Docker on AWS
Docker on AWS
Running Containers in the Cloud
Brandon Chavis – Solutions Architect
Thomas Jones – Solutions Architect
April 2015
Amazon Web Services – Docker on AWS: Running Containers in the Cloud
April 2015
3 Introduction
3 Container Benefits
4 Speed
4 Consistency
4 Density and Resource Efficiency
5 Flexibility
6 Containers in AWS
6 AWS Elastic Beanstalk
6 Amazon EC2 Container Service
7 Container-Enabled AMIs ................................................................................ 8 Container Management .................................................................................. 8 Scheduling ................................................................................................... 10 Container Repositories ................................................................................. 10 Logging and Monitoring ................................................................................ 10 Storage ......................................................................................................... 11 Networking ................................................................................................... 11 Page 2 of 23
Container Security
12 Container Use Cases
12 Batch Jobs
12 Distributed Applications
12 Continuous Integration and Deployment
13 Platform as a Service
14 Architectural Considerations
14 Walkthough
15 Conclusion
22 Further Reading
23 Document Revisions
23 Appendix A: Supported Linux Versions
23 Amazon Web Services – Docker on AWS: Running Containers in the Cloud
April 2015
AWS is a natural complement to Linux containers because of the wide range of scalable
infrastructure services upon which containers can be deployed. AWS Elastic Beanstalk
includes integrated support for Docker containers and Amazon EC2 Container Service
(Amazon ECS) was designed from the ground up to manage Docker containers at scale.
Amazon ECS is built upon several key features, such as cluster management and
support for multiple container schedulers, to facilitate and orchestrate large-scale
deployments of containers across managed clusters of Amazon EC2 instances.
Prior to the introduction of containers, developers and administrators were often faced
with a challenging web of complex compatibility restrictions, in which an application or
workload was built specifically for its predetermined environment. If this workload
needed to be migrated, for example from bare metal to a virtual machine, or from a VM
to the cloud, or between service providers, then this generally meant rebuilding the
application or workload entirely to ensure compatibility in the new environment.
Containers were conceived as an answer to this matrix of restriction and incompatibilities
by providing a common interface for migrating applications between environments.
With the release of Docker, interest in Linux containers has rapidly increased. Docker
facilitates isolation of an application by packaging the application itself with its
dependencies, configuration files, and interfaces into a single image that can then be
moved seamlessly between different hosts, ensuring a consistent runtime environment
between test and production versions of the application. Docker applies several userlevel features to containers that make containers more accessible and useable to the
community. For example, the Docker stack provides a CLI to manage containers, ways
to provision containers declaratively or interactively, and a repository to host container
At its core, Docker is an open source project that uses several resource isolation
features of the Linux kernel to sandbox an application, its dependencies, and interfaces
inside of an atomic unit called a container. Because containers roll an application
together with its dependencies and interfaces into a single re-deployable unit, a
container can be run on any host system with the appropriate kernel components, while
shielding the application from behavioral inconsistencies due to variances in software
installed on the host. Multiple containers can be run on a single host OS without the
need of a hypervisor while being isolated from neighboring containers. This layer of
isolation introduces consistency, flexibility, and portability that enable rapid software
deployment and testing.
Page 3 of 23
Amazon Web Services – Docker on AWS: Running Containers in the Cloud
April 2015
Container Benefits
The rapid growth of Linux containers, and Docker specifically, is being fueled by the
many benefits that containers provide. These benefits are apparent across an
organization: from developers and operations, to QA. The primary benefits of Docker are
speed, consistency, density, and flexibility.
Perhaps the most highly touted benefit of containers is speed. When highlighting the
advantages of using Docker, it’s impossible to avoid referencing speed at some point in
the discussion. However, simply stating that Docker is fast sells the technology short, as
speed is apparent in both performance characteristics and in application lifecycle and
deployment benefits. These are the two major ways that the speed of containers can
help developers, system administrators, and entire organizations to act quickly.
Firstly, because the architecture of containers allows for full process isolation by using
the Linux kernel namespaces and cgroups, containers do not require a full virtualization
stack or hypervisor to interface with host hardware. Containers simply share the kernel
of the underlying host OS, and processes running in a container are unaware of their
silo. This results in no discernable performance impact for applications running inside a
Secondly, containers can be created very quickly—they generally start in less than a
second. The speed implications become quite apparent and beneficial in the
development and deployment lifecycle; due to their lightweight and modular nature,
containers can enable rapid iteration of your applications. Development speed is
improved by the ability to deconstruct applications into smaller units. These smaller units
are advantageous because of a reduction in shared resources between application
components, ideally leading to fewer compatibility issues between required libraries or
Launching a container with a new release of code can be done without significant
deployment overhead. Operational speed is improved, because code built in a container
on a developer’s local machine can be easily moved to a test server by simply moving
the container. At build time, this container can be linked to other containers required to
run the application stack.
