of services and resources, in some cases, AWS might not offer a solution(resource) for your use-case, and often time a third-party solution better fits your needs. For example: using Auth0 over Cognito or using Algolia over AWS Elastisearch. These non-AWS resources can be provisioned using custom resources.

2. Performing provisioning steps not related to infrastructure - Deployment of infrastructure is only a part of having the application successfully running. You might need to seed the database or run migrations, both of which are a great fit for a custom resource. Another example might be running a smoke test from your custom resource to ensure that everything is running correctly after deployment.

3. Any other use case - Since you are fully in control of the code and therefore of what the custom resource does, there is virtually no limit of how to use custom resources. Custom resources can be powerful, but we advise users to be reasonable with them.

Usage

In Stacktape, custom resources are made of two parts:

1. custom-resource-definition - definition is special type of lambda function. It contains the custom logic - the code, which gets executed each time custom-resource-instance is created, updated, or deleted. To put into perspective of standard programming languages: custom-resource-definition is like a class and custom-resource-instance is instance of this class.

2. custom-resource-instance - when you specify instance, you must specify custom-resource-definition defintionName (which you want to instantiate) and resourceProperties(which can be thought of as arguments), which will be passed to the executing code.

All parts of a custom resource are shown in the following snippet. Individual parts are explained in the subsequent sections.

The following example shows a custom resource that is used to seed a mongo cluster.

resources:
  myMongoCluster:
    type: mongo-db-atlas-cluster
    properties:
      clusterTier: M10

  mongoSeeder:
    type: custom-resource-definition
    properties:
      packageConfig:
        filePath: seed-the-mongo-cluster.ts
      accessControl:
        allowAccessTo:
          - myMongoCluster

  seedMongoCluster:
    type: custom-resource-instance
    properties:
      definitionName: mongoSeeder
      resourceProperties:
        mongoConnectionString: $Param('myMongoCluster', 'AtlasMongoCluster::SrvConnectionString')

Custom resource definition

Custom resource definition is similar to function definition. This is because on the background custom-resource-definition leverages lambda functions.

resources:
  myMongoCluster:
    type: mongo-db-atlas-cluster
    properties:
      clusterTier: M10

  mongoSeeder:
    type: custom-resource-definition
    properties:
      packageConfig:
        filePath: seed-the-mongo-cluster.ts
      accessControl:
        allowAccessTo:
          - myMongoCluster

  seedMongoCluster:
    type: custom-resource-instance
    properties:
      definitionName: mongoSeeder
      resourceProperties:
        mongoConnectionString: $Param('myMongoCluster', 'AtlasMongoCluster::SrvConnectionString')

Code of custom resource

Package configuration allows you to specify a path to your code and other package properties. The code contains a logic of your custom resource.

resources:
  myMongoCluster:
    type: mongo-db-atlas-cluster
    properties:
      clusterTier: M10

  mongoSeeder:
    type: custom-resource-definition
    properties:
      packageConfig:
        filePath: seed-the-mongo-cluster.ts
      accessControl:
        allowAccessTo:
          - myMongoCluster

  seedMongoCluster:
    type: custom-resource-instance
    properties:
      definitionName: mongoSeeder
      resourceProperties:
        mongoConnectionString: $Param('myMongoCluster', 'AtlasMongoCluster::SrvConnectionString')

Code example

The following example shows the code of our seeding custom resource in file seed-the-mongo-cluster.ts

See AWS documentation to learn more about:

• the type of requests (events) custom resource receives during stack operations
• response objects which you should return during respective operations

import mongoose from 'mongoose';
import { SUCCESS, FAILED, send } from 'cfn-response-promise';

// incoming event is in following form
// {
//   "RequestType" : "Create" || "Update" || "Delete",
//   "RequestId" : "9db53695-b0a0-47d6-908a-ea2d8a3ab5d7",
//   "ResponseURL" : "https://...",
//   "ResourceType" : "AWS::Cloudformation::CustomResource",
//   "LogicalResourceId" : "...",
//   "StackId" : "arn:aws:cloudformation:...",
//   "ResourceProperties" : {
//      ... properties of custom-resource-instance
//   }
// }

export default async (event, context) => {
  // custom resource definition code
  let success = true;
  let dataToReturn = {};

  try {
    // we are only seeding database if the operation is Create
    if (event.RequestType === 'Create') {
      // we are using the "mongoConnectionString" property passed by custom-resource-instance to create connection
      const connection = await mongoose.connect(event.ResourceProperties.mongoConnectionString, {
        authMechanism: 'MONGODB-AWS',
        authSource: '$external',
        useNewUrlParser: true,
        useUnifiedTopology: true,
        dbName: 'my-test-database'
      });
      // code with seeding the database ...
      // ...
    }
  } catch (err) {
    success = false;
  }
  await send(event, context, success ? SUCCESS : FAILED, dataToReturn, 'customresourceid');
};

// function must respond to "ResponseURL" with response in following form
// we are using "cfn-response-promise" library which formats response for us
// {
//   "Status" : "SUCCESS" || "FAILED",
//   "RequestId" : "9db53695-b0a0-47d6-908a-ea2d8a3ab5d7",
//   "LogicalResourceId" : "...",
//   "StackId" : "arn:aws:cloudformation:...",
//   "PhysicalResourceId" : "...",
//   "Data" : {
//     ... attributes which can be queried in template using $Param
//   }
// }

Access control

Access control is configured same as access control of functions or other workloads.

In this example, we are granting access to the mongo cluster. This will inject credentials for access which are used in code.

resources:
  myMongoCluster:
    type: mongo-db-atlas-cluster
    properties:
      clusterTier: M10

  mongoSeeder:
    type: custom-resource-definition
    properties:
      packageConfig:
        filePath: seed-the-mongo-cluster.ts
      accessControl:
        allowAccessTo:
          - myMongoCluster

  seedMongoCluster:
    type: custom-resource-instance
    properties:
      definitionName: mongoSeeder
      resourceProperties:
        mongoConnectionString: $Param('myMongoCluster', 'AtlasMongoCluster::SrvConnectionString')

Environment

Use environment variables for injecting information that resource needs during code execution.

Custom resource instance

Custom-resource-instance, creates instance of specified custom-resource-definition. Depending on the action (create/update/delete) stack is undergoing, the instance receives event (see code example) and executes the code.

resources:
  myMongoCluster:
    type: mongo-db-atlas-cluster
    properties:
      clusterTier: M10

  mongoSeeder:
    type: custom-resource-definition
    properties:
      packageConfig:
        filePath: seed-the-mongo-cluster.ts
      accessControl:
        allowAccessTo:
          - myMongoCluster

  seedMongoCluster:
    type: custom-resource-instance
    properties:
      definitionName: mongoSeeder
      resourceProperties:
        mongoConnectionString: $Param('myMongoCluster', 'AtlasMongoCluster::SrvConnectionString')

API reference