Skip to main content

Provision infrastructure with Dagger and AWS CloudFormation

In this guide, you will learn how to automatically provision infrastructure on AWS by integrating Amazon Cloudformation in your Dagger environment.

We will start with something simple: provisioning a new bucket on Amazon S3. But Cloudformation can provision almost any AWS resource, and Dagger can integrate with the full Cloudformation API.

Prerequisites​

Reminder​

Guidelines​

The provisioning strategy detailed below follows S3 best practices. However, to remain agnostic of your current AWS level, it profoundly relies on S3 and Cloudformation documentation.

Relays​

The first thing to consider when developing a plan based on relays is to read their universe reference: it summarizes the expected inputs and their corresponding formats. Here is the Cloudformation one.

Initialize a Dagger Project and Environment​

(optional) Setup example app​

You will need the local copy of the Dagger examples repository used in previous guides

git clone https://github.com/dagger/examples

Make sure to run all commands from the todoapp directory:

cd examples/todoapp

Organize your package​

Let's create a new directory for our Cue package:

mkdir cloudformation

Create a basic plan​

Let's implement the Cloudformation template and convert it to a Cue definition for further flexibility.

Setup the template and the environment​

Setup the template​

The idea here is to follow best practices in S3 buckets provisioning. Thankfully, the AWS documentation contains a working Cloudformation template that fits 95% of our needs.

1. Tweaking the template: output bucket name only​

Create a file named template.cue and add the following configuration to it.

todoapp/cloudformation/template.cue
package main

// inlined s3 cloudformation template as a string
template: """
      {
        "AWSTemplateFormatVersion": "2010-09-09",
        "Resources": {
          "S3Bucket": {
            "Type": "AWS::S3::Bucket",
            "Properties": {
              "AccessControl": "PublicRead",
              "WebsiteConfiguration": {
                "IndexDocument": "index.html",
                "ErrorDocument": "error.html"
              }
            },
            "DeletionPolicy": "Retain"
          },
          "BucketPolicy": {
            "Type": "AWS::S3::BucketPolicy",
            "Properties": {
              "PolicyDocument": {
                "Id": "MyPolicy",
                "Version": "2012-10-17",
                "Statement": [
                  {
                    "Sid": "PublicReadForGetBucketObjects",
                    "Effect": "Allow",
                    "Principal": "*",
                    "Action": "s3:GetObject",
                    "Resource": {
                      "Fn::Join": [
                        "",
                        [
                          "arn:aws:s3:::",
                          {
                            "Ref": "S3Bucket"
                          },
                          "/*"
                        ]
                      ]
                    }
                  }
                ]
              },
              "Bucket": {
                "Ref": "S3Bucket"
              }
            }
          }
        },
        "Outputs": {
          "Name": {
            "Value": {
              "Fn::GetAtt": ["S3Bucket", "Arn"]
            },
            "Description": "Name S3 Bucket"
          }
        }
      }
    """
2. Cloudformation relay​

As our plan relies on Cloudformation's relay, let's dissect the expected inputs by gradually incorporating them into our plan.

dagger doc alpha.dagger.io/aws/cloudformation
# Inputs:
#     config.region       string                                   AWS region
#     config.accessKey    dagger.#Secret                           AWS access key
#     config.secretKey    dagger.#Secret                           AWS secret key
#     source              string                                   Source is the Cloudformation template (JSON/YAML…
#     stackName           string                                   Stackname is the cloudformation stack
#     parameters          struct                                   Stack parameters
#     onFailure           *"DO_NOTHING" | "ROLLBACK" | "DELETE"    Behavior when failure to create/update the Stack
#     timeout             *10 | >=0 & int                          Maximum waiting time until stack creation/update…
#     neverUpdate         *false | true                            Never update the stack if already exists
1. General insights​

As seen above in the documentation, values starting with * are default values. However, as a plan developer, we may need to add default values to inputs from relays without one: Cue gives you this flexibility.

2. The config value​

The config values are all part of the aws relay. Regarding this package, as you can see above, five of the required inputs miss default options (parameters field is optional):

  • config.region
  • config.accessKey
  • config.secretKey
  • source
  • stackName

Let's implement the first step, use the aws.#Config relay, and request its first inputs: the region to deploy and the AWS credentials.

todoapp/cloudformation/source.cue
package main

import (
    "alpha.dagger.io/aws"
)

// AWS account: credentials and region
awsConfig: aws.#Config

This defines:

  • awsConfig: AWS CLI Configuration step using the package alpha.dagger.io/aws. It takes three user inputs: a region, an accessKey, and a secretKey