The ability to relocate entire development environments by moving a container between
systems highlights yet another major advantage of Docker: the consistency and fidelity
of a modular development environment provides predictable results when moving code
between development, test, and production systems. By ensuring that the container
Page 4 of 23
Amazon Web Services – Docker on AWS: Running Containers in the Cloud
April 2015
encapsulates exact versions of necessary libraries and packages, it is possible to
minimize the risk of bugs due to slightly different dependency revisions.
This concept easily lends itself to a “disposable system” approach, in which patching
individual containers is less preferable than building new containers in parallel, testing,
and then replacing the old. Again, this practice helps avoid “drift” of packages across a
fleet of containers, versions of your application, or dev/test/prod environments; the end
result is more consistent, predictable, and (ideally) stable applications.
Density and Resource Efficiency
The release of Docker and the enormous community support for the project over the last
eighteen months signify another step in the direction of ever-increasing density and
modularity of computing resources. Much like the invention of blade servers, the
pervasiveness of virtual machines, or the industry shift to cloud computing, container
technology promises increases in efficiency and agility through the ability to abstract
applications further from underlying operating systems and hardware.
A common approach that highlights this change is allocation of a single service or
application to a container. A one-service-to-one-container mapping allows for minimalist
environments that are easy for ops teams to manage and for dev teams to upgrade. This
approach also allows for modular deployments of a single service or application without
affecting the other services running in different containers. For example, there is limited
risk of changing a common dependency shared between two applications or restarting a
shared application server.
Containers facilitate enhanced resource efficiency by allowing multiple containers to run
on a single system. Resource efficiency is a natural result of the isolation and allocation
techniques that containers use. Containers can easily be restricted to a certain number
of CPUs and allocated specific amounts of memory. By understanding what resources a
container needs and what resources are available to your VM or underlying host server,
it’s possible to maximize the containers running on a single host, resulting in higher
density, increased efficiency of compute resources, and less money wasted on excess
The ability to “right-size” an instance via container scheduling is now far more granular
and can amplify the benefits of running on a flexible and scalable platform such as
Amazon EC2. Perhaps instead of scaling up and down an instance, changing the
number of containers assigned to a host could be the first step of achieving optimal
utilization. One of the techniques that Docker uses to drive density is utilizing a copy-onwrite file system, which is both efficient and fast because it simply creates a pointer to
files that are shared by multiple containers. Copy-on-write creates a copy of a file only if
it is changed.
Page 5 of 23
Amazon Web Services – Docker on AWS: Running Containers in the Cloud
April 2015
Additionally, because containers are host-platform agnostic (so long as the host is a
version of Linux running the necessary Docker software) and resource efficient, they
allow for an increase in the density of applications when deployed on both VMs and
physical servers without incurring the performance loss of additional overhead.
The flexibility of Docker containers is based on their portability, ease of deployment, and
small size. Similar to Git, Docker provides a simple mechanism for developers to
download and install Docker containers and their subsequent applications using the
command docker pull. This is in contrast to the sometimes extensive installation
instructions that other applications may require.
Because Docker provides a standard interface, it makes containers easy to deploy
wherever you like, providing portability between different versions of Linux, your laptop,
or the cloud. You can run the same Docker container on any supported version of Linux,
assuming that you have the Docker stack installed on the host.
Containers also provide flexibility by making microservice architecture possible. In
contrast to common infrastructure models in which a virtual machine runs multiple
services, packaging services inside their own container on top of a host OS allows a
service to be moved between hosts, isolated from failure of other adjacent services, and
protected from errant patches or software upgrades on the host system.
Because Docker provides clean, reproducible, and modular environments, it streamlines
both code deployment and infrastructure management. Docker offers numerous benefits
for a variety of use cases, whether in development, testing, deployment, or production.
Containers in AWS
Amazon Web Services is an elastic, secure, flexible, and developer-centric ecosystem
that serves as an ideal platform for Docker deployments. AWS offers the scalable
infrastructure, APIs, and SDKs that integrate tightly into a development lifecycle and
accentuate the benefits of the lightweight and portable containers that Docker offers to
its users. In this section, we discuss two possibilities for container deployment in AWS:
AWS Elastic Beanstalk, and Amazon ECS.
