Samouczek dotyczący platformy Docker: wprowadzenie do platformy Docker

Kontenery zapewniają lekki sposób przenoszenia obciążeń aplikacji, takich jak maszyna wirtualna, ale bez kosztów ogólnych i zbiorczych zwykle związanych z maszynami wirtualnymi. Dzięki kontenerom aplikacje i usługi można pakować i swobodnie przenosić między środowiskami fizycznymi, wirtualnymi lub chmurowymi.

Docker, system do tworzenia i zarządzania kontenerami stworzony przez firmę Docker Inc., wykorzystuje natywną funkcjonalność kontenera znajdującą się w systemie Linux i udostępnia ją użytkownikom końcowym za pośrednictwem interfejsu wiersza polecenia i zestawu interfejsów API.

Wiele typowych składników aplikacji jest teraz dostępnych jako wstępnie spakowane kontenery platformy Docker, co ułatwia wdrażanie stosów oprogramowania jako komponentów odsprzężonych (model mikrousług). To powiedziawszy, pomaga wiedzieć, jak elementy pasują do siebie od wewnątrz.

Dlatego w tym przewodniku instaluję serwer WWW Apache w kontenerze Docker i badam, jak działa Docker po drodze.

Zainstaluj Docker

Używam Ubuntu jako podstawy do kompilacji Dockera. Ubuntu to nie tylko popularna i szeroko stosowana dystrybucja, ale sam zespół Dockera używa Ubuntu do programowania, a Docker jest obsługiwany na Ubuntu Server od wersji 12.04 i nowszych. Dla uproszczenia zacznę od instrukcji dotyczących nowej instalacji Ubuntu 16.04.

Przygotuj Ubuntu Linux dla Dockera

Pierwszą rzeczą do zrobienia jest uzyskanie odpowiedniej wersji jądra i jego nagłówków:

$ sudo apt-get install --install-recommends linux-generic-hwe-16.04

Ten proces może zająć trochę czasu i będzie wymagał ponownego uruchomienia po zakończeniu:

$ sudo reboot

Może być konieczne późniejsze zaktualizowanie innych pakietów w systemie:

$ sudo apt-get update

$ sudo apt-get upgrade

Zainstaluj Docker na Ubuntu

Instalacja Dockera w dystrybucjach CentOS, Fedora, Debian, Ubuntu i Raspbian Linux jest łatwa dzięki skryptowi powłoki, który można pobrać z //get.docker.com/. Do tego będziesz potrzebować curlpolecenia. Aby uzyskać najnowszą wersję curl:

sudo apt-get install curl

Po curlzainstalowaniu pobierz skrypt instalacyjny