How to troubleshoot logging to cloudwatch from docker in an ECS container? - amazon-cloudwatch

Question

I'm tearing my hair out over this.
Here's my setup:
ECS Container on an EC2 Instance that contains an
ECS Task Definition that runs a
Docker instance with a python script that logs to stderr
I see a Cloudwatch log group fo the ECS Task get created, but nothing I print to stderr appears.
My ECS container has the default ecsInstanceRole.
The ecsInstanceRole has the AmazonEC2ContainerServiceforEC2Role policy which is as follows:
{
"Version": "2012-10-17",
"Statement": [
{
"Effect": "Allow",
"Action": [
"ecs:CreateCluster",
"ecs:DeregisterContainerInstance",
"ecs:DiscoverPollEndpoint",
"ecs:Poll",
"ecs:RegisterContainerInstance",
"ecs:StartTelemetrySession",
"ecs:UpdateContainerInstancesState",
"ecs:Submit*",
"ecr:GetAuthorizationToken",
"ecr:BatchCheckLayerAvailability",
"ecr:GetDownloadUrlForLayer",
"ecr:BatchGetImage",
"logs:CreateLogStream",
"logs:PutLogEvents"
],
"Resource": "*"
}
]
}
Task definition
{
"containerDefinitions": [
{
"dnsSearchDomains": [],
"logConfiguration": {
"logDriver": "awslogs",
"options": {
"awslogs-group": "torres",
"awslogs-region": "us-west-2",
"awslogs-stream-prefix": "torres"
}
},
"command": [
"sleep",
"infinity"
],
"cpu": 100,
"mountPoints": [
{
"readOnly": null,
"containerPath": "/mnt/efs",
"sourceVolume": "efs"
}
],
"memoryReservation": 512,
"image": "X.dkr.ecr.us-west-2.amazonaws.com/torres-docker:latest",
"interactive": true,
"essential": true,
"pseudoTerminal": true,
"readonlyRootFilesystem": false,
"privileged": false,
"name": "torres"
}
],
"family": "torres",
"volumes": [
{
"name": "efs",
"host": {
"sourcePath": "/mnt/efs"
}
}
]
}

Answer 1

I figured out the issue, but I don't know why it solves it.
When I changed my command from sleep infinity to just running a piece of code that logs python script.py the logging in CloudWatch started working.
Before I was launching the container with sleep infinity ssh-ing into the container and launching script.py from the shell and this was NOT logging.