AWS Elastic Beanstalk
AWS Elastic Beanstalk is a management layer for AWS services like Amazon Elastic
Compute Cloud (Amazon EC2), Amazon Relational Database Service (Amazon RDS),
and Elastic Load Balancing. AWS Elastic Beanstalk removes the requirement of
manually launching the AWS resources required to run your application. Instead, you
upload your application while AWS Elastic Beanstalk handles the details of capacity
provisioning, load balancing, scaling, and application health monitoring. You can define
Page 6 of 23
Amazon Web Services – Docker on AWS: Running Containers in the Cloud
April 2015
and configure resources programmatically using AWS APIs, the eb command line tool,
AWS CloudFormation, or the AWS Management Console.
Additionally, AWS Elastic Beanstalk offers the ability to deploy and manage
containerized applications, and the command line utility eb can be used to deploy both
the AWS Elastic Beanstalk environment and Docker containers. AWS Elastic Beanstalk
can deploy containerized applications to Amazon ECS by provisioning the cluster and
any other infrastructure components that your application may require. This means that
you can simply specify which container images are to be deployed, CPU and memory
requirements, port mappings, container links; and AWS Elastic Beanstalk places your
containers across your cluster and monitors the health of each container. With AWS
Elastic Beanstalk, you can easily deploy and scale containerized web applications and
avoid the complexities of provisioning the underlying infrastructure.
In situations where organizations want to leverage the benefits of containers and want
the simplicity of deploying applications to AWS by uploading a container image, AWS
Elastic Beanstalk may be the right choice. If more granular control over containers or
custom application architectures are needed, then consider working with Amazon ECS
Amazon ECS
While AWS Elastic Beanstalk is useful for deploying a limited number of containers, a
web application, or for the development lifecycle, the way to run and operate containerenabled applications with more flexibility and scale is by using Amazon EC2 Container
Service (Amazon ECS). Amazon ECS is designed specifically to allow you to run and
manage containers across a number of hosts that are grouped into clusters.
As the number of active containers being run increases, so does the management
overhead involved. Creating, maintaining, and deploying containers, implementing and
administrating cluster management software, monitoring your containers, and other
undifferentiated heavy lifting can detract from your organization's ability to focus on your
core business. This problem is solved with the release of Amazon ECS, a highly
scalable, high performance container management service that supports Docker
containers and allows you to easily run distributed applications on a managed cluster of
EC2 instances.
Amazon ECS provides three unique capabilities for users that want to run Docker
containers. First, there is no need to install, operate, and scale your own cluster
management infrastructure. With simple API calls, you can launch and stop containerenabled applications and query the complete state of your cluster. Second, Amazon
ECS is designed for use with other AWS services and includes access to many familiar
features like Elastic Load Balancing, EBS volumes, EC2 security groups, and IAM roles.
Third, Amazon ECS offers multiple ways to manage container scheduling, supporting a
wide variety of applications.
Page 7 of 23
Amazon Web Services – Docker on AWS: Running Containers in the Cloud
April 2015
Amazon ECS makes it easy to launch containers across multiple hosts, isolate
applications and users, and scale rapidly to meet changing demands of your applications
and users. Using Amazon ECS, you can now launch, manage, and orchestrate
thousands of Linux containers on a cluster of EC2 instances, without having to build your
own cluster management backend.
Clusters are made up of container instances, which are EC2 instances running the
Amazon ECS container agent that communicates instance and container state
information to the cluster manager and dockerd (the Docker daemon). Instances
running the Amazon ECS container agent automatically register with the default or
specified cluster. The Amazon ECS container agent is open source and freely available
(, and as such, can be built into any AMI
intended for use with Amazon ECS. When a cluster is created using the Amazon ECS
console, an Auto Scaling group is also associated with the cluster. This ensures that the
cluster grows in response to the needs of the container workload.
After instances have been deployed to a cluster, a task definition is used to define the
application or service to run by defining the containers and volumes that are deployed to
a host. Task definitions consist of one or more container definitions, which specify the
name and location of Docker images, how to allocate resources to each container, and
any links and other placement constraints. You can also specify any volume
requirements you might have. Running instances of a task definition are called tasks and
are the minimum unit of work in Amazon ECS.
You can run tasks by using a scheduler. Amazon ECS includes two built-in schedulers
or you can write your own or integrate with third-party schedulers.
Container-Enabled AMIs
AWS has developed a streamlined, purpose-built operating system for use with Amazon
ECS. The Amazon ECS-Optimized Amazon Linux AMI includes the Amazon ECS
container agent (running inside a Docker container), dockerd (the Docker daemon), and
removes a number of packages that are not required for running containers. Because
the Amazon ECS container agent has been released under the open source Apache 2.0
license anyone can build the agent into an AMI and deploy it to an ECS cluster. As the
ecosystem around Amazon ECS develops, we anticipate container-optimized partner
AMIs from RedHat, Ubuntu, and others. CoreOS has already announced support for
Amazon ECS.