Setup the environment​

1. Create a new environment​

Let's create a project:

dagger init

Let's create an environment to run it:

dagger new 'cloudformation' -p ./cloudformation
2. Check plan​

Pro tips: To check whether it worked or not, these three commands might help

dagger input list -e cloudformation # List our personal plan's inputs
# Input                Value                  Set by user  Description
# awsConfig.region     string                 false        AWS region
# awsConfig.accessKey  dagger.#Secret         false        AWS access key
# awsConfig.secretKey  dagger.#Secret         false        AWS secret key

dagger query -e cloudformation # Query values / inspect default values (Instrumental in case of conflict)
# {}

dagger up -e cloudformation # Try to run the plan. As expected, we encounter a failure because some user inputs haven't been set
# 4:11PM ERR system | required input is missing    input=awsConfig.region
# 4:11PM ERR system | required input is missing    input=awsConfig.accessKey
# 4:11PM ERR system | required input is missing    input=awsConfig.secretKey
# 4:11PM FTL system | some required inputs are not set, please re-run with `--force` if you think it's a mistake    missing=0s

Finish template setup​

Now that we have the config definition properly configured, let's modify the Cloudformation one:

todoapp/cloudformation/source.cue
package main

import (
    "alpha.dagger.io/aws"
    "alpha.dagger.io/dagger"
    "alpha.dagger.io/random"
    "alpha.dagger.io/aws/cloudformation"
)

// AWS account: credentials and region
awsConfig: aws.#Config

// Create a random suffix
suffix: random.#String & {
    seed: ""
}

// Query the Cloudformation stackname, or create one with a random suffix for uniqueness
cfnStackName: *"stack-\(suffix.out)" | string & dagger.#Input

// AWS Cloudformation stdlib
cfnStack: cloudformation.#Stack & {
    config:    awsConfig
    stackName: cfnStackName
    source:    template
}

This defines:

  • suffix: random suffix leveraging the random relay. It doesn't have a seed because we don't care about predictability
  • cfnStackName: Name of the stack, either a default value stack-suffix or user input
  • cfnStack: Cloudformation relay with AWS config, stackName and JSON template as inputs

Configure the environment​

Before bringing up the deployment, we need to provide the cfnStack inputs declared in the configuration. Otherwise, Dagger will complain about missing inputs.

dagger up -e cloudformation
# 3:34PM ERR system | required input is missing    input=awsConfig.region
# 3:34PM ERR system | required input is missing    input=awsConfig.accessKey
# 3:34PM ERR system | required input is missing    input=awsConfig.secretKey
# 3:34PM FTL system | some required inputs are not set, please re-run with `--force` if you think it's a mistake    missing=0s

You can inspect the list of inputs (both required and optional) using dagger input list:

dagger input list -e cloudformation
# Input                 Value                                  Set by user  Description
# awsConfig.region      string                                 false        AWS region
# awsConfig.accessKey   dagger.#Secret                         false        AWS access key
# awsConfig.secretKey   dagger.#Secret                         false        AWS secret key
# suffix.length         *12 | number                           false        length of the string
# cfnStack.onFailure    *"DO_NOTHING" | "ROLLBACK" | "DELETE"  false        Behavior when failure to create/update the Stack
# cfnStack.timeout      *10 | >=0 & int                        false        Maximum waiting time until stack creation/update (in minutes)
# cfnStack.neverUpdate  *false | true                          false        Never update the stack if already exists

Let's provide the missing inputs:

dagger input text awsConfig.region us-east-2 -e cloudformation
dagger input secret awsConfig.accessKey yourAccessKey -e cloudformation
dagger input secret awsConfig.secretKey yourSecretKey -e cloudformation

Deploying​

Finally ! We now have a working template ready to be used to provision S3 infrastructures. Let's deploy it:

dagger up -e cloudformation
#2:22PM INF suffix.out | computing
#2:22PM INF suffix.out | completed    duration=200ms
#2:22PM INF cfnStack.outputs | computing
#2:22PM INF cfnStack.outputs | #15 1.304 {
#2:22PM INF cfnStack.outputs | #15 1.304     "Parameters": []
#2:22PM INF cfnStack.outputs | #15 1.304 }
#2:22PM INF cfnStack.outputs | #15 2.948 {
#2:22PM INF cfnStack.outputs | #15 2.948     "StackId": "arn:aws:cloudformation:us-east-2:817126022176:stack/stack-emktqcfwksng/207d29a0-cd0b-11eb-aafd-0a6bae5481b4"
#2:22PM INF cfnStack.outputs | #15 2.948 }
#2:22PM INF cfnStack.outputs | completed    duration=35s

dagger output list -e cloudformation
# Output                 Value                                                    Description
# suffix.out             "emktqcfwksng"                                           generated random string
# cfnStack.outputs.Name  "arn:aws:s3:::stack-emktqcfwksng-s3bucket-9eiowjs1jab4"  -

The deployment went well!

