Pragmatic Pineapple 🍍

Building and Deploying Microservices with AWS Lambda and Semaphore

Published Last updated loading views

AWS Lambda is a service from the Amazon Web Services family which runs your code based on various events. When you create a Lambda function and deploy your code to it, AWS Lambda takes care of provisioning and managing servers that run your code.

This tutorial will show you how to develop a Node.js Lambda function from scratch and deploy it to AWS Lambda. To achieve a continuous delivery workflow for our function, we will automate the deployment process through Semaphore’s integration with Lambda.

Once the function is complete, it will text message anyone who breaks the build on your project on Semaphore.

The function we are developing will use:

  • Grunt (a task runner for Javascript, which we will use to invoke our function locally)
  • NPM (the primary package manager for Node.js),
  • Twilio API (we will use Twilio’s API to send text messages)
  • AWS SDK for Node.js (a simple and easy-to-use API for AWS services)

Prerequisites

To develop this function, you will need:

Initializing a Lambda Function

A Lambda function is basically the code that is run when a function is invoked. Typically, it is invoked by the events that you define.

Let’s create a new Lambda function. To do so, we need to go to the Lambda console and click on Get Started Now.

To start with, we will select the Hello World Node.js blueprint for our function.

After you have selected a blueprint, you should get the following screen:

Now we need to name our function, give it a description, and choose the runtime language, as shown below.

We will leave the Lambda function code section as is, and we will replace that code with our own once we are ready.

Under Lambda function handler and role we will keep the default index.handler handler, but we will need to create a new execution role. You can learn more about execution roles in the AWS Lambda documentation on granting permissions using the execution role.

We will use the S3 API in our code, so we will need an S3 execution role for the function. More on that later on.

Select the S3 execution role in the dropdown menu to create a role.

This will take you to a dialog where you can create a new IAM Role. We will name it semaphore-deploy-role. This role has a set of permissions that will both let us use AWS SDK in our code and let Semaphore do the deployment for us. After you have created a new IAM Role, you can click on Allow.

Now that you’ve created an IAM role for your function, you can go ahead and select it.

Next, we will increase timeout to 6 seconds, and leave the required memory at 128 MB under Advanced settings.

After we’ve done all of this, our settings should look like this:

Click on Next, review the function details and click on Create function.

Writing the Code

Storing Phone Numbers

Now that we have created a Lambda function, it’s time to write some code. We will create a function that retrieves the JSON file from AWS S3. We will add the emails and phone numbers of the users who are working on our project on Semaphore to this file. This file will also contain the Twilio credentials (the Account SID, the Authentication Token and the Twilio Number). This data is stored on S3 so that no one can see it in our code.

Retrieving Twilio Credentials

You can retrieve your Twilio credentials from their website, given you have previously created a (free) Twilio account. You will also need to enable sending SMS messages to your country in the international settings for your account and verify your number and every other number to which you will need to send text messages. You can do that on the Twilio verified numbers page.

Next, open an editor of your choice and make a file named numbers.json. Add your email, the phone numbers and Twilio credentials to this file. Here’s an example of what this file should look like:

{
  "twilio": {
    "twilio_account_sid": "12345678",
    "twilio_auth_token": "12345678",
    "twilio_number": "+420123456"
  },
  "[email protected]": "+3333333",
  "[email protected]": "+2222222",
  "[email protected]": "+1111111"
}

Go to the AWS S3 console and upload the numbers.json to a bucket of your choice.

Defining Dependencies

Next, we will define the dependencies that we are using in our Node.js application in package.json. This file should look similar to the following:

{
  "name": "congrats-you-broke-the-build",
  "version": "0.0.1",
  "main": "index.js",
  "devDependencies": {
    "aws-sdk": "2.0.23"
  },
  "dependencies": {
    "twilio": "2.1.0"
  }
}

Installing Grunt CLI

We will also need to install Grunt in order to test our function. Grunt is a task runner for JavaScript, which is used for easy automation of project tasks such as building and packaging. Install it with the following code:

$ sudo npm install -g grunt-cli

If you’re having trouble with installing Grunt, you can find more information on the NPM Grunt page.

Defining the Lambda Function Logic

Now that we have uploaded numbers.json to S3 and installed Grunt CLI, we can implement some code. When we invoke our Lambda function, exports.handler will handle the request that invoked it. The event parameter is the JSON file that was included in the request sent to our Lambda function.