Container Management
Amazon ECS provides cluster management capabilities to offset the administration
overhead of managing a large fleet of containers. For example, imagine running ten,
twenty, or hundreds of containers on a single host, and having ten, twenty, or hundreds
of hosts in a cluster. Docker deployments of this scale are not uncommon, and as
container technology continues to mature and becomes more widely implemented, scale
Page 8 of 23
Amazon Web Services – Docker on AWS: Running Containers in the Cloud
April 2015
and density of deployments will continue to increase.
Instead of requiring organizations to learn about a container management suite and
writing drivers or other components to interface with their applications,
Amazon ECS provides optimal control and visibility over your containers, your cluster,
and ultimately your applications with a simple, detailed API.
The following API operations are available for the core components of Amazon ECS
(clusters, container instances, task definitions, and tasks):
These actions eliminate the need to build and scale your own cluster and container
management solutions. You no longer have to worry about the availability of your master
node or your state management solution; call the relevant actions to carry out
management tasks that scale with the size of your cluster.
Page 9 of 23
Amazon Web Services – Docker on AWS: Running Containers in the Cloud
April 2015
Amazon ECS is a shared state, optimistic concurrency system and provides very flexible
scheduling capabilities. Amazon ECS schedulers leverage cluster state information
provided by the Amazon ECS API actions to make appropriate placement decision.
Currently, Amazon ECS provides two scheduler options. The RunTask action randomly
distributes tasks across your cluster and tries to minimize that possibility that a single
instance on your cluster will get a disproportionate number of tasks. The service
scheduler, CreateService, is ideally suited to long-running stateless services. The
service scheduler ensures that an appropriate number of tasks are constantly running
and reschedules tasks when a task fails (e.g., if the underlying container instance fails
for some reason). The service scheduler can also make sure that tasks are registered
against one or more load balancers.
Containers can be mapped to a load balancer using a service. Services provide Elastic
Load Balancing integration to distribute incoming traffic across containers and reduce
the overhead of tracking which instance hosts which container. Services optionally
provide fault-tolerance by checking the status of running tasks, restarting them if they
have failed, and re-associating them to their load balancer. Services also let you deploy
updates, such as changing the number of running tasks or the task definition version that
should be running.
In addition to providing a set of default schedulers, Amazon ECS allows for integration
with both custom schedulers and existing third party schedulers. As a proof of concept,
the ECSSchedulerDriver has been written to demonstrate integration of Amazon ECS
with a third-party scheduler; in this case, the Apache Mesos Framework. The Amazon
ECS API actions function as primitives for development of additional custom schedulers.
The List and Describe actions can obtain information about the state of a cluster, while
the StartTask action can place tasks on appropriate resources based on business and
application requirements.
Container Repositories
Amazon ECS is repository-agnostic, so customers can use a repository of their choice.
Some repositories, such as Docker Hub, also support integration with Github or
Bitbucket for automatic builds of containers when changes are pushed to your
codebase. Amazon ECS can integrate with private Docker repositories, including Docker
Hub Enterprise, running in AWS or an on-premises data center.
Logging and Monitoring
Amazon ECS supports monitoring of the underlying EC2 instances in a cluster using
Amazon CloudWatch.
Page 10 of 23
Amazon Web Services – Docker on AWS: Running Containers in the Cloud
April 2015
Amazon ECS allows you to store and share information using data volumes. Data
volumes are useful as a persistent data store that can be shared between multiple
containers on a host, as empty, non-persistent scratch space for containers, or as an
exported volume from one container to be mounted by other containers. ECS task
definitions allow you to reference the location of an appropriate location on the host
(either on instance storage or using EBS volumes). You can then reference the volume
from specific container definitions. You can also let Docker manage the volume and
share volumes between containers.
Volumes have an optional sourcePath parameter that references a directory on the
underlying host, and a containerPath parameter that references the mountpoint
inside the container. From within the container, writes (for example) toward the
containerPath value are persisted to the underlying volume defined in sourcePath.
If a sourcePath value is not provided, Docker treats the defined data volumes as
scratch space, and the data is not persisted past the life of the container. This is useful if
multiple containers in a task need to reference a common temporary cache for the
duration of their task.
Data volumes can also define the relationship of storage between two containers by
using the volumesFrom parameter. In this configuration, container “A” can have a data
volume defined on the underlying host, and other containers can reference the data
volume presented through container “A”. The volumesFrom parameter requires a
sourceContainer argument to specify which container's data volume should be
mounted. The mountpoint for the exported volume is defined by the container that is
exporting the shared volume.
Amazon ECS allows you to take advantage of native Docker features like port mapping
and container linking, while building on host-level Amazon EC2 networking features such
as security groups and IP addresses.
