Episode 1:Core Concepts

This is the first video in a 10 part series by Mohamed Labouardy, showing how to build a simple DevOps pipeline, including built-in security at each stage. To follow the series, please join the community to receive notification as each episode is released.



Highly Available Docker Registry on AWS with Nexus

Have you ever wondered how you can build a highly available & resilient Docker Repository to store your Docker Images ?



In this post, we will setup an EC2 instance inside a Security Group and create an A record pointing to the server Elastic IP address as follow:



To provision the infrastructure, we will use Terraform as IaC (Infrastructure as Code) tool. The advantage of using this kind of tools is the ability to spin up a new environment quickly in different AWS region (or different IaaS provider) in case of incident (Disaster recovery).

Start by cloning the following Github repository:

1
git clone https://github.com/mlabouardy/terraform-aws-labs.git

Inside docker-registry folder, update the variables.tfvars with your own AWS credentials (make sure you have the right IAM policies).

1
2
3
4
5
6
7
8
9
10
11
12
resource "aws_instance" "default" {
ami = "${lookup(var.amis, var.region)}"
instance_type = "${var.instance_type}"
key_name = "${aws_key_pair.default.id}"
security_groups = ["${aws_security_group.default.name}"]

user_data = "${file("setup.sh")}"

tags {
Name = "registry"
}
}

I specified a shell script to be used as user_data when launching the instance. It will simply install the latest version of Docker CE and turn the instance to Docker Swarm Mode (to benefit from replication & high availability of Nexus container)

1
2
3
4
5
6
7
#!/bin/sh
yum update -y
yum install -y docker
service docker start
usermod -aG docker ec2-user
docker swarm init
docker service create --replicas 1 --name registry --publish 5000:5000 --publish 8081:8081 sonatype/nexus3:3.6.2

Note: Surely, you can use a Configuration Management Tools like Ansible or Chef to provision the server once created.

Then, issue the following command to create the infrastructure:

1
terraform apply -var-file=variables.tfvars

Once created, you should see the Elastic IP of your instance:



Connect to your instance via SSH:

1
ssh ec2-user@35.177.167.36

Verify that the Docker Engine is running in Swarm Mode:



Check if Nexus service is running:



If you go back to your AWS Management Console. Then, navigate to Route53 Dashboard, you should see a new A record has been created which points to the instance IP address.



Point your favorite browser to the Nexus Dashboard URL (registry.slowcoder.com:8081). Login and create a Docker hosted registry as below:



Edit the /etc/docker/daemon.json file, it should have the following content:

1
2
3
{
"insecure-registries" : ["registry.slowcoder.com:5000"]
}

Note: For production it’s highly recommended to secure your registry using a TLS certificate issued by a known CA.

Restart Docker for the changes to take effect:

1
service docker restart

Login to your registry with Nexus Credentials (admin/admin123):



In order to push a new image to the registry:

1
docker push registry.slowcoder.com:5000/mlabouardy/movies-api:1.0.0-beta


Verify that the image has been pushed to the remote repository:



To pull the Docker image:

1
docker pull registry.slowcoder.com:5000/mlabouardy/movies-api:1.0.0-beta


Note: Sometimes you end up with many unused & dangling images that can quickly take significant amount of disk space:



You can either use the Nexus CLI tool or create a Nexus Task to cleanup old Docker Images:



Populate the form as below:



The task above will run everyday at midnight to purge unused docker images from “mlabouardy” registry.

Drop your comments, feedback, or suggestions below — or connect with me directly on Twitter @mlabouardy.

Build a CI/CD pipeline for Dockerized Microservices and Serverless Functions in AWS

Learn DevSecOps best practices with free hands-on labs and workshops. My workshop will focus on a range of subjects from building a CI/CD pipeline in AWS for Dockerized Microservices and Serverless functions to Monitoring and Logging. Join for the rollout this Wednesday, January 16, with new sessions added each week for the next 10 weeks.



There is no cost to participate in the workshops other than your contact information.


Join Here

Hosting a Free Static Website on Google Cloud Storage

This guide walks you through setting up a free bucket to serve a static website through a custom domain name using Google Cloud Platform services.

Sign in to Google Cloud Platform, navigate to Cloud DNS service and create a new public DNS zone:



By default it will have a NS (Nameserver) and a SOA (Start of Authority) records:



Go to you domain registrar, in my case I purchased a domain name from GoDaddy (super cheap). Add the nameserver names that were listed in your NS record:



PS: It can take some time for the changes on GoDaddy to propagate through to Google Cloud DNS.

Next, verify you own the domain name using the Open Search Console. Many methods are available (HTML Meta data, Google Analytics, etc). The easiest one is DNS verification through a TXT record:



Add the TXT record to your DNS zone created earlier:



DNS changes might take some time to propagate:



Once you have verified domain, you can create a bucket with Cloud Storage under the verified domain name. The storage class should be “Multi-Regional” (geo redundant bucket, in case of outage) :



Copy the website static files to the bucket using the following command:

1
gsutil rsync -R . gs://www.serverlessmovies.com/

After the upload completes, your static files should be available on the bucket as follows:



Next, make the files publicly accessible by adding allUsers entity with Object Viewer role to the bucket permissions:



Once shared publicly, a link icon appears for each object in the public access column. You can click on this icon to get the URL for the object:



Verify that content is served from the bucket by requesting the index.html link in you browser:



Next, set the main page to be index.html from “Edit website configuration” section:



Now, we need to map our domain name with the bucket we created earlier. Create a CNAME record that points to c.storage.googleapis.com:



Point your browser to your domain name, your website should be served:



While our solution works like a charm, we can access our content through HTTP only (Google Cloud Storage only supports HTTP when using it through a CNAME record). In the next post, we will serve our content through a custom domain over SSL using a Content Delivery Network (CDN).

Drop your comments, feedback, or suggestions below — or connect with me directly on Twitter @mlabouardy.