We need to create a file named index.js to hold our Lambda function logic. First, let’s fetch numbers.json from S3 using the following code segment:

var AWS = require("aws-sdk")

exports.handler = function (event, context) {
  console.log("JSON API from Semaphore: %j", event)

  AWS.config.apiVersions = {
    s3: "2006-03-01",
  }

  // My bucket with numbers.json is located in 'us-west-2' region
  var s3 = new AWS.S3({ region: "us-west-2" })
  // This is where you define bucket and a file for S3 to get
  var params = { Bucket: "congrats-you-broke-the-build", Key: "numbers.json" }

  s3.getObject(params, function (err, data) {
    if (err) console.log(err, err.stack) // an error has happened on AWS

    // Parse JSON file and put it in numbers variable
    var numbers = JSON.parse(data.Body)
  })
}

Now that we have parsed our JSON file, it’s time do something with those numbers. We will add a new function called manipulateNumbers to our code. This function will receive numbers from our numbers.json file.

var AWS = require("aws-sdk")

exports.handler = function (event, context) {
  console.log("JSON API from Semaphore: %j", event)

  AWS.config.apiVersions = {
    s3: "2006-03-01",
  }

  // My bucket with numbers.json is located in 'us-west-2' region
  var s3 = new AWS.S3({ region: "us-west-2" })
  // This is where you define the location of the bucket and the file S3 needs to retrieve
  var params = { Bucket: "congrats-you-broke-the-build", Key: "numbers.json" }

  s3.getObject(params, function (err, data) {
    if (err) console.log(err, err.stack) // an error has happened on AWS

    // Parse JSON file and put it in numbers variable
    var numbers = JSON.parse(data.Body)

    manipulateNumbers(numbers)
  })

  function manipulateNumbers(numbers) {
    // If someone breaks the master build on Semaphore, get inside the if statement
    if (event.branch_name == "master" && event.result == "failed") {
      // We get the name of a user who broke the build
      var blame = event.commit.author_name

      // message that is sent to the developer who broke the master branch
      var message =
        "Congrats " +
        blame +
        ", you managed to break the master branch on SemaphoreCI!."
    }
  }
}

This function checks if the branch_name is master and if the build has failed. It includes the message that will be sent to the user who broke the build.

We now have the message we will send to the user, but we still need to write the code that will send it:

var AWS = require("aws-sdk")
var twilio = require("twilio")

exports.handler = function (event, context) {
  console.log("JSON API from Semaphore: %j", event)

  AWS.config.apiVersions = {
    s3: "2006-03-01",
  }

  // My bucket with numbers.json is located in the 'us-west-2' region
  var s3 = new AWS.S3({ region: "us-west-2" })
  // This is where you define the location of the bucket and the file S3 needs to retrieve
  var params = { Bucket: "congrats-you-broke-the-build", Key: "numbers.json" }

  s3.getObject(params, function (err, data) {
    if (err) console.log(err, err.stack) // an error has happened on AWS

    // Parse JSON file and put it in numbers variable
    var numbers = JSON.parse(data.Body)

    manipulateNumbers(numbers)
  })

  function manipulateNumbers(numbers) {
    // If someone breaks the master build on Semaphore, enter the if statement
    if (event.branch_name == "master" && event.result == "failed") {
      // We get the name of the user who broke a build
      var blame = event.commit.author_name

      // The message that is sent to the developer who broke the master branch
      var message =
        "Congrats " +
        blame +
        ", you managed to brake master branch on SemaphoreCI!."

      twilioHandler(numbers, message)
    }
  }

  function twilioHandler(numbers, message) {
    var blame_mail = event.commit.author_email
    // twilio credentials
    var twilio_account_sid = numbers.twilio.twilio_account_sid
    var twilio_auth_token = numbers.twilio.twilio_auth_token
    var twilio_number = numbers.twilio.twilio_number

    var client = twilio(twilio_account_sid, twilio_auth_token)

    // Send SMS
    client.sendSms(
      {
        to: numbers[blame_mail],
        from: twilio_number,
        body: message,
      },
      function (err, responseData) {
        // this function is executed when a response is received from Twilio
        if (!err) {
          console.log(responseData)
          context.done(null, "Message sent to " + numbers[blame_mail] + "!")
        } else {
          console.log(err)
          context.done(null, "There was an error, message not sent!")
        }
      }
    )
  }
}

We requested a Twilio package, which we will use to send the message. We also added the twilioHandler function. This function creates a Twilio client and extracts the phone number from the numbers variable based on author_email from event. After it’s done retrieving the data needed for sending the text message, it calls client.sendSms(), which sends the text message to the user who broke the build.

Now we’re ready to test the function.

Testing the Function Locally

Let’s create a event.json file, so we can test our function locally. We will try to mimic the JSON that Semaphore API will send us. In that file, we will set branch_name to master and the result of the build to failed. This is the scenario that will trigger sending an SMS to the user who breaks the master branch.