Amazon ECS describe* API actions facilitate service discovery so that applications in a
cluster can easily find one another as they are started, stopped, or moved.
Docker links are also supported by Amazon ECS, including support for injected
environment variables and /etc/hosts.conf entries to allow for simple discovery of other
linked containers. For more advanced users, it is possible to specify the exact security
group, network interface, and IP address resources that should be used by a container.
Page 11 of 23
Amazon Web Services – Docker on AWS: Running Containers in the Cloud
April 2015
Container Security
Running Docker containers within the execution environment of customer-controlled
EC2 instances builds on the familiar Amazon EC2 isolation frameworks such as AWS
Identity and Access Management (IAM), security groups, and Amazon Virtual Private
Cloud (Amazon VPC).
AWS customers maintain control over the Docker daemon itself, the operating system
and the underlying EC2 instance when using Amazon ECS. Customers also have
control via the AWS deployment & management service family as well as through native
configuration via the tooling of their choice. This operational model enables customers to
leverage the isolation capabilities of Docker containers and other software isolation
frameworks (such as iptables, SELinux, and AppArmor) in conjunction with the
underlying AWS security controls in order to meet their particular security, risk and
compliance requirements. For example, customers can provision EC2 instances with
different configurations in order to satisfy security, segregation or functional
Amazon ECS helps scale this approach through the concept of clusters, which act as a
placement boundary for the execution of any given task definition. This relationship
enables customers to provision different sets of security configurations and isolation
frameworks (such as assignment to different VPCs or EC2 instances) seamlessly and
assign them to different sets of task definitions, even as the number of running tasks and
container instances scales elastically.
Container Use Cases
There are many ways in which using containers on AWS can benefit your organization.
Batch Jobs
Containers can be used for Batch and ETL (Extract, Transform, Load) jobs by packaging
the job into a container and deploying the container into a cluster. It is possible to
improve the resource utilization of the cluster by running different versions of the same
job or multiple jobs on the cluster. By running job containers on an existing cluster, the
jobs start quickly; it is possible to grow capacity dynamically in response to demand.
Batch and ETL job capacity can be shared with other processes such as application
containers, taking advantage of fluctuations in demand on the existing cluster or using
Amazon EC2 Spot pricing to service the load.
Distributed Applications
Containers are an excellent building block for distributed applications and microservice
architecture. Distributed applications use design patterns of loose coupling, elasticity,
Page 12 of 23
Amazon Web Services – Docker on AWS: Running Containers in the Cloud
April 2015
message passing, and queuing to scale when required. The strengths of containers
(speed, density, consistency, and flexibility) help enable distributed application
deployments by allowing for quick provisioning and updates across a cluster of
heterogeneous servers. Microservice architecture goes one step further by breaking
each service down into its own encapsulated unit; containers are a perfect match for this
Batch processing is another area where containers make a lot of sense. It is quick and
easy to start up a large quantity of containers to process a batch of jobs and then stop
them after the processing is complete.
Amazon ECS provides an easy way to manage a large cluster of instances and
containers programmatically. With Amazon ECS, it is possible to launch thousands of
containers in seconds, allowing rapid scaling and iteration. The cluster can support a mix
of applications—for example, both short and long running tasks on the same EC2
instance. Amazon ECS is designed to enable provisioning of new containers into the
cluster in the most efficient way. If more control is required, this behavior can be
overridden by using the Amazon ECS API, which also allows for integration with other
tools—for example, connecting Amazon ECS to your current software delivery process.
Continuous Integration and Deployment
Containers are a keystone component of continuous integration (CI) and continuous
deployment (CD) workflows. Because containers can be built programmatically using
Dockerfiles, containers can be automatically rebuilt any time a new code revision is
Immutable deployments are natural with Docker. Each deployment is a new set of
containers, and it’s easy to roll back by deploying containers that reference previous
images. Automation simplifies frequent deployments. Amazon ECS provides API actions
that make deployments easy by providing the complete state of the cluster and the
ability to deploy containers using one of the built-in schedulers or a custom scheduler.
Amazon ECS can leverage existing CI features in tools like GitHub, Jenkins, and Docker
Hub while adding the ability to deploy images to your Amazon ECS clusters
automatically and continuously.
Docker Hub supports CI through automated builds. Automated builds leverage the
combination of a Dockerfile that specifies how the container should be built, and a
GitHub repository that specifies which code should be deployed in the container. Using a
GitHub service hook, Docker Hub can monitor the repository for new commits of code
and take a specified action—usually deploying the code into a freshly built container.