Deploy Private Docker Registry on GCP with Nexus, Terraform and Packer

In this post, I will walk you through how to deploy Sonatype Nexus OSS 3 on Google Cloud Platform and how to create a private Docker hosted repository to store your Docker images and other build artifacts (maven, npm and pypi, etc). To achieve this, we need to bake our machine image using Packer to create a gold image with Nexus preinstalled and configured. Terraform will be used to deploy a Google compute instance based on the baked image. The following schema describes the build workflow:



PS : All the templates used in this tutorial, can be found on my GitHub.

To get started, we need to create the machine image to be used with Google Compute Engine (GCE). Packer will create a temporary instance based on the CentOS image and use a shell script to provision the instance:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
{
"variables" : {
"zone" : "YOUR ZONE",
"project" : "YOUR PROJECT ID",
"source_image" : "centos-7-v20181210",
"ssh_username" : "packer",
"credentials_path" : "PATH/account.json"
},
"builders" : [
{
"type": "googlecompute",
"account_file": "{{user `credentials_path`}}",
"project_id": "{{user `project`}}",
"source_image": "{{user `source_image`}}",
"ssh_username": "{{user `ssh_username`}}",
"zone": "{{user `zone`}}",
"image_name" : "nexus-v3-14-0-04"
}
],
"provisioners" : [
{
"type" : "file",
"source" : "./nexus.rc",
"destination" : "/tmp/nexus.rc"
},
{
"type" : "file",
"source" : "./repository.json",
"destination" : "/tmp/repository.json"
},
{
"type" : "shell",
"script" : "./setup.sh",
"execute_command" : "sudo -E -S sh '{{ .Path }}'"
}
]
}

The shell script, will install the latest stable version of Nexus OSS based on their official documentation and wait for the service to be up and running, then it will use the Scripting API to post a groovy script:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
#!/bin/bash

NEXUS_USERNAME="admin"
NEXUS_PASSWORD="admin123"

echo "Install Java JDK 8"
yum update -y
yum install -y java-1.8.0-openjdk wget

echo "Install Nexus OSS"
mkdir /opt/nexus
cd /opt/nexus
wget https://download.sonatype.com/nexus/3/latest-unix.tar.gz
tar -xvf latest-unix.tar.gz
rm latest-unix.tar.gz
mv nexus-3.14.0-04 nexus
useradd nexus
chown -R nexus:nexus /opt/nexus/
ln -s /opt/nexus/nexus/bin/nexus /etc/init.d/nexus
cd /etc/init.d
chkconfig --add nexus
chkconfig --levels 345 nexus on
mv /tmp/nexus.rc /opt/nexus/nexus/bin/nexus.rc
service nexus restart

until $(curl --output /dev/null --silent --head --fail http://localhost:8081); do
printf '.'
sleep 2
done


echo "Upload Groovy Script"
curl -v -X POST -u $NEXUS_USERNAME:$NEXUS_PASSWORD --header "Content-Type: application/json" 'http://localhost:8081/service/rest/v1/script' -d @/tmp/repository.json

echo "Execute it"
curl -v -X POST -u $NEXUS_USERNAME:$NEXUS_PASSWORD --header "Content-Type: text/plain" 'http://localhost:8081/service/rest/v1/script/docker-repository/run'

The script will create a Docker private registry listening on port 5000:

1
2
3
4
import org.sonatype.nexus.blobstore.api.BlobStoreManager; 
import org.sonatype.nexus.repository.storage.WritePolicy;

repository.createDockerHosted('mlabouardy', 5000, 443, BlobStoreManager.DEFAULT_BLOBSTORE_NAME, true, true, WritePolicy.ALLOW)

Once the template files are defined, issue packer build command to bake our machine image:



If you head back to Images section from Compute Engine dashboard, a new image called nexus should be created:



Now we are ready to deploy Nexus, we will create a Nexus server based on the machine image we baked with Packer. The template file is self-explanatory, it creates a set of firewall rules to allow inbound traffic on port 8081 (Nexus GUI) and 22 (SSH) from anywhere, and creates a google compute instance based on the Nexus image:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
provider "google" {
credentials = "${file("${var.credentials}")}"
project = "${var.project}"
region = "${var.region}"
}

resource "google_compute_firewall" "nexus" {
name = "nexus-firewall"
network = "${google_compute_network.nexus.name}"

allow {
protocol = "tcp"
ports = ["22", "8081"]
}

source_ranges = ["0.0.0.0/0"]
}

resource "google_compute_network" "nexus" {
name = "nexus-network"
}

resource "google_compute_instance" "nexus" {
name = "nexus"
machine_type = "${var.instance_type}"
zone = "${var.zone}"

boot_disk {
initialize_params {
image = "${var.image_name}"
size = 100
}
}

metadata {
sshKeys = "${var.ssh_user}:${file(var.ssh_pub_key_file)}"
}

network_interface {
network = "${google_compute_network.nexus.name}"
access_config = {}
}
}

On the terminal, run the terraform init command to download and install the Google provider, shown as follows:



Create an execution plan (dry run) with the terraform plan command. It shows you things that will be created in advance, which is good for debugging and ensuring that you’re not doing anything wrong, as shown in the next screenshot:



When you’re ready, go ahead and apply the changes by issuing terraform apply:



Terraform will create the needed resources and display the public ip address of the nexus instance on the output section. Jump back to GCP Console, your nexus instance should be created:



If you point your favorite browser to http://instance_ip:8081, you should see the Sonatype Nexus Repository Manager interface:



Click the “Sign in” button in the upper right corner and use the username “admin” and the password “admin123”. Then, click on the cogwheel to go to the server administration and configuration section. Navigate to “Repositories”, our private Docker repository should be created as follows:



The docker repository is published as expected on port 5000:



Hence, we need to allow inbound traffic on that port, so update the firewall rules accordingly:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
resource "google_compute_firewall" "nexus" {
name = "nexus-firewall"
network = "${google_compute_network.nexus.name}"

allow {
protocol = "tcp"
ports = ["22", "8081", "5000"]
}

source_ranges = ["0.0.0.0/0"]
}

resource "google_compute_network" "nexus" {
name = "nexus-network"
}

Issue terrafrom apply command to apply the changes:



Your private docker registry is ready to work at instance_ip:5000, let’s test it by pushing a docker image.

