AWS Fargate

What is AWS Fargate and Why Do You Need It?

AWS Fargate is the serverless container service/technology provided by AWS. It is also compatible with both the Elastic Container Service and Elastic Kubernetes services. Fargate can quickly launch tens of thousands of container instances/tasks and will seamlessly scale to meet our application compute requirements.

Components

1. Cluster
2. Task Definition
3. Task
4. Service
5. Repositories

Clusters

• Clusters contain services that run Tasks or servers.
• The task is considered a Serverless Container Provisioned by AWS Fargate whereas the Server meant the Ec2 instance provisioned by us.
• Clusters are often considered logical entities for the collection of tasks or servers in a service.

Task definitions

• Task definition is used as a blueprint to run our application in fargate tasks.
• Task definition primarily includes the Docker image of our application.
• It also includes how much VCpu (virtual CPU )and memory (RAM) should use for each task or container along with many other attributes.

Reference:Task Definition Parameters

Tasks

The task is the containers server launched by AWS Fargate based on the task definition we defined

Services

• Services help us to maintain and run the desired /specified number of container tasks or servers as per our needs in a cluster.
• Load balancer can be used in services to distribute the traffic among the task within the services.
• ECS Services is using a rollback deployment strategy for every task definition revision updated in services by default.

For Example:

Assume, we are having an ECS service named service-1 which is having task definition named task-def-1:1. Service is running with 1 task If the developer/Devops engineer made the task revision from task-def-1:1 to task-def-1:2 and apply the changes in services ECS service will launch the new task. Once the new task will become healthy, then only, the existing or old task will replace and terminate. If the new task failed to launch, the old task will be running state until the new task becomes healthy. Thus we can achieve zero downtime for our deployment.

We can also choose Blue/Green deployment with CodeDeploy or External deployment while creating services.

Repositories

• Repositories contain docker image URLs that will be used in task definition to deploy the application.
• Repositories can be public or private.

Operating System and CPU Architecture

There are two architectures and 3 Operating systems supported by AWS fargate to containerize our application.

• We don’t need to buy any license from Microsoft for using windows containers. The license fee will include in the bill.
• AWS will take care of OS License Management.

Working of AWS Fargate

AWS Fargate working components can be majorly classified as below mentioned three components

1. Front-end service
2. Cluster Manger-sub system
3. Capacity Manager-sub system

Front-end service

• This is the entry point (public endpoints) for aws fargate.
• The service is responsible for performing IAM authentication/authorization and limit enforcement.

Cluster Manager-Sub system

This is the backend service of aws fargate working under the hood which keeps track of clusters and tasks which are running within the cluster

Capacity Manager-Sub system

• This service is responsible for picking available instances which are in the warm pool and available instances that are ready to launch.
• This service is also responsible for picking the correct instance type and size based on the task definition defined in the services.
• It controls the replenishing of instance/ task.

Sample workflow under the hood of aws fargate

Windows containers on AWS Fargate Considerations

• We won’t be able login into the windows container server as aws fargate does not support direct access to RDP or RPE to the underlying OS.
• However, using ALB in the ECS services as Front-end, we can expose our windows container task in the target group as a backend to ALB.
• We will be able to access the windows server via the ALB DNS URL.
• Using the AWS logs driver in the windows container, we will be able to send log stream information to CloudWatch Logs.

Reference 1

Reference 2

AWS Fargate Demo ( Deploy a Simple Web Application on Amazon ECS using AWS Fargate Launch Type)

Remember the below workflow once you created the cluster in aws ECS for fargate.

Repositories–>Task definition–>Services

For custom images or private images Create an ECS cluster with fargate as compute options Step 1: Create a repositories Step 2: Create a task definition Step 3: Create a services

For this demo , we are using public docker image only so, we don't need to create repositories.

Cluster

• Login to aws console and choose ECS or Elastic container service.
• In the left navigation pane, click cluster and click create cluster.
• Enter the desired cluster name and choose the vpc and subnets.

Pricing

Pricing will based on the below-mentioned 5 independent dimensions we configured in our aws fargate services.

They are Vcpu, Memory, Storage, Operating system and Cpu Architecture.

By Default, 20GB of Ephermal storage will be provided by Fargate task and pods.

Bills are charged only for additional storage we configure.