Amazon ECS includes the ability to create services and manage long-running
applications. A task definition defines the containers to run, while a service consists of a
Page 13 of 23
Amazon Web Services – Docker on AWS: Running Containers in the Cloud
April 2015
task definition, a count of how many tasks to run, and optionally the Elastic Load
Balancing load balancers associated with the running containers. Amazon ECS then
ensures that the correct number of tasks are running, auto-recovers stopped tasks, and
registers the EC2 instances with the load balancers. In a CI model, the container
definition references a specific image that is tagged by the build pipeline. Each task
definition is idempotent; when there is a new image, you can create a new task
definition. You can then deploy updates by updating the service’s task definition and
Amazon ECS stops tasks running the old task definition and start tasks with the new
task definition.
Because of how easy Amazon ECS makes deploying new containers, it is a perfect fit
for CI/CD tools that exist in the Docker ecosystem today.
Platform as a Service
Platform as a service (PaaS) is generally defined as a set of software, tools, and the
underlying infrastructure packaged as a complete unit. In this type of service model,
customers have a common base stack upon which they can deploy their application,
which may remove much of the overhead of purchasing, licensing, and deploying
individual components of the stack. Typical offerings in a PaaS stack might include
libraries, middleware, databases, and the underlying OS, which are provided by the
PaaS organization; the consumer usually performs resource configuration.
For PaaS providers, there may be significant challenges in quickly deploying large
numbers of identical resources in a modular fashion, isolating end users from each
other, and integrating new iterations of platform offerings into production. Because
containers inherently facilitate isolation of resources and users, allow for extremely
efficient utilization of resources, and make creating and deploying template resources a
simple process, containers are a logical choice for a PaaS offering.
Each individual product offered by the PaaS provider could be built into its own
container, and then deployed on demand with very little boot time in response to a
customer need. Amazon ECS can take much of the undifferentiated heavy lifting off the
shoulders of the PaaS provider by handling the scheduling and cluster management.
Because Amazon ECS leverages the same physical datacenters that power Amazon
EC2, the PaaS provider only has to worry about deploying the products that define their
platform offerings, while leaving the infrastructure and container management to AWS.
Architectural Considerations
When a task is built, it is important to keep in mind that all containers defined in the task
are placed onto a single instance in the cluster. Because of this, it’s logical that a task
represents a single contiguous application that has multiple tiers requiring inter-container
communication. A LAMP stack that runs a particular application should have all
Page 14 of 23
Amazon Web Services – Docker on AWS: Running Containers in the Cloud
April 2015
components—an Apache container, a PHP container, and a MYSQL container—defined
in the same task. Tasks give users the ability to allocate resources to containers, so
containers can be evaluated on resource requirements and collocated to use available
compute and memory resources fully. CPU resources are allocated in units of CPU
cores, where 1024 units represent a single core. Memory can be allocated on a per
megabyte basis. This granularity allows balancing of host resources and container
Amazon ECS provides three API actions for placing containers onto hosts: StartTask,
RunTask, and CreateService. The RunTask action uses Amazon ECS scheduler logic to
place a task on an open host. If more control is needed, use the StartTask action, as this
allows a specific cluster instance to be passed as a value in the API call.
The CreateService action allows for the creation of a Service object, which is the
combination of a TaskDefinition object and an existing Elastic Load Balancing load
balancer. By creating a Service object, multiple instances of application stacks can be
deployed and associated with load balancers. A container's host port mapping can be
mapped directly to an Elastic Load Balancing instance listener port. Multiple containers
can be mapped to the same load balancer via different listeners. For example, to expose
both an Apache container listening on host port 80 and a Tomcat container listening on
host port 8080 through the same load balancer, two listeners need to be created and
mapped to the container's host ports. There is a fixed relationship between container
ports and listeners.
Because containers cannot use links to communicate across host (instance) boundaries,
there can be challenges with advertising internal container state, such as current IP
address and application status, to other containers running on separate hosts within the
cluster. In a dynamic container cluster, containers may frequently change state or obtain
a new IP address and port mappings as a result of termination and re-instantiation. The
concept of “service discovery” exists with the intention of addressing these challenges.
At a high-level, service discovery patterns generally involve containers updating a
common key-value store with information about their state (e.g., available or running),
the port they are currently listening to, and their current IP address. The Amazon ECS
describe API actions can serve as primitives for service discovery functionality. For
example, the describe-service action returns the load balancer used for a service, which
can provide a simple service discovery method.
This walkthrough helps you create your first cluster, register a task definition, and deploy
a container. The following steps guide you through launching a simple NGINX webserver
inside an ECS container using an ECS task and a community dockerfile.