Since we have exposed the private Docker registry on a plain HTTP endpoint, we need to configure the Docker daemon that will act as client to the private Docker registry as to allow for insecure connections.



  • On Windows or Mac OS X: Click on the Docker icon in the tray to open Preferences. Click on the Daemon tab and add the IP address on which the Nexus GUI is exposed along with the port number 5000 in Insecure registries section. Don’t forget to Apply & Restart for the changes to take effect and you’re ready to go.
  • Other OS: Follow the official guide.

You should now be able to log in to your private Docker registry using the following command:



And push your docker images to the registry with the docker push command:



If you head back to Nexus Dashboard, your docker image should be stored with the latest tag:



Drop your comments, feedback, or suggestions below — or connect with me directly on Twitter @mlabouardy.

Deploy a Docker Swarm cluster on GCP using Terraform in 8 steps

Kubernetes might be the ultimate choice when deploying heavy workloads on Google Cloud Platform. However, Docker Swarm has always been quite popular among developers who prefer fast deployments and simplicity— and among ops who are learning to get comfortable with an orchestrated environment.

In this post, we will walk through how to deploy a Docker Swarm cluster on GCP using Terraform from scratch. Let’s do it!



All the templates and playbooks used in this tutorial, can be found on my GitHub.

Get Started

To get started, sign in to your Google Cloud Platform console and create a service account private key from IAM:



Download the JSON file and store it in a secure folder.

For simplicity, I have divided my Swarm cluster components to multiple template files — each file is responsible for creating a specific Google Compute resource.

1. Setup your swarm managers

In this example, I have defined the Docker Swarm managers based on the CoreOS image:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
resource "google_compute_instance" "managers" {
count = "${var.swarm_managers}"
name = "manager"
machine_type = "${var.swarm_managers_instance_type}"
zone = "${var.zone}"

boot_disk {
initialize_params {
image = "${var.image_name}"
size = 100
}
}

metadata {
sshKeys = "${var.ssh_user}:${file(var.ssh_pub_key_file)}"
}

network_interface {
network = "${google_compute_network.swarm.name}"
access_config = {}
}
}

2. Setup your swarm workers

Similarly, a set of Swarm workers based on CoreOS image, and I have used the resource dependencies feature of Terraform to ensure the Swarm managers are deployed first. Please note the usage of depends_on keyword:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
resource "google_compute_instance" "workers" {
count = "${var.swarm_workers}"
name = "worker${count.index + 1}"
machine_type = "${var.swarm_workers_instance_type}"
zone = "${var.zone}"

depends_on = ["google_compute_instance.managers"]

boot_disk {
initialize_params {
image = "${var.image_name}"
size = 100
}
}

metadata {
sshKeys = "${var.ssh_user}:${file(var.ssh_pub_key_file)}"
}

network_interface {
network = "${google_compute_network.swarm.name}"
access_config = {}
}
}

3. Define your network rules

Also, I have defined a network interface with a list of firewall rules that allows inbound traffic for cluster management, raft sync communications, docker overlay network traffic and ssh from anywhere:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
resource "google_compute_firewall" "swarm" {
name = "swarm-firewall"
network = "${google_compute_network.swarm.name}"

allow {
protocol = "icmp"
}

allow {
protocol = "tcp"
ports = ["22", "2377", "7946"]
}

allow {
protocol = "udp"
ports = ["7946", "4789"]
}

source_ranges = ["0.0.0.0/0"]
}

resource "google_compute_network" "swarm" {
name = "swarm-network"
}

4. Automate your inventory with Terraform

In order to take automation to the next level, let’s use Terraform template_file data source to generate a dynamic Ansible inventory from Terraform state file:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
data "template_file" "inventory" {
template = "${file("templates/inventory.tpl")}"

depends_on = [
"google_compute_instance.managers",
"google_compute_instance.workers",
]

vars {
managers = "${join("\n", google_compute_instance.managers.*.network_interface.0.access_config.0.nat_ip)}"
workers = "${join("\n", google_compute_instance.workers.*.network_interface.0.access_config.0.nat_ip)}"
}
}

resource "null_resource" "cmd" {
triggers {
template_rendered = "${data.template_file.inventory.rendered}"
}

provisioner "local-exec" {
command = "echo '${data.template_file.inventory.rendered}' > ../ansible/inventory"
}
}

The template file has the following format, and it will be replaced by the Swarm managers and workers IP addresses at runtime:

1
2
3
4
5
[managers]
${managers}

[workers]
${workers}

Finally, let’s define Google Cloud to be the default provider:

1
2
3
4
5
provider "google" {
credentials = "${file("${var.credentials}")}"
project = "${var.project}"
region = "${var.region}"
}

5. Setup Ansible roles to provision instances

Once the templates are defined, we will use Ansible to provision our instances and turn them to a Swarm cluster. Hence, I created 3 Ansible roles:

  • python: as its name implies, it will install Python on the machine. CoreOS ships only with the basics, it’s a minimal linux distribution without much except tools centered around running containers.
  • swarm-init: execute the docker swarm init command on the first manager and store the swarm join tokens.
  • swarm-join: join the node to the cluster using the token generated previously.

By now, your main playbook will look something like:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
---
- name: Install Python
hosts: managers:workers
gather_facts: False
roles:
- python

- name: Init Swarm cluster
hosts: managers
gather_facts: False
roles:
- swarm-init

- name: Join Swarm cluster
hosts: workers
gather_facts: False
vars:
token: ""
manager: ""
roles:
- swarm-join

6. Test your configuration

To test it out, open a new terminal session and issue terraform init command to download the google provider:



Create an execution plan (dry run) with the terraform plan command. It shows you things that will be created in advance, which is good for debugging and ensuring that you’re not doing anything wrong, as shown in the next screenshot:



You will be able to examine Terraform’s execution plan before you deploy it to GCP. When you’re ready, go ahead and apply the changes by issuing terraform apply command.