{
  "branch_name": "master",
  "result": "failed",
  "commit": {
    "id": "ce03782d581ed985caf9c479173a14962b0fe941",
    "url": "https://github.com/lazy_user//commit/ce03782d581ed985caf9c479173a14962b0fe941",
    "author_name": "Lazy user",
    "author_email": "[email protected]",
    "message": "Add index.js",
    "timestamp": "2015-12-07T11:31:38+01:00"
  }
}

Next, we will create Gruntfile.js so we can invoke our Lambda function locally.

var grunt = require("grunt")
grunt.loadNpmTasks("grunt-aws-lambda")

grunt.initConfig({
  lambda_invoke: {
    default: {},
  },
})

We’re now ready to test our function. Run:

$ npm install

When the NPM package installation finishes, we can invoke our Lambda function by executing the following line:

$ grunt lambda_invoke

The output should look similar to this:

Success!  Message:
------------------
Message sent to +2222222!

Deploying a Lambda Function

Now that we’ve developed the Lambda function and tested it locally, it’s time to deploy it to the internet.

Pushing the Code to GitHub

To integrate the code with Semaphore, we need to first push it to GitHub. You can find detailed instructions on how to do that in GitHub’s documentation

Building the Project on Semaphore

Now that we’ve pushed our code to GitHub, we can add the project from GitHub to Semaphore.

Once the analysis is done and you have the build commands in front of you, you can click on Build With These Settings. The build should pass, and you should a page that looks like this:

Setting Up Deployment to AWS Lambda

To set up deployment, you need to go to your project page on Semaphore and click on Set Up Deployment. You will get a screen with different deployment options. We will choose AWS Lambda.

Selecting the Deployment Method

There are two deployment methods available on Semaphore: automatic and manual deployment.

Automatic deployment means that a deploy will be triggered after every passed build on the selected branch. In addition, you can also manually deploy any build from any branch at any time.

For automatic deployment, you will be asked to select which branch will be automatically deployed after each passed build.

Manual deployment requires the manual selection of the builds to deploy.

We will choose Automatic deployment.

Note: You can change the deployment strategy at any time in server settings once the setup is complete.

Selecting the Branch

In this step, we will choose from which branch we wish to deploy the application. We’ll select the master branch.

Entering the AWS Credentials

Next, enter your AWS credentials on the following screen. If you need help getting these credentials, consult the Retrieving Security Credentials section of our tutorial on deploying a web application to Elastic Beanstalk. The process for retrieving credentials is the same when deploying to Lambda.

After entering the credentials, select the region your application resides in on the same screen. This will enable Semaphore to list the functions you have in the region you have specified.

Selecting the Function

The next step is to select the function to which we will deploy our code. We’ll select congrats-you-broke-the-build. Next, type npm install in the dependencies installation commands section. This will install the packages we are using in our application, so that they are deployed to Lambda.

Naming the Server on Semaphore

In this step, you need to provide a name for your server. This name will be used on your Semaphore dashboard and deployment timeline.

The setup is now complete — you are ready to automatically deploy to Lambda. All you need to do is click on Deploy.

Testing the Function on AWS Lambda

You can now go to your Lambda function to test it. Click on Actions and select Configure test event.

Copy the contents of event.json to the Input test event editor. Enter the email address and phone number you defined in your numbers.json in order to be able to receive a message.

Click on Save and test. Congrats, you managed to test your first AWS Lambda function.

Enabling Notifications

Next, we will tell Semaphore to send a POST request to our Lambda function every time a build is finished. Let’s make an Amazon API Gateway for our Lambda functions. Go to the API endpoints tab of the Lambda function and click on Add API endpoint. The configuration should look as follows:

Click on Submit and copy the API endpoint URL. We will need this URL to add a webhook to Semaphore.

Now, go to your project settings on Semaphore. Click on Notifications, and then on Webhooks. Click on Add Webhook and paste the link you copied from Lambda to the URL text field. Your screen should look similar to this:

Click on Test. This will send a test request to the API end point that you copied. Next, go to the Monitoring page on AWS Lambda and click on View logs in CloudWatch, where you can see that the function has been invoked.

Breaking the Build

Let’s see if this works in a real test case scenario. For this purpose, we will break a build on Semaphore. Go to your project settings, click on Build Settings and add a new command line to any thread:

$ false

just to see how it breaks. Now click on Start. Your build should fail, and this will cause a request to be sent to your Lambda function. All that’s left to do is to wait for the text message to arrive on your phone.

Congrats, you’ve managed to successfully develop, test and deploy your first Lambda function through Semaphore!

The code from this tutorial is available in this GitHub repository.


This article was originally posted on Semaphore


Nikola Đuza

Written by Nikola Đuza who helps developers improve their productivity by sharing pragmatic advice & applicable knowledge on JavaScript and Ruby. You can connect with him on Twitter.

© 2024 Nikola Đuza