Page 15 of 23
Amazon Web Services – Docker on AWS: Running Containers in the Cloud
April 2015
NOTE: For all preliminary setup requirements, such as registering an AWS account,
configuring IAM, setting up a VPC, and installing the CLI, see the Setting Up topic in the
Amazon EC2 Container Service Developer Guide.
For more information about any of the specific commands, see the Amazon EC2
Container Service API Reference.
1) Create a new cluster:
Every account is automatically given a default cluster (named "default”). However, you
can create your own with the create-cluster command:
$ aws ecs create-cluster --cluster-name Walkthrough
"cluster": {
"clusterName": "Walkthrough",
"status": "ACTIVE",
Note: By default, each AWS account is limited to two clusters.
2) Add instances:
Next, launch instances into your cluster. If you simply launch a container-optimized AMI
and do nothing else, the instance automatically registers to the default cluster.
If you would like to control which cluster the instances register to, you need to input
UserData to populate the cluster name into the /etc/ecs/ecs.config file. For more
information, see the Running Commands on Your Linux Instance at Launch topic in the
Amazon EC2 User Guide for Linux Instances.
Paste the following into the userdata field when you launch your instances:
echo ECS_CLUSTER=your_cluster_name >> /etc/ecs/ecs.config
For this walkthrough, you will launch a webserver so it is important to configure the
correct security group permissions and allow inbound access from anywhere on port 80.
3) Run a quick check:
Page 16 of 23
Amazon Web Services – Docker on AWS: Running Containers in the Cloud
April 2015
After a few moments, your instances should have registered to the cluster. To check,
use the list-container-instances command:
$ aws ecs list-container-instances --cluster Walkthrough
"containerInstanceArns": [
To dig into the instances more, you can also describe the instances using the
describe-container-instances command:
$ aws ecs describe-container-instances --cluster default -container-instances container_instance_UUID
$ ecs describe-container-instances --cluster Walkthrough -container-instances 9235f233-d23e-4c36-9130-81d072b2d630
"failures": [],
"containerInstances": [
"status": "ACTIVE",
"registeredResources": [
"integerValue": 1024,
"longValue": 0,
"type": "INTEGER",
"name": "CPU",
"doubleValue": 0.0
"integerValue": 996,
"longValue": 0,
"type": "INTEGER",
"name": "MEMORY",
"doubleValue": 0.0
Page 17 of 23
Amazon Web Services – Docker on AWS: Running Containers in the Cloud
April 2015
"name": "PORTS",
"longValue": 0,
"doubleValue": 0.0,
"stringSetValue": [
"type": "STRINGSET",
"integerValue": 0
"ec2InstanceId": "i-c5690f29",
"agentConnected": true,
"containerInstanceArn": "arn:aws:ecs:us-east1:accountid:container-instance/9235f233-d23e-4c36-913081d072b2d630",
"remainingResources": [
"integerValue": 1024,
"longValue": 0,
"type": "INTEGER",
"name": "CPU",
"doubleValue": 0.0
"integerValue": 996,
"longValue": 0,
"type": "INTEGER",
"name": "MEMORY",
"doubleValue": 0.0
"name": "PORTS",
"longValue": 0,
"doubleValue": 0.0,
"stringSetValue": [
"type": "STRINGSET",
"integerValue": 0
Page 18 of 23
Amazon Web Services – Docker on AWS: Running Containers in the Cloud
April 2015
4) Register the task:
Before you can run a task on your ECS cluster, you must register a task definition. Task
definitions are lists of containers grouped together. For more information about the
available task definition parameters, see Amazon ECS Task Definitions in the Amazon
EC2 Container Service Developer Guide.
The following example creates a JSON file with the task definition and names it
nginx_task.json. This specific task launches a pre-configured NGINX container from the
Docker Hub repository:
"image": "dockerfile/nginx",
"name": "nginx",
"cpu": 1000,
"memory": 500,
"essential": true,
"links": [
"entryPoint": [
"environment": [
"portMappings": [
"containerPort": 80,
"hostPort": 80
“family”: “whitepaper”
Page 19 of 23
Amazon Web Services – Docker on AWS: Running Containers in the Cloud
April 2015
After creating the task definition file, register it with Amazon ECS:
$ aws ecs register-task-definition --family Walkthrough -container-definitions file://$HOME/nginx_task.json
"taskDefinition": {
"taskDefinitionArn": "arn:aws:ecs:us-east1:accountid:task-definition/Walkthrough:1",
"containerDefinitions": [
"environment": [],
"name": "nginx",
"links": [],
"image": "dockerfile/nginx",
"essential": true,
"portMappings": [
"containerPort": 80,
"hostPort": 80
"entryPoint": [
"memory": 500,
"cpu": 1000
"family": "Walkthrough",
"revision": 1
5) Run the task:
To run your task on your cluster, take note of the taskDefinition instance value
(Walkthrough:1) returned after task registration in the previous step. The instance of the
task definition is called by run-task. You can also obtain the ARN by using the aws
ecs list-task-definitions command.