The following output will be displayed (some parts were cropped for brevity):



If you head back to Compute Engine Dashboard, your instances should be successfully created:



7. Create your Swarm cluster with Ansible

Now our instances are created, we need to turn them to a Swarm cluster with Ansible. Issue the following command:

1
ansible-playbook -i inventory main.yml


Next, SSH to the manager instance using it’s public IP address:



If you run docker node ls, you will get a list of nodes in the swarm:



Deploy the visualizer service with the following command:

1
2
3
docker service create --name=visualizer --publish=8080:8080/tcp \
--constraint=node.role==manager --mount=type=bind,src=/var/run/docker.sock,dst=/var/run/docker.sock \
dockersamples/visualizer


8. Update your network rules

The service is exposed on port 8080 of the instance. Therefore, we need to allow inbound traffic on that port, you can use Terraform to update the existing firewall rules:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
resource "google_compute_firewall" "swarm" {
name = "swarm-firewall"
network = "${google_compute_network.swarm.name}"

allow {
protocol = "icmp"
}

allow {
protocol = "tcp"
ports = ["22", "2377", "7946", "8080"]
}

allow {
protocol = "udp"
ports = ["7946", "4789"]
}

source_ranges = ["0.0.0.0/0"]
}

resource "google_compute_network" "swarm" {
name = "swarm-network"
}

Run terraform apply again to create the new ingress rule, it will detect the changes and ask you to confirm it:



If you point your favorite browser to your http://instance_ip:8080, the following dashboard will be displayed which confirms our cluster is fully setup:



In an upcoming post, we will see how we can take this further by creating a production-ready Swarm cluster on GCP inside a VPC — and how to provision Swarm managers and workers on-demand using instance groups based on increases or decreases in load.

We will also learn how to bake a CoreOS machine image with Python preinstalled with Packer, and how to use Terraform and Jenkins to automate the infrastructure deployment!

Drop your comments, feedback, or suggestions below — or connect with me directly on Twitter @mlabouardy.

Build a Ruby based Lambda Function

At AWS re:Invent 2018, it was announced that Ruby is now a supported language for AWS Lambda. In this post, I walk you through how to write your very first Ruby-based Lambda function from scratch, followed by how to configure, deploy, and test a Lambda function.



API Gateway will forward incoming requests to the target Ruby based Lambda function, which will call the corresponding DynamoDB operation on the movies table.

To get started, create a Lambda execution role with permission to invoke the Scan operation on the DynamoDB table:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
{
"Version": "2012-10-17",
"Statement": [
{
"Sid": "1",
"Effect": "Allow",
"Action": "dynamodb:Scan",
"Resource": [
"arn:aws:dynamodb:eu-west-3:*:table/movies",
"arn:aws:dynamodb:eu-west-3:*:table/movies/index/*"
]
}
]
}

The function entry-point below is is self explanatory, it uses the AWS SDK (the package is pre-installed in Lambda) to instantiate a DynamoDB client in the appropriate region and issues the Scan operation on the DynamoDB table (defined in an environment variable):

1
2
3
4
5
6
7
8
9
10
11
require 'aws-sdk'
require 'json'

def lambda_handler(event:, context:)
dynamodb = Aws::DynamoDB::Client.new(region: ENV['AWS_REGION'])

resp = dynamodb.scan({
table_name: ENV['TABLE_NAME'],
})
{ statusCode: 200, body: JSON.generate(resp.items) }
end

The AWS SDK for Ruby is included in the Lambda execution environment by default.

Now that our handler is defined, head to the Lambda form creation and select the IAM role (you might need to refresh the page for the changes to take effect) from the Existing role drop-down list. Then, click the Create function button:



Set the table name as an environment variable:



The movies table contains a set of movies:



Create a deployment package (zip file) and update the function’s code using the AWS CLI command:

1
2
zip -r deployment.zip handler.rb
aws lambda update-function-code --function-name ScanMovies --zip-file fileb://./deployment.zip

Make sure to set the Lambda function handler to handler.lambda_handler

Once the function has been deployed, invoke it manually using the sample event data by clicking on the “Test” button in the top right of the console.



So far, we learned how to build our first Lambda function with Ruby. We also learned how to invoke it manually from the console. To leverage the power of Lambda, we are going to learn how to trigger this Lambda function in response to incoming HTTP requests (event-driven architecture) using the AWS API Gateway service:



Create a deployment stage and open your favorite browser with the API Invoke URL; you should see a message like the one shown in the following screenshot:



The following screenshot shows a properly configured Ruby based Lambda function with IAM access to DynamoDB:



Like what you’re read­ing? Check out my book and learn how to build, secure, deploy and manage production-ready Serverless applications in Golang with AWS Lambda.

Drop your comments, feedback, or suggestions below — or connect with me directly on Twitter @mlabouardy.

Full guide to building a Serverless API with zero code

A common use case of API Gateway is building API endpoints in top of Lambda functions. It can also be used as an API proxy to connect to AWS services. In this guide, I will walk you through how to create your own API using API Gateway and DynamoDB only and go through advanced features to enhance your API endpoints such as:

  • Mapping templates, Integration Request and Integration Response.
  • Error handling and request validation.
  • Authentication with AWS Cognito and Lambda Authorizer.
  • API Throttling with Plan usage and API keys.
  • API documentation generation.
  • API Gateway custom domain.

Setting up DynamoDB

To get started, create a DynamoDB table called movies with an id as a partition key (leave the read/write capacity to default values):



Next, insert few items into the table, it should look something like this:



Next, we need to grant the API Gateway access to DynamoDB table. Therefore we need to create an IAM role assumable by API Gateway:

1
2
3
4
5
6
7
8
9
10
11
12
13
{
"Version": "2012–10–17",
"Statement": [
{
"Sid": "",
"Effect": "Allow",
"Principal": {
"Service": "apigateway.amazonaws.com"
},
"Action": "sts:AssumeRole"
}
]
}

The role will give API Gateway permission to invoke the following DynamoDB operations on movies table:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
{
"Version": "2012–10–17",
"Statement": [
{
"Sid": "VisualEditor0",
"Effect": "Allow",
"Action": [
"dynamodb:PutItem",
"dynamodb:GetItem",
"dynamodb:DeleteItem",
"dynamodb:Scan",
"dynamodb:Query"
],
"Resource": [
"arn:aws:dynamodb:eu-west-3:*:table/movies",
"arn:aws:dynamodb:eu-west-3:*:table/movies/index/*"
]
}
]
}

API Endpoints

Before going into further detail about the architecture, the following diagram shows how API Gateway and DynamoDB will fit into the API architecture:



When calling the API endpoints, the request will go through the API Gateway, which will invoke the appropriate DynamoDB operation. This returns a response which is proxied by the API Gateway to the client in a JSON format.

GET /MOVIES

Create new API called MoviesAPI from API Gateway Console, and create a new resource, let’s call it movies:



Expose a GET method on /movies resource by clicking on “Create Method”. Select AWS Service under the “Integration type” section, choose the DynamoDB service, set the HTTP method to be POST and action type to be a Scan operation.



Next, we need to transform the HTTP request coming into API Gateway to a proper Scan request for DynamoDB. In the API Gateway console, select the “Integration Request”. All the way at the bottom we can select the Body Mapping Templates. Here, create a new application/json mapping template with the following configuration:



Deploy the API from “Actions” and create a new deployment stage, an invocation URL will be displayed:



Point your browser to the URL given or use a modern REST client like Postman. The endpoint will return a list of movies in a JSON format:



The output is returned in DynamoDB response format, in order to map the raw response to traditional JSON object structure, we will use Integration Response feature.

Click on “GET” method and navigate to “Integration Response”, expand the 200 response code. Expand the “Mapping Templates” section. In Content-Type choose application/json and create a mapping template that loop through each item from the Items array, extracts the relevant attributes of the movie’s item and places them into a response structure:



Mapping template is a script expressed in Velocity Template Language (VTL) and applied to the payload using JSONPath expressions.

As a result, you should now see a formatted response.



GET /MOVIES/:ID

The second endpoint will be responsible of fetching a movie based on an ID provided by the client. Hence, a new resource with a path parameter should be created. The value of ID will be made available via the $input.params(‘id’) method:



Expose a GET method, and then link the resource to the DynamoDB service. The action will be GetItem operation:



Again, specify a body mapping template for the integration request, now with the following template:



When the API URL is invoked with an ID, the movie corresponding to the ID is returned if it exists.



Similarly we will use integration response to map the raw DynamoDB response to the similar JSON object structure we defined earlier:



If you test it out once again, the following JSON will be returned:



POST /MOVIES

Now we know how the GET method works with and without path parameters. The next step will be to insert a new item to the table. Create a POST method with PutItem as an action:



We will create a mapping template to transform the client request into the structure that the DynamoDB API PutItem requires. The below mapping template creates the JSON structure required by the DynamoDB PutItem API. The three input variables are referenced from the request JSON using the $input variable:



Back in the “Method Execution” pane click “TEST”. Create an example request body that matches the API definition documented above and then choose “Test”. For example, your request body could be:



Navigate to the DynamoDB console and view the movies table to show that the request really was successfully processed:



Try to insert a new movie without giving a movie’s name attribute. The following error will returned:



It’s a DynamoDB PutItem error. Fortuently, API Gateway allows you to validate your request body before invoking the downstream resources (In our example the DynamoDB table). To achieve this, we will use API Gateway Models. A Model defines the payload data structure. Models definitions are written using JSON Schema draft 4.

In the API Gateway, navigate to the Models tab and create a new model. Fill in the form as so:



The model above defines a movie entity with 3 attributes and requires id and name attributes to be defined (used during validation).

Head back to “Resources” page and click on “Method Request” from the POST method, enable the request validator option as below:



If you try to insert a new movie without providing the required parameters, a bad request message error will be returned:



You can override the default 400 message from the “Gateway Responses” as follows:



As a result, the user defined error message will be returned:



Great! Try implementing the PUT and DELETE methods:



Authentication

The serverless API that we have built so far works like a charms. However, its open to the public, anyone can insert data into DynamoDB table if he/she has the API Gateway invocation URL. Luckily, API Gateway offers two ways to handle authentication:



API Gateway Authentication with Cognito and Lambda Authorizer

AMAZON COGNITO

Create a new user pool, click on “Review defaults” to create a pool with default settings. A success message should be displayed at the end of the creation process:



After creating your first user pool, register your serverless API from “App clients” under “General settings” and select “Add an app client”. Give the application a name and check the server-based authentication ADMIN_NO_SRP_AUTH option:



Create a new user using the AWS command line:

1
2
3
4
5
6
7
# Create a user
aws cognito-idp sign-up -region AWS_REGION -client-id CLIENT_ID \
-username USERNAME -password PASSWORD -user-attributes Name=email,Value=EMAIL

# Confirm sign up
aws cognito-idp admin-confirm-sign-up -region AWS_REGION -user-pool-id USER_POOL \
-username USERNAME

Now that the user pool has been created, we can configure the API Gateway to validate access tokens from a successful user pool authentication before granting access to DynamoDB.