In case of a failure, the Debug deploy tab shows the command to get more information. The name of the provisioned S3 instance lies in the cfnStack.outputs.Name output key, without arn:aws:s3:::

With this provisioning infrastructure, your dev team will easily be able to instantiate aws infrastructures: all they need to know is dagger input list -e cloudformation and dagger up -e cloudformation isn't that awesome? :-D

Cue Cloudformation template​

This section will convert the inlined JSON template to CUE to take advantage of the language features.

To do so quickly, we will first transform the template from JSON format to Cue format, then optimize it to leverage Cue's forces.

1. Create convert.cue​

We will create a new convert.cue file to process the conversion

todoapp/cloudformation/convert.cue
package main

import "encoding/json"

s3Template: json.Unmarshal(template)

This defines:

  • s3Template: contains the unmarshalled template.

You need to empty the plan and copy the convert.cue file to the plan for Dagger to reference it

mv cloudformation/source.cue ~/tmp/

2. Retrieve the Unmarshalled JSON​

Then, still in the same folder, query the s3Template value to retrieve the Unmarshalled result of s3:

dagger query s3Template -e cloudformation
# {
#   "AWSTemplateFormatVersion": "2010-09-09",
#   "Outputs": {
#     "Name": {
#       "Description": "Name S3 Bucket",
#       "Value": {
#         "Fn::GetAtt": [
#           "S3Bucket",
#           "Arn"
#           ...

The commented output above is the cue version of the JSON Template, copy it

3. Remove convert.cue​

rm cloudformation/convert.cue

4. Store the output​

Open cloudformation/template.cue and append below elements with copied Cue definition of the JSON:

todoapp/cloudformation/template.cue
// Add this line, to make it part to the cloudformation template
package main

import "encoding/json"

// Wrap exported Cue in previous point inside the `s3` value
s3: {
    AWSTemplateFormatVersion: "2010-09-09"
    Outputs: Name: {
        Description: "Name S3 Bucket"
        Value: "Fn::GetAtt": [
            "S3Bucket",
            "Arn",
        ]
    }
    Resources: {
        BucketPolicy: {
            Properties: {
                Bucket: Ref: "S3Bucket"
                PolicyDocument: {
                    Id: "MyPolicy"
                    Statement: [
                        {
                            Action:    "s3:GetObject"
                            Effect:    "Allow"
                            Principal: "*"
                            Resource: "Fn::Join": [
                                "",
                                [
                                    "arn:aws:s3:::",
                                    {
                                        Ref: "S3Bucket"
                                    },
                                    "/*",
                                ],
                            ]
                            Sid: "PublicReadForGetBucketObjects"
                        },
                    ]
                    Version: "2012-10-17"
                }
            }
            Type: "AWS::S3::BucketPolicy"
        }
        S3Bucket: {
            DeletionPolicy: "Retain"
            Properties: {
                AccessControl: "PublicRead"
                WebsiteConfiguration: {
                    ErrorDocument: "error.html"
                    IndexDocument: "index.html"
                }
            }
            Type: "AWS::S3::Bucket"
        }
    }
}

// Template contains the marshalled value of the s3 template
template: json.Marshal(s3)

We're using the built-in json.Marshal function to convert CUE back to JSON, so Cloudformation still receives the same template.

You can inspect the configuration using dagger query -e cloudformation to verify it produces the same manifest:

dagger query template -f text -e cloudformation

Now that the template is defined in CUE, we can use the language to add more flexibility to our template.

Let's define a re-usable #Deployment definition in todoapp/cloudformation/deployment.cue:

todoapp/cloudformation/deployment.cue
package main

#Deployment: {

    // Bucket's output description
    description: string

    // index file
    indexDocument: *"index.html" | string

    // error file
    errorDocument: *"error.html" | string

    // Bucket policy version
    version: *"2012-10-17" | string

    // Retain as default deletion policy. Delete is also accepted but requires the s3 bucket to be empty
    deletionPolicy: *"Retain" | "Delete"

    // Canned access control list (ACL) that grants predefined permissions to the bucket
    accessControl: *"PublicRead" | "Private" | "PublicReadWrite" | "AuthenticatedRead" | "LogDeliveryWrite" | "BucketOwnerRead" | "BucketOwnerFullControl" | "AwsExecRead"