$ aws ecs run-task --cluster Walkthrough --task
Walkthrough:1 --count 1
Page 20 of 23
Amazon Web Services – Docker on AWS: Running Containers in the Cloud
April 2015
"tasks": [
"taskArn": "arn:aws:ecs:us-east1:accountid:task/c79b8d02-1969-40c0-be7b-a9c0261104d2",
"overrides": {
"containerOverrides": [
"name": "nginx"
"lastStatus": "PENDING",
"containerInstanceArn": "arn:aws:ecs:us-east1:accountid:container-instance/54353373-c32c-463e-96d90ac7db5f6a9d",
"desiredStatus": "RUNNING",
"taskDefinitionArn": "arn:aws:ecs:us-east1:accountid:task-definition/Walkthrough:1",
"containers": [
"containerArn": "arn:aws:ecs:us-east1:accountid:container/d31b241d-8166-4f12-b900c1187bed3c1e",
"taskArn": "arn:aws:ecs:us-east1:accountid:task/c79b8d02-1969-40c0-be7b-a9c0261104d2",
"lastStatus": "PENDING",
"name": "nginx"
6) Test your container:
Lastly, test that your container is responding. Because you mapped the container port 80
to the instance port 80, you can use the curl utility to test the public IP address of the
instance and get a response.
$ curl -v >> /dev/null
* Rebuilt URL to:
* Hostname was NOT found in DNS cache
Page 21 of 23
Amazon Web Services – Docker on AWS: Running Containers in the Cloud
% Total
% Received % Xferd
Time Current
April 2015
Average Speed
Dload Upload
Left Speed
0 --:--:-- --:-:-- --:--:-0* Connected to
( port 80 (#0)
> GET / HTTP/1.1
> User-Agent: curl/7.38.0
> Host:
> Accept: */*
< HTTP/1.1 200 OK
* Server nginx/1.6.2 is not blacklisted
< Server: nginx/1.6.2
< Date: Wed, 17 Dec 2014 19:33:33 GMT
< Content-Type: text/html
< Content-Length: 867
< Last-Modified: Wed, 17 Dec 2014 17:06:20 GMT
< Connection: keep-alive
< ETag: "5491b80c-363"
< Accept-Ranges: bytes
{ [data not shown]
867 100
0 --:--:-- --:-:-- --:--:-- 6568
* Connection #0 to host left intact
Using containers in conjunction with AWS can accelerate your software development by
creating synergy between your development and operations teams. The efficient and
rapid provisioning, the promise of “build once, run anywhere”, the separation of duties
via a common standard, and the flexibility of portability that containers provide offer
advantages to organizations of all sizes. By providing a range of services that support
containers along with an ecosystem of complimentary services, AWS makes it easy to
get started with containers while providing the necessary tools to run containers at scale.
Page 22 of 23
Amazon Web Services – Docker on AWS: Running Containers in the Cloud
April 2015
Further Reading
For more information, see the following resources:
Amazon EC2 Container Service documentation
Docker (
Deploying AWS Elastic Beanstalk Applications from Docker Containers (AWS Elastic
Beanstalk Developer Guide)
CoreOS Linux AMI (AWS Marketplace)
ECS Mesos Scheduler Driver
© 2015, Amazon Web Services, Inc. or its affiliates. All rights reserved. This document is
provided for informational purposes only. It represents AWS’s current product offerings
and practices as of the date of issue of this document, which are subject to change
without notice. Customers are responsible for making their own independent
assessment of the information in this document and any use of AWS’s products or
services, each of which is provided “as is” without warranty of any kind, whether express
or implied. This document does not create any warranties, representations, contractual
commitments, conditions or assurances from AWS, its affiliates, suppliers or licensors.
The responsibilities and liabilities of AWS to its customers are controlled by AWS
agreements, and this document is not part of, nor does it modify, any agreement
between AWS and its customers.
Document Revisions
Version 1.0 – initial release
Appendix A: Supported Linux Versions
Amazon is a contributor to the official Docker Repos Program. At the time of publication,
Amazon Linux 2014.03 includes the docker-0.9.0 package, and the most up-to-date
versions are available in our repositories. Other Linux distributions in AWS support
Docker as well; Docker is included with RHEL 7 and is available for RHEL 6 in the Extra
Packages for Enterprise Linux (EPEL) repositories. CoreOS, a minimal Linux OS
designed specifically for Docker, is available in the AWS Marketplace.
Page 23 of 23
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF