homepage-docker-logo

A Practical Introduction to Docker Containers

Background

Why

Docker has quite an amount of buzz around it today because it makes so many things easy that were difficult with virtual machines.

Docker containers makes it easy for Developers, Systems Administrators, Architects, Consultants and others to quickly test a piece of software in a container; much quicker than a virtual machine, and using less resources. The average command in Docker takes under a second to complete.

[root@keith]# time docker run fedora cat /etc/redhat-release
Fedora release 20 (Heisenbug)
real 0m0.715s
user 0m0.004s
sys 0m0.004s

 

What

Simple Use Cases

  • I need to see the man page from a specific version of RHEL, CentOS or Fedora
  • I need to quickly verify the command line options of a program
  • I need to test the functionality of a specific version of software
  • I need a scratch pad that is NOT my system
  • I need a single daemon running, and I don’t care what distribution of Linux it runs on (see registry below)

 

Complex Use Cases

Docker containers run a single process when started, but complex installations of software which require multiple daemons running simultaneously (RHEV-M, Satellite, etc) can be done. However, they require more engineering work using either Bash scripting or running SystemD in the container.

Production vs. Development

The Docker tooling reached 1.0 in June, 2014 and with some precautions, can be used in production. There is a lot to think about with a production deployment, and I suggest readying the Architecting Containers series if you are seriously considering a production deployment.

CentOS and Red Hat Enterprise Linux

This tutorial will focus on integration with Red Hat technologies including CentOS and Red Hat Enterprise Linux. The Docker tooling (daemon and client) is available in Red Hat Enterprise Linux, CentOS, and Fedora. Also, base Docker images are available for Red Hat Enterprise Linux (6 and 7), CentOS and Fedora.

Architecture

One of the key advantages of using Docker is it’s centralized image management server, called a Registry Server. The Docker project maintains a public registry server which hosts images they maintain, as well as images created by the community. This registry service is free to use, as long as the images are public.

As one builds images, they are made up of layers. These layers are shared together in, what is called a repository. Users on the registry can share multiple repositories.

Docker has official CentOS and Fedora repositories which they support:

 

This tutorial will use public CentOS Fedora repositories, but you may also run through this tutorial using Red Hat Enterprise Linux:

 

OS Virtualization vs. Application Virtualization

Why separate RPMs for each major version of Red Hat Enterprise Linux or CentOS into different YUM repositories? This was a conscious decision because we are working with a full enterprise Linux distribution versus a single application. Historically, it has been best practice to install a fresh copy when upgrading major versions of Red Hat Enterprise Linux or CentOS. While it is possible to upgrade in place and share multiple versions in a single repository, this is not the preferred method with an enterprise operating system.

However, when virtualizing individual applications, it may be very appropriate to upgrade in place and share all versions within a single repository (See section: Set Up a Registry Server). Each version will be tagged in place and it’s easy to upgrade or downgrade atomically.

Registry Servers

After reading this tutorial, you may be excited to set up your own registry server. This is easy to do, but has some caveats, so think through whether it is worth just using a hosted registry server. If you want to deploy an onsite registry server, there are several options:

 

Basic Operations

Now we will run through some basic operations to get you up and running.

Install Docker

In CentOS 7, this requires the user space tools and daemon to be installed from an RPM. On Red Hat Enterprise Linux, this RPM is contained in the rhel-7-server-extras-rpms/x86_64 channel.

yum install docker
systemctl enable docker.service
systemctl start docker.service

 

Test Docker

This will automatically pull the latest CentOS 4 image from the remote repository and cache it locally.

docker run -it centos cat /etc/redhat-release

 

Pull an Image

This will pull the latest CentOS 5 image from the remote repository and cache it in the local index. If the image you are pulling is made up of layers, all of the layers will be pulled.

docker pull centos

 

List Images

This will list all of the images in the local index and display how they are linked to each other. Every time a new container is spawned from an image, it will create another copy on write image to save it’s changes too. The tree structure will help make things clear.

Since, Docker 1.7, there is no native tooling to inspect image layers in a local cache, but with the help of a tool called dockviz, you can quickly inspect all of the layers in a local repository. The following command will returned shortened versions of the UUID that are typically unique enough to work with on a single machine. If you need to the full UUID, use the –no-trunc option.

docker run --rm --privileged -v /var/run/docker.sock:/var/run/docker.sock nate/dockviz images -t

 

Tag an Image

It makes it easier to deal with images if they are tagged with simple names.

docker tag fatherlinux/centos5-base centos5-base

 

Run a Container

Notice how easy it is to test a command in CentOS 4

docker run -i -t -rm centos man rsync

 

Log in to the Hosted Docker Registry

To create repositories on the public Docker Registry, it is necessary to sign up at:

https://www.docker.io/account/signup/

 

Once you have created an account, you will need to login from the command line

docker login index.docker.io

 

If login is successful.

Username: fatherlinux
Password: 
Email: smccarty@redhat.com
Login Succeeded

 

Dockerfile: Commit an Image

Once logged in to the public Docker Registry, new images can be built and committed with code using a Dockerfile. This allows an administrator to automatically rebuild a known good starting point quickly and easily. It is recommended to always start with an image defined by a Dockerfile.

There are a couple of important things to notice with this Dockerfile. First, the FROM directive specifies centos as the base image. This will pull the latest image from the centos repostiory on DockerHub. In this example, I have provided the source repository for you. Second, the only change we have specified in the Dockerfile, is to update CentOS to latest available packages. Finally, notice we have a commented out ENTRYPOINT at the end of the file. Play with this option to get a feel for it while building from this Dockerfile.

vi Dockerfile

 

#
# Version 1

# Pull from CentOS RPM Build Image
FROM centos

MAINTAINER Scott McCarty smccarty@redhat.com

# Update the image
RUN yum update -y

# Output
# ENTRYPOINT tail /var/log/yum.log

 

Build and tag the image

Execute the following command in same directory as the Dockerfile.

docker build -t centos-updated .

 

Inspect the new image

This will list all of the layers in an image

docker run --rm --privileged -v /var/run/docker.sock:/var/run/docker.sock nate/dockviz images -t
└─60e65a8e4030 Virtual Size: 196.6 MB Tags: docker.io/centos:latest
  └─6d8919b62698 Virtual Size: 196.6 MB
    └─05192ebc2b2d Virtual Size: 264.0 MB Tags: centos-updated:latest

 

Notice that the new image is now available for deployment.

Test the new image

docker run -it --rm centos-updated tail /var/log/yum.log
Jan 13 09:57:25 Updated: nss-sysinit-3.19.1-19.el7_2.x86_64
Jan 13 09:57:26 Updated: nss-3.19.1-19.el7_2.x86_64
Jan 13 09:57:27 Updated: nss-tools-3.19.1-19.el7_2.x86_64
Jan 13 09:57:28 Updated: 1:openssl-libs-1.0.1e-51.el7_2.2.x86_64

 

Manually: Commit an Image

Once a container has changes made locally, they can be committed to the local index. This allows you to check point and continue. It also allows you to create new images based off of this modified container.

docker run -it centos bash

 

Modify the image

Make some changes inside the container. In this example, create a test file and exit

echo test file > /etc/test.cfg
exit

 

Commit the container

First, get a list of containers. Notice that every container has a CONTAINTER ID and a STATUS. A status of Up means the container is currently running, while a status of Exit indicates that the container has been stopped. Think of the CONTAINER ID as a branch from the base image that contains all of the changes that were made to the container while it was running. By default this data is saved even after the container is shut down.

docker ps -a
CONTAINER ID        IMAGE               COMMAND                  CREATED              STATUS                     PORTS               NAMES
c65fe9f4194b        centos              "bash"                   About a minute ago   Exited (0) 3 seconds ago                       condescending_galileo
620e24fc14c2        centos-updated      "tail /var/log/yum.lo"   6 minutes ago        Exited (0) 6 minutes ago                       determined_perlman
...

 

Now, commit the container back as a branch of it’s base image

docker commit c65fe9f4194b

 

Notice that the image is now available in the tree output. Also, notice that the newly created image layer (6d8919b62698) is a branch of the root centos:latest image, not the centos-updated:latest image, which we previously built with a Dockerfile.

docker run --rm --privileged -v /var/run/docker.sock:/var/run/docker.sock nate/dockviz images -t
└─60e65a8e4030 Virtual Size: 196.6 MB Tags: docker.io/centos:latest
  ├─5b6001fad9a6 Virtual Size: 196.6 MB
  └─6d8919b62698 Virtual Size: 196.6 MB
    └─05192ebc2b2d Virtual Size: 264.0 MB Tags: centos-updated:latest

 

Tag the new image with something meaningful

docker tag 5b6001fad9a6 centos-test

 

Push a Container

Once a container is committed locally, it can be pushed back to the registry server to be shared. The changes will be pushed as a layered image. Notice how quickly it is able to push only the differences between your modified image and the base image. This is a big part of the value.

docker tag centos-test fatherlinux/centos-test
docker push fatherlinux/centos-test
The push refers to a repository [fatherlinux/centos-test] (len: 1)
Sending image list
Pushing repository fatherlinux/centos-test (1 tags)
e39724bc32b2: Image already pushed, skipping 
c9bfb69481a8: Image successfully pushed 
Pushing tag for rev [c9bfb69481a8] on {https://registry-1.docker.io/v1/repositories/fatherlinux/centos6-base-hostname/tags/latest}

 

Advanced Operations

Pull All Standard Images

These repositories (images) are publicly available from DockerHub and the Red Hat Registry.

docker pull centos
docker pull fedora
docker pull registry.access.redhat.com/rhel

 

Create Base Image

This method was developed with guidance from this script. This example is based on CentOS 6.

Create a tar file of the system

tar --numeric-owner --exclude=/proc --exclude=/sys -cvf centos6-base.tar /

 

Copy the tar file to where the consuming system and Import the image

cat centos6-base.tar | docker import - centos6-base

 

Test

docker run -i -t centos6-base cat /etc/redhat-release

 

Set Up a Registry Server

Notice that the entire application is packaged up and ran from inside of a docker container. This has the interesting consequence that we are not even concerned with what operating system is hosting this registry application. Also, notice that port 5000 in the docker container is mapped to port 5000 on the hosting virtual machine, which makes the application running in the container transparently appear to be running on the virtual machine.

docker run -p 5000:5000 registry

 

Search Private Registry

List all images in the repository

http://registry.example.com:5000/v1/search?

 

Search for all repositories with the word “rhel” in their name

http://registry.example.com:5000/v1/search?q=rhel

 

Remove Old Docker Containers

By default, Docker keeps changes for every container which is instantiated. When testing, this can be undesirable. Be careful because this will remove all branches/data. Any containers which have not been committed will have all data deleted:

docker rm `docker ps --no-trunc -a -q`

 

Links


Join Red Hat Developers, a developer program for you to learn, share, and code faster – and get access to Red Hat software for your development.  The developer program and software are both free!

Take advantage of your Red Hat Developers membership and download RHEL today at no cost.

    1. So, Docker does not handle updates to an application/container. The original idea is that the entire environment would be burned down and recreated. Very similar to the cloud model.

      That said, there are a ton of projects working on solving this. One of them is the Fedora/RHEL Atomic project.

      Like

  1. Hi, i deployed two docker containers in two different virtual machines, virtual machines having same network in two docker containers i am running redis when i try to run second docker container redis slaveof mode it failed to identify first container could you please help me….

    Like

  2. Hello all , I have ubuntu machine and I have deployed some containers , when I run docker ps command it gives me names and other details , and by ‘docker ps -a’ command gives me image size with other details. I want full size of a container. with volumes and image. and I also I want a to make a application for do backup of that all containers. will u please someone help me !

    Like

    1. So, Bharat Karale, getting the size of a Docker image and it’s associated -v bind mounts is not trivial. Also, with immutable infrastructure I believe this is the wrong path. Essentially, there should always be good separation of code, configuration and data in most (but not all) applications you containerize.

      For example, with a web application composed of Varnish, Memcached, MySQL, and Apache, you should always make sure that/etc/my.cnf, /var/lib/mysql, /var/www/html, and /etc/httpd are all bind mounted into the container. If you do this here is the breakdown of your code, config and data:

      Code: Docker images/repositories containing Varnish, Memcachd, MySQL, and Apache.
      Config: /etc/my.cnf and /etc/httpd
      Data: /var/lib/mysql and /var/www/html (could also be considered application and built in a layered image, but I would confuse here)

      If this is your architecture, to back up each of these you would:

      Code: Backup your registry server and/or the volumes it uses to store images/repositories
      Config: Back up NFS, Git, or iSCSI (many others) volumes that you use to mount in configuration files (done any number of ways, but easiest in a Kubernetes environment)
      Data: Back up NFS, iSCSI, or EBS volumes that you use to mount in data

      Some of these questions get easier with something like Kubernetes/Atomic Enterprise Platform/OpenShift. But, with raw Docker, you have to build out some of the solution yourself. This really requires a book to describe, but hopefully this gets you pointed in the right direction.

      [1]: Paths from the RHEL/CentOS/Fedora world because it’s what I know off of the top of my head

      Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s