To begin securing API access, go to API Gateway console, choose the RESTful API that we built in the previously, and click on “Authorizers” from the navigation bar. Click on the “Create New Authorizer” button and select “Cognito”. Then, select the user pool that we created earlier and set the token source field to Authorization. This defines the name of the incoming request header containing the API caller’s identity token for Authorization:



You can now secure all of the endpoints, for instance, in order to secure the endpoint responsible for creating an new movie. Click on the corresponding POST method under the /movies resource. Click on the “Method Request” box, then on “Authorization”, and select the user pool we created previously:



Once done, redeploy the API and try to insert a new movie using the API invocation URL. This time, the endpoint is secured and requires authentication:



In order to authenticate, we need to obtain an identity token for the signed-in user from the the user pool and include the identity token in the Authorization header for the API Gateway requests. Issue the following AWS CLI command to get a new token:

1
aws cognito-idp admin-initiate-auth -region AWS_REGION -cli-input-json file://input.json

The command above takes a JSON file with the following attributes:

1
2
3
4
5
6
7
8
9
{
"UserPoolId": "USER_POOL",
"ClientId": "CLIENT_ID",
"AuthFlow": "ADMIN_NO_SRP_AUTH",
"AuthParameters": {
"USERNAME": "USERNAME",
"PASSWORD": "PASSWORD"
}
}

Once executed, the preceding command will return the following JSON response:

1
2
3
4
5
6
7
8
9
10
{
"AuthenticationResult": {
"ExpiresIn": 3600,
"IdToken": "ID_TOKEN",
"RefreshToken": "REFRESH_TOKEN",
"TokenType": "Bearer",
"AccessToken": "ACCESS_TOKEN"
},
"ChallengeParameters": {}
}

Copy the ID token and add it to the Authorization header of your request:



The API Gateway will verify the token and will invoke the PutItem operation on the movies table, which will insert a new movie into the table:



LAMBDA AUTHORIZER

When a client sends a request to your API, it will go through the API Gateway, which will extracts the token from the request and calls your Lambda function authorizer with it. The function evaluates the token, generates a policy and sends it back to API Gateway. API Gateway evaluates the policy and invoke the DynamoDB action registered for the API endpoint.

For the sake of simplicity, our function will verify if the token provided by the client equals to our secret (environment variable) and returns a policy document based on the result. The following is the function handler source code written in Node.JS:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
const TOKEN = process.env.TOKEN;

const generatePolicy = (effect, methodArn) => {
return {
'policyDocument': {
'Version': '2012-10-17',
'Statement': [
{
'Sid': '1',
'Action': 'execute-api:Invoke',
'Effect': effect,
'Resource': methodArn
}
]
}
}
}

exports.handler = async (event, context) => {
if(event.authorizationToken == TOKEN){
return generatePolicy('ALLOW', event.methodArn)
}
return generatePolicy('DENY', event.methodArn)
}

Head back to API Gateway and created a new “Lambda Authorizer” and set Authorization to be the header API Gateway will extract the token from:



Choose the method you want to secure, let’s say, it will be the endpoint responsible of deleting a movie from the table. Click on “Method Request” and under Authorization select your new authorizer:



Let’s try calling the endpoint, As expected, we’re not getting through to our real endpoint:



If you include the secret token to the Authorization header of your request, you should be able to delete an item:



Looks good!

API Throttling

You can use usage plans combined with API keys to set method-level throttling limits for your API and define how much and how fast clients can access your API (request rates and quotas).

The following procedure describes how to create a usage plan:

API USAGE

Create a usage plan called basic, with a throttling limit of 1 request per second and quota limit of 10000 requests per day:



Create a 2nd usage plan called premium, with a throlling limit of 10 requests per second and a quota limit of 1 million requests per day:



API KEYS

Next, create two API keys:



Assign the first API key to basic usage plan and second key to premium usage plan:



Associate the usage plans we created to the API deployment stage:



Configure an API method to require an API key:



Deploy or redeploy the API for the requirement to take effect:



Now if you added the x-api-key header. If all goes well you will receive output like this:



If you exceed the rate limit or quota limit associated with your API key, a “Too many requests” HTTP error will be returned:



Custom Domains

You can use your own domain name for an API and deployment stage, create a Custom Domain Name backed by an ACM (Amazon Certificate Manager) certificate:



Create a new custom domain name from API Gateway Console:



Add a path mapping to map your domain name to your API deployment stage:



Once configured, you can query your API using your custom domain name as follows: https://api.serverlessmovies.com/movies

Documentation

Before finishing this guide, we will go through how to create documentation for the serverless API we’ve built so far.

On the API Gateway console, select the deployment stage that you’re interested in generating documentation for. In the following example, I chose the sandbox environment. Then, click on the Export tab and click on the Export as Swagger section:



Swagger is an implementation of the OpenAPI, which is a standard defined by the Linux Foundation on how to describe and define APIs. This definition is called the OpenAPI specification document.

You can save the document in either a JSON or YAML file. Then, navigate to https://editor.swagger.io/ and paste the content on the website editor, it will be compiled and an HTML page will be generated as follows:



Like what you’re read­ing? Check out my book and learn how to build, secure, deploy and manage production-ready Serverless applications in Golang with AWS Lambda.

Drop your comments, feedback, or suggestions below — or connect with me directly on Twitter @mlabouardy.

Build real-world, production-ready applications with AWS Lambda

Serverless architecture is popular in the tech community due to AWS Lambda. Go is simple to learn, straightforward to work with, and easy to read for other developers; and now it’s been heralded as a supported language for AWS Lambda. This book is your optimal guide to designing a Go serverless application and deploying it to Lambda.



This book starts with a quick introduction to the world of serverless architecture and its benefits, and then delves into AWS Lambda using practical examples. You’ll then learn how to design and build a production-ready application in Go using AWS serverless services with zero upfront infrastructure investment. The book will help you learn how to scale up serverless applications and handle distributed serverless systems in production. You will also learn how to log and test your application.

Along the way, you’ll also discover how to set up a CI/CD pipeline to automate the deployment process of your Lambda functions. Moreover, you’ll learn how to troubleshoot and monitor your apps in near real-time with services such as AWS CloudWatch and X-ray. This book will also teach you how to secure the access with AWS Cognito.

By the end of this book, you will have mastered designing, building, and deploying a Go serverless application.

