Deploying Ecs with CodePipelin may fail and time out due to lack of CPU units

My configuration is to upload an Image from CircleCI to ECR, catch the changes in Codepipeline and deploy to ECS.

At that time, an error occurs when deploying to ECS.

service dev-test was unable to place a task because no container instance met all of its requirements. The closest matching container-instance xxxxxxxx has insufficient CPU units available. For more information, see the Troubleshooting section of the Amazon ECS Developer Guide.

I can expect that the task cannot be executed due to lack of CPU, but this error does not occur when deploying from CircleCI to ECS or deploying manually.
What is the solution to this problem?

Also, I don't know if this is related to this issue, but I have AutoScaling set to 2 for minimum and 4 for maximum, but AutoScaling is not running even though there is not enough CPU, and I get an error.
Is this problem also related to this issue?

Container instance

CPU: .5 vCPU
Memory: 3GB

CPU free space: 512
Free Memory: 769

task_definition.json

{
    "family": "development",
    "containerDefinitions": [
        {
            "name": "nginx",
            "cpu": 512,
            "memoryReservation": 1024
        },
        {
            "name": "app",
            "cpu": 512,
            "memoryReservation": 1024
        },
        {
            "name": "aws-otel-collector",
            "cpu": 512,
            "memoryReservation": 1024
        }
    ],
    "requiresCompatibilities": [
        "EC2"
    ],
    "cpu": "1536",
    "memory": "3072",
    "runtimePlatform": {
        "cpuArchitecture": "X86_64",
        "operatingSystemFamily": "LINUX"
    }
}

I have configured the cluster's capacity provider to resolve this issue.

Required size: 2
Minimum size: 2
Maximum size: 4

AutoScaling for the service is also set.

Minimum 2
Maximum 4

This results in two container instances being launched permanently.

With this setup, I had assumed that if there was not enough CPU at deployment time, a new container instance would be launched.

However, in reality, the error I mentioned earlier was displayed and there was no indication that a scale-out had taken place.

So several questions arose.

• Would capacity providers and AutoScaling not work during deployment?
• Would increasing the CPU capacity of each container instance solve this problem?
• Is it possible to temporarily increase CPU capacity only during deployment?

If anyone can help me understand this issue, please answer.

Your ECS Task is too big and cannot fit on the EC2 instance. Try creating a new revision of the task with more CPU and memory

> I have AutoScaling set to 2 for minimum and 4 for maximum

This sounds like you have configured ECS Service autoscaling. The error is that you don't have an EC2 instance registered with the cluster that has available space to run the task. You need to configure a cluster capacity provider with an EC2 auto-scaling group in order for ECS to automatically add EC2 instances to the cluster when they are needed.