Docker and Fedora
For the last couple of days I have been playing with Docker. Docker is a project that helps you create images and run containers. This sounds like virtualization, but it isn't. It uses Linux Containers (LXC) under the hood to do all the magic. What kind of magic? Read on.
The first question you might ask is: how is Docker/LXC different from virtualization?
LXC is something between chroot and full virtualization. You don't run your applications in a virtual machine (inside a process controlled by the virtualization engine). Instead your applications are run in isolated (by the kernel itself) environments. Processes that run in one container do not have any access to processes in other containers. From a cointainer POV it looks like virtualization, but when you look from the host side - you can see all of the processes (applications) running on the host, directly. I'm pretty new to the LXC technology, but it's very promising, especially with regards to speed.
The most visible difference between virtualization and LXC for the end user is boot time. LXC overhead is literally zero compared to the few seconds up to minutes to boot your operating system in a virtualized environment. When you run a container, it's ready to do the stuff immediately - you don't need to wait at all.
This single feature is a great reason to take a look at LXC. But there is also Docker, which is a great extension of what we already have in LXC.
Docker builds upon LXC. It uses it to create images and later run and manage
them. What makes Docker especially nice is its lightweight and easy to use. The
good folks at Docker made Ubuntu their distribution of choice, but Fedora isn't
sleeping. Just recently Lokesh Mandvekar created a
docker-io package and it
was reviewed and
accepted for Fedora.
It'll take a week or two for it become available in the Fedora 19+ repos. If
you're eager (and brave) - I prepared my own Docker
repository with RPMs for
Fedora 19, 20 and Rawhide. This repo will become unavailable after the official
Docker RPMs hit Fedora repos.
Please note that I used a Fedora 20 host, but it should work on Fedora 19 too.
All commands below should be executed with root privileges.
curl http://goldmann.fedorapeople.org/repos/docker.repo > /etc/yum.repos.d/docker-goldmann.repo yum install docker-io
When the install finishes - start the
docker systemd service:
systemctl start docker.service
And if you want to enable it on boot:
systemctl enable docker.service
Docker should be running now.
Run your first container
Docker offers a central repository with images. This makes it easy to download (and publish) images. Matthew Miller (Fedora Cloud Architect) prepared a Fedora 19 image. This image will be updated to Fedora 20 once it is released.
Let's grab the
$ docker pull mattdm/fedora Pulling repository mattdm/fedora 22a514a5aa4c: Download complete 50f374c05c2c: Download complete 97fc5bf7f8d4: Download complete
Done! Now you have the image locally (in
/var/lib/docker) and you can immediately start a container based on it:
docker run -i -t mattdm/fedora /bin/bash
Let's look at the parameters:
run- runs a container,
-i- keeps the stdin open, even if there is nothing attached,
-t- allocates a pseudo terminal, so we can interact with the container directly,
mattdm/fedora- ID of the image, it can be a tag or a hash (
22a514a5aa4cin this case),
/bin/bash- the command to run after the container boots.
After you run the command, you'll be greeted by the bash prompt from inside the container, where you can do whatever you want. There are different types of images, some of them have an entry point, some not. I hope to discuss this further in a different blog post.
$ docker run -i -t mattdm/fedora /bin/bash bash-4.2#
If you see an error similar to this, try again.
2013/09/25 13:22:02 Error: Error starting container 4b9cdcc43f43: fork/exec /usr/bin/unshare: operation not permitted
This is a known bug and I hope it will be fixed soon.
Basic container management
To stop the container, just press
CTRL+D. Please note that the container is
now stopped, but that does not mean that it no longer exists. Stopped container
can be started or removed.
To remove a stopped container you need to know the container ID. You can see it
by using the
docker ps command. By default the
docker ps command will show
only running containers. To see all containers (including stopped) run
$ docker ps -a ID IMAGE COMMAND CREATED STATUS PORTS 15bd697c7174 mattdm/fedora:latest /bin/bash 22 minutes ago Exit 0 5ab7c7a95885 mattdm/fedora:latest /bin/bash 23 minutes ago Exit 0 4b9cdcc43f43 mattdm/fedora:latest /bin/bash 23 minutes ago Exit 0 0fdab01e4eaa mattdm/fedora:latest /bin/bash 24 minutes ago Exit 0
Now you can remove the container by executing the
docker rm command and
specifying the ID, for example
docker rm 15bd697c7174. The
container is now gone.
By default Fedora disables IP forwarding which will prevent you from accessing the Internet from inside of the container. In most (all?) cases this is not what you want. To enable IP forwarding you can run this command:
sysctl -w net.ipv4.ip_forward=1
After restarting, forwarding will be disabled again. To make it persistent,
/etc/sysctl.d/80-docker.conf file and put the following line in it:
net.ipv4.ip_forward = 1
There is an open bug to fix this in Fedora.
Build your first image
What we have done so far is run an image made by someone else. Let's create our own image now.
Docker uses plain text files to describe the image which can contain various
commands. To build an image,
let's create an empty directory and place a file in it called
# Base on the Fedora image created by Matthew FROM mattdm/fedora # Install the JBoss Application Server 7 RUN yum install -y jboss-as # Run the JBoss AS after the container boots ENTRYPOINT /usr/share/jboss-as/bin/launch.sh standalone standalone.xml 0.0.0.0
FROM command is
required and tells Docker which image should be used as a base for our new
RUN command is used
to modify the image by running a command inside the container at the time of
command specifies which command should be executed after the container fully
The next step is to build the image itself. In the directory execute the
docker build . command:
$ docker build . Uploading context 10240 bytes Step 1 : FROM mattdm/fedora ---> 22a514a5aa4c Step 2 : RUN yum install -y jboss-as ---> Running in 4e4d90823207 Resolving Dependencies --> Running transaction check ---> Package jboss-as.noarch 0:7.1.1-21.fc19 will be installed --> Processing Dependency: wss4j >= 1.6.7 for package: jboss-as-7.1.1-21.fc19.noarch --> Processing Dependency: wsdl4j >= 1.6.2-5 for package: jboss-as-7.1.1-21.fc19.noarch --> Processing Dependency: resteasy >= 2.3.2-7 for package: jboss-as-7.1.1-21.fc19.noarch --> Processing Dependency: mod_cluster-java >= 1.2.1-2 for package: jboss-as-7.1.1-21.fc19.noarch --> Processing Dependency: jython >= 2.2.1-9 for package: jboss-as-7.1.1-21.fc19.noarch --> Processing Dependency: jbossws-spi >= 2.1.0 for package: jboss-as-7.1.1-21.fc19.noarch --> Processing Dependency: jbossws-native >= 4.1.0 for package: jboss-as-7.1.1-21.fc19.noarch --> Processing Dependency: jbossws-cxf >= 4.1.0 for package: jboss-as-7.1.1-21.fc19.noarch --> Processing Dependency: jbossws-common >= 2.0.4-3 for package: jboss-as-7.1.1-21.fc19.noarch [....SNIP...] xpp3.noarch 0:220.127.116.11-8.fc19 xpp3-minimal.noarch 0:18.104.22.168-8.fc19 xsom.noarch 0:0-9.20110809svn.fc19 xstream.noarch 0:1.3.1-5.fc19 zip.x86_64 0:3.0-7.fc19 Complete! ---> fafccbe2bffc Step 3 : ENTRYPOINT /usr/share/jboss-as/bin/launch.sh standalone standalone.xml 0.0.0.0 ---> Running in 055d264ab953 ---> 366ff524eea0 Successfully built 366ff524eea0
Please note that after every command Docker commits the changes (in a manner similar to Git). Future executions of the same command will use the cached result.
Now if we run
docker run -i -t 366ff524eea0 (please note that we don't
/bin/bash command, since our image has an entry point and it will
be executed for us) we'll see JBoss AS booting:
$ docker run -i -t 366ff524eea0 ========================================================================= JBoss Bootstrap Environment JBOSS_HOME: /usr/share/jboss-as JAVA: java [...SNIP...] 13:28:15,433 WARN [org.jboss.as.domain.http.api] (MSC service thread 1-4) JBAS015102: Unable to load console module for slot main, disabling console 13:28:15,442 INFO [org.jboss.as.server.deployment.scanner] (MSC service thread 1-1) JBAS015012: Started FileSystemDeploymentService for directory /usr/share/jboss-as/standalone/deployments 13:28:15,520 INFO [org.jboss.as.connector.subsystems.datasources] (MSC service thread 1-2) JBAS010400: Bound data source [java:jboss/datasources/ExampleDS] 13:28:15,552 INFO [org.jboss.as] (Controller Boot Thread) JBAS015951: Admin console listening on http://127.0.0.1:9990 13:28:15,553 INFO [org.jboss.as] (Controller Boot Thread) JBAS015874: JBoss AS 7.1.1.Final "Brontes" started in 2328ms - Started 133 of 208 services (74 services are passive or on-demand)
That's it, JBoss AS is running.
I highly recommend the try-and-fail method. Read the Docker docs, try to build your own images. In future blog posts I'll get a bit more into Docker details (and we'll build a cluster).
Hope you enjoyed this quick ride with Docker and Fedora!