Simply speaking, ECS provides container (docker) service on top of EC2 without extra cost. So most EC2 supported features are available in ECS as well, eg. clustering, ELB, auto scale and monitoring.

I created my first ECS container instance in the US region, and had a bit play with it. Here are my notes:

1) AWS provides ECS-optimized AMI. It basically is a EC2 instance – Amazon Linux with docker installed, and has the ecs agent which itself is a container as well. User should be able to use their own AMI as ECS container instance, as long as docker (1.5+) and AWS ECS container agent are installed. Also the ecs-agent container needs to be created as following (the cluster name is kept in the file /etc/ecs/ecs.config). I will try to use RHEL7 as the container host on AWS, see how it goes.

touch /etc/ecs/ecs.config; mkdir -p /var/log/ecs; docker run –name ecs-agent -d -v /var/run/docker.sock:/var/run/docker.sock -v /var/log/ecs:/log -p 127.0.0.1:51678:51678 –env-file /etc/ecs/ecs.config -e ECS_LOGFILE=/log/ecs-agent.log amazon/amazon-ecs-agent

2) ECS uses json file to create task definition which includes one or more containers, it allows allocating CPU/RAM usage to each container, inter-connection between containers, mounting volume from host or other containers on the same host, port mapping… This is a sample of a task definition. It has two containers, one is the web server, the other is the busy box which generates a simple html page to display on the web-server container.

{
“taskDefinitionArn”: “arn:aws:ecs:us-west-2:092666830250:task-definition/console-sample-app-static:1”,
“revision”: 1,
“containerDefinitions”: [
{
“volumesFrom”: [],
“portMappings”: [
{
“hostPort”: 80,
“containerPort”: 80
}
],
“command”: null,
“environment”: [],
“essential”: true,
“entryPoint”: null,
“links”: [],
“mountPoints”: [
{
“containerPath”: “/usr/local/apache2/htdocs”,
“sourceVolume”: “my-vol”,
“readOnly”: null
}
],
“memory”: 300,
“name”: “simple-app”,
“cpu”: 10,
“image”: “httpd:2.4”
},
{
“volumesFrom”: [
{
“readOnly”: null,
“sourceContainer”: “simple-app”
}
],
“portMappings”: [],
“command”: [
“/bin/sh -c \”while true; do echo ‘Amazon ECS Sample App

‘ > bottom; cat top date bottom > /usr/local/apache2/htdocs/index.html ; sleep 1; done\””
],
“environment”: [],
“essential”: false,
“entryPoint”: [
“sh”,
“-c”
],
“links”: [],
“mountPoints”: [],
“memory”: 200,
“name”: “busybox”,
“cpu”: 10,
“image”: “busybox”
}
],
“volumes”: [
{
“host”: {
“sourcePath”: null
},
“name”: “my-vol”
}
],
“family”: “console-sample-app-static”
}

3) I guess you can not patch the container that is created from AWS docker images, as I don’t see any yum file or directory within the container that is based on httpd:latest image. Replacing the container with a new one might be their way. In the in-house environment, I can patch the container that is created from the Redhat provided RHEL docker image.

This is on AWS:
hostname; ip addr show eth0
6ecda488108c
44: eth0: <BROADCAST,UP,LOWER_UP> mtu 9001 qdisc noqueue state UP group default
link/ether 02:42:ac:11:00:17 brd ff:ff:ff:ff:ff:ff
inet 172.17.0.23/16 scope global eth0
valid_lft forever preferred_lft forever
inet6 fe80::42:acff:fe11:17/64 scope link
valid_lft forever preferred_lft forever
ls /etc/yum*
ls: cannot access /etc/yum*: No such file or directory

This is in-house environment:
hostname; ip addr show eth0
fa3339efba15
48: eth0: <BROADCAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default
link/ether 02:42:ac:11:00:18 brd ff:ff:ff:ff:ff:ff
inet 172.17.0.24/16 scope global eth0
valid_lft forever preferred_lft forever
inet6 fe80::42:acff:fe11:18/64 scope link
valid_lft forever preferred_lft foreveryum update
Loaded plugins: product-id, subscription-manager
Resolving Dependencies
–> Running transaction check
—> Package libxml2.x86_64 0:2.9.1-5.el7_0.1 will be updated
—> Package libxml2.x86_64 0:2.9.1-5.el7_1.2 will be an update
—> Package libxml2-python.x86_64 0:2.9.1-5.el7_0.1 will be updated

4) I noticed that the task/containers that has been deleted from the AWS console are actually still on the host, but with ‘exited’ status. I can start the container in the host by typing ‘docker start ‘, but the AWS console does not show the manually started container. So always using the console might be a better idea to keep the environment consistent.

5) To support ECS, AWS CLI has to be 1.7.21 or greater.

[jchen@mylab]$ aws –version
aws-cli/1.7.23 Python/2.7.8 Linux/3.19.3-100.fc20.x86_64

[jchen@mylab software]$ aws ecs list-clusters
{
“clusterArns”: [
“arn:aws:ecs:us-west-2:092666830250:cluster/default”
]
}
[jchen@mylab software]$ aws ecs list-services
{
“serviceArns”: [
“arn:aws:ecs:us-west-2:092666830250:service/sample-web”
]
}