Hands-On Serverless Applications with Go is available at the online stores below:







CI/CD for Lambda Functions with Jenkins

The following post will walk you through how to build a CI/CD pipeline to automate the deployment process of your Serverless applications and how to use features like code promotion, rollbacks, versions, aliases and blue/green deployment. At the end of this post, you will be able to build a pipeline similar to the following figure:



For the sake of simplicity, I wrote a simple Go based Lambda function that calculates the Fibonacci number:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
package main

import (
"errors"

"github.com/aws/aws-lambda-go/lambda"
)

func fibonacci(n int) int {
if n <= 1 {
return n
}
return fibonacci(n-1) + fibonacci(n-2)
}

func handler(n int) (int, error) {
if n < 0 {
return -1, errors.New("Input must be a positive number")
}
return fibonacci(n), nil
}

func main() {
lambda.Start(handler)
}

I implemented also a couple of unit tests for both the Fibonacci recursive and Lambda handler functions:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
package main

import (
"errors"
"testing"

"github.com/stretchr/testify/assert"
)

func TestFibonnaciInputLessOrEqualToOne(t *testing.T) {
assert.Equal(t, 1, fibonacci(1))
}

func TestFibonnaciInputGreatherThanOne(t *testing.T) {
assert.Equal(t, 13, fibonacci(7))
}

func TestHandlerNegativeNumber(t *testing.T) {
responseNumber, responseError := handler(-1)
assert.Equal(t, -1, responseNumber)
assert.Equal(t, errors.New("Input must be a positive number"), responseError)
}

func TestHandlerPositiveNumber(t *testing.T) {
responseNumber, responseError := handler(5)
assert.Equal(t, 5, responseNumber)
assert.Nil(t, responseError)
}

To create the function in AWS Lambda and all the necessary AWS services, I used Terraform. An S3 bucket is needed to store all the deployment packages generated through the development lifecycle of the Lambda function:

1
2
3
4
5
// S3 bucket
resource "aws_s3_bucket" "bucket" {
bucket = "${var.bucket}"
acl = "private"
}

The build server needs to interact with S3 bucket and Lambda functions. Therefore, an IAM instance role must be created with S3 and Lambda permissions:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
// Jenkins slave instance profile
resource "aws_iam_instance_profile" "worker_profile" {
name = "JenkinsWorkerProfile"
role = "${aws_iam_role.worker_role.name}"
}

resource "aws_iam_role" "worker_role" {
name = "JenkinsBuildRole"
path = "/"

assume_role_policy = <<EOF
{
"Version": "2012-10-17",
"Statement": [
{
"Action": "sts:AssumeRole",
"Principal": {
"Service": "ec2.amazonaws.com"
},
"Effect": "Allow",
"Sid": ""
}
]
}
EOF
}

resource "aws_iam_policy" "s3_policy" {
name = "PushToS3Policy"
path = "/"

policy = <<EOF
{
"Version": "2012-10-17",
"Statement": [
{
"Action": [
"s3:PutObject",
"s3:GetObject"
],
"Effect": "Allow",
"Resource": "${aws_s3_bucket.bucket.arn}/*"
}
]
}
EOF
}

resource "aws_iam_policy" "lambda_policy" {
name = "DeployLambdaPolicy"
path = "/"

policy = <<EOF
{
"Version": "2012-10-17",
"Statement": [
{
"Action": [
"lambda:UpdateFunctionCode",
"lambda:PublishVersion",
"lambda:UpdateAlias"
],
"Effect": "Allow",
"Resource": "*"
}
]
}
EOF
}

resource "aws_iam_role_policy_attachment" "worker_s3_attachment" {
role = "${aws_iam_role.worker_role.name}"
policy_arn = "${aws_iam_policy.s3_policy.arn}"
}

resource "aws_iam_role_policy_attachment" "worker_lambda_attachment" {
role = "${aws_iam_role.worker_role.name}"
policy_arn = "${aws_iam_policy.lambda_policy.arn}"
}

An IAM role is needed for the Lambda function as well:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
// Lambda IAM role
resource "aws_iam_role" "lambda_role" {
name = "FibonacciFunctionRole"
path = "/"

assume_role_policy = <<EOF
{
"Version": "2012-10-17",
"Statement": [
{
"Action": "sts:AssumeRole",
"Principal": {
"Service": "lambda.amazonaws.com"
},
"Effect": "Allow",
"Sid": ""
}
]
}
EOF
}

Finally, a Go-based Lambda function will be created with the following properties:

1
2
3
4
5
6
7
8
// Lambda function
resource "aws_lambda_function" "function" {
filename = "deployment.zip"
function_name = "Fibonacci"
role = "${aws_iam_role.lambda_role.arn}"
handler = "main"
runtime = "go1.x"
}

Next, build the deployment package with the following commands:

1
2
3
4
# Build linux binary
GOOS=linux go build -o main main.go
# Create a zip file
zip deployment.zip main

Then, issue the terraform apply command to create the resources:



Sign in to AWS Management Console and navigate to Lambda Console, a new function called “Fibonacci” should be created:



You can test it out, by mocking the input from the “Select a test event” dropdown list:



If you click on “Test” button the Fibonacci number of 7 will be returned:



So far our function is working as expected. However, how can we ensure each changes to our codebase doesn’t break things ? That’s where CI/CD comes into play, the idea is making all code changes and features go through a complex pipeline before integrating them to the master branch and deploying it to production.

You need a Jenkins cluster with at least a single worker (with Go preinstalled), you can follow my previous post for a step by step guide on how to build a Jenkins cluster on AWS from scratch.