    // Modified copy of s3 value in `todoapp/cloudformation/template.cue`
    template: {
        AWSTemplateFormatVersion: "2010-09-09"
        Outputs: Name: {
            Description: description
            Value: "Fn::GetAtt": [
                "S3Bucket",
                "Arn",
            ]
        }
        Resources: {
            BucketPolicy: {
                Properties: {
                    Bucket: Ref: "S3Bucket"
                    PolicyDocument: {
                        Id: "MyPolicy"
                        Statement: [
                            {
                                Action:    "s3:GetObject"
                                Effect:    "Allow"
                                Principal: "*"
                                Resource: "Fn::Join": [
                                    "",
                                    [
                                        "arn:aws:s3:::",
                                        {
                                            Ref: "S3Bucket"
                                        },
                                        "/*",
                                    ],
                                ]
                                Sid: "PublicReadForGetBucketObjects"
                            },
                        ]
                        Version: version
                    }
                }
                Type: "AWS::S3::BucketPolicy"
            }
            S3Bucket: {
                DeletionPolicy: deletionPolicy
                Properties: {
                    AccessControl: "PublicRead"
                    WebsiteConfiguration: {
                        ErrorDocument: errorDocument
                        IndexDocument: indexDocument
                    }
                }
                Type: "AWS::S3::Bucket"
            }
        }
    }
}

template.cue can be rewritten as follows:

todoapp/cloudformation/template.cue
package main

import "encoding/json"

s3: #Deployment & {
    description: "Name S3 Bucket"
}

// Template contains the marshalled value of the s3 template
template: json.Marshal(s3.template)

Verify template

Double-checks at the template level can be done with manual uploads on Cloudformation's web interface or by executing the below command locally:

tmpfile=$(mktemp ./tmp.XXXXXX) && dagger query template -f text -e cloudformation > "$tmpfile" && aws cloudformation validate-template  --template-body file://"$tmpfile" ; rm "$tmpfile"

Let's make sure it yields the same result:

dagger query template -f text -e cloudformation
# {
#   "description": "Name S3 Bucket",
#   "indexDocument": "index.html",
#   "errorDocument": "error.html",
#   "version": "2012-10-17",
#   "deletionPolicy": "Retain",
#   "accessControl": "PublicRead",
#   "template": {
#     "AWSTemplateFormatVersion": "2010-09-09",
#     "Outputs": {
#       "Name": {
#         "Description": "Name S3 Bucket",
#         "Value": {

Reimplement source.cue:

todoapp/cloudformation/source.cue
package main

import (
    "alpha.dagger.io/aws"
    "alpha.dagger.io/dagger"
    "alpha.dagger.io/random"
    "alpha.dagger.io/aws/cloudformation"
)

// AWS account: credentials and region
awsConfig: aws.#Config

// Create a random suffix
suffix: random.#String & {
    seed: ""
}

// Query the Cloudformation stackname, or create one with a random suffix for uniqueness
cfnStackName: *"stack-\(suffix.out)" | string & dagger.#Input

// AWS Cloudformation stdlib
cfnStack: cloudformation.#Stack & {
    config:    awsConfig
    stackName: cfnStackName
    source:    template
}

And we can now deploy it:

dagger up -e cloudformation
#2:22PM INF suffix.out | computing
#2:22PM INF suffix.out | completed    duration=200ms
#2:22PM INF cfnStack.outputs | computing
#2:22PM INF cfnStack.outputs | #15 1.304 {
#2:22PM INF cfnStack.outputs | #15 1.304     "Parameters": []
#2:22PM INF cfnStack.outputs | #15 1.304 }
#2:22PM INF cfnStack.outputs | #15 2.948 {
#2:22PM INF cfnStack.outputs | #15 2.948     "StackId": "arn:aws:cloudformation:us-east-2:817126022176:stack/stack-emktqcfwksng/207d29a0-cd0b-11eb-aafd-0a6bae5481b4"
#2:22PM INF cfnStack.outputs | #15 2.948 }
#2:22PM INF cfnStack.outputs | completed    duration=35s

Name of the deployed bucket:

dagger output list -e cloudformation
# Output                 Value                                                    Description
# suffix.out             "ucwcecwwshdl"                                           generated random string
# cfnStack.outputs.Name  "arn:aws:s3:::stack-ucwcecwwshdl-s3bucket-gaqmj8rzsl08"  -

The name of the provisioned S3 instance lies in the cfnStack.outputs.Name output key, without arn:aws:s3:::

PS: This plan could be further extended with the AWS S3 example. It could provide infrastructure and quickly deploy it.

PS1: As it could be an excellent first exercise for you, this won't be detailed here. However, we're interested in your imagination: let us know your implementations :-)