Prior to the build, the IAM instance role (created with Terraform) with the write access to S3 and the update operations to Lambda must be configured on the Jenkins workers:



Jump back to Jenkins Dashboard and create new multi-branch project and configure the GitHub repository where the code source is versioned as follows:



Create a new file called Jenkinsfile, it defines a set of steps that will be executed on Jenkins (This definition file must be committed to the Lambda function’s code repository):

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
def bucket = 'deployment-packages-mlabouardy'
def functionName = 'Fibonacci'
def region = 'eu-west-3'

node('slaves'){
stage('Checkout'){
checkout scm
}

stage('Test'){
sh 'go get -u github.com/golang/lint/golint'
sh 'go get -t ./...'
sh 'golint -set_exit_status'
sh 'go vet .'
sh 'go test .'
}

stage('Build'){
sh 'GOOS=linux go build -o main main.go'
sh "zip ${commitID()}.zip main"
}

stage('Push'){
sh "aws s3 cp ${commitID()}.zip s3://${bucket}"
}

stage('Deploy'){
sh "aws lambda update-function-code --function-name ${functionName} \
--s3-bucket ${bucket} \
--s3-key ${commitID()}.zip \
--region ${region}"
}
}

def commitID() {
sh 'git rev-parse HEAD > .git/commitID'
def commitID = readFile('.git/commitID').trim()
sh 'rm .git/commitID'
commitID
}

The pipeline is divided into 5 stages:

  • Checkout: clone the GitHub repository.
  • Test: check whether our code is well formatted and follows Go best practices and run unit tests.
  • Build: build a binary and create the deployment package.
  • Push: store the deployment package (.zip file) to an S3 bucket.
  • Deploy: update the Lambda function’s code with the new artifact.

Note the usage of the git commit ID as a name for the deployment package to give a meaningful and significant name for each release and be able to roll back to a specific commit if things go wrong.

Once the project is saved, a new pipeline should be created as follows:



Once the pipeline is completed, all stages should be passed, as shown in the next screenshot:



At the end, Jenkins will update the Lambda function’s code with the update-function-code command:



If you open the S3 Console, then click on the bucket used by the pipeline, a new deployment package should be stored with a key name identical to the commit ID:



Finally, to make Jenkins trigger the build when you push to the code repository, click on “Settings” from your GitHub repository, then create a new webhook from “Webhooks”, and fill it in with a URL similar to the following:



In case you’re using Git branching workflows (you should), Jenkins will discover automatically the new branches:



Hence, you must separate your deployment environments to test new changes without impacting your production. Therefore, having multiple versions of your Lambda functions makes sense.

Update the Jenkinsfile to add a new stage to publish a new Lambda function’s version, every-time you push (or merge) to the master branch:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
def bucket = 'deployment-packages-mlabouardy'
def functionName = 'Fibonacci'
def region = 'eu-west-3'

node('slaves'){
stage('Checkout'){
checkout scm
}

stage('Test'){
sh 'go get -u github.com/golang/lint/golint'
sh 'go get -t ./...'
sh 'golint -set_exit_status'
sh 'go vet .'
sh 'go test .'
}

stage('Build'){
sh 'GOOS=linux go build -o main main.go'
sh "zip ${commitID()}.zip main"
}

stage('Push'){
sh "aws s3 cp ${commitID()}.zip s3://${bucket}"
}

stage('Deploy'){
sh "aws lambda update-function-code --function-name ${functionName} \
--s3-bucket ${bucket} \
--s3-key ${commitID()}.zip \
--region ${region}"
}

if (env.BRANCH_NAME == 'master') {
stage('Publish') {
sh "aws lambda publish-version --function-name ${functionName} \
--region ${region}"
}
}
}

def commitID() {
sh 'git rev-parse HEAD > .git/commitID'
def commitID = readFile('.git/commitID').trim()
sh 'rm .git/commitID'
commitID
}

On the master branch, a new stage called “Published” will be added:



As a result, a new version will be published based on the master branch source code:



However, in agile based environment (Extreme programming). The development team needs to release iterative versions of the system often to help the customer to gain confidence in the progress of the project, receive feedback and detect bugs in earlier stage of development. As a result, small releases can be frequent:



AWS services using Lambda functions as downstream resources (API Gateway as an example) need to be updated every-time a new version is published -> operational overhead and downtime. USE aliases !!!

The alias is a pointer to a specific version, it allows you to promote a function from one environment to another (such as staging to production). Aliases are mutable, unlike versions, which are immutable.

That being said, create an alias for the production environment that points to the latest version published using the AWS command line:

1
2
3
aws lambda create-alias --function-name Fibonacci \
--name production --function-version 2 \
--region eu-west-3

You can now easily promote the latest version published into production by updating the production alias pointer’s value:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
def bucket = 'deployment-packages-mlabouardy'
def functionName = 'Fibonacci'
def region = 'eu-west-3'

node('slaves'){
stage('Checkout'){
checkout scm
}

stage('Test'){
sh 'go get -u github.com/golang/lint/golint'
sh 'go get -t ./...'
sh 'golint -set_exit_status'
sh 'go vet .'
sh 'go test .'
}

stage('Build'){
sh 'GOOS=linux go build -o main main.go'
sh "zip ${commitID()}.zip main"
}

stage('Push'){
sh "aws s3 cp ${commitID()}.zip s3://${bucket}"
}

stage('Deploy'){
sh "aws lambda update-function-code --function-name ${functionName} \
--s3-bucket ${bucket} \
--s3-key ${commitID()}.zip \
--region ${region}"
}

if (env.BRANCH_NAME == 'master') {
stage('Publish') {
def lambdaVersion = sh(
script: "aws lambda publish-version --function-name ${functionName} --region ${region} | jq -r '.Version'",
returnStdout: true
)
sh "aws lambda update-alias --function-name ${functionName} --name production --region ${region} --function-version ${lambdaVersion}"
}
}
}

def commitID() {
sh 'git rev-parse HEAD > .git/commitID'
def commitID = readFile('.git/commitID').trim()
sh 'rm .git/commitID'
commitID
}

Like what you’re read­ing? Check out my book and learn how to build, secure, deploy and manage production-ready Serverless applications in Golang with AWS Lambda.

Drop your comments, feedback, or suggestions below — or connect with me directly on Twitter @mlabouardy.

Your browser is out-of-date!

Update your browser to view this website correctly. Update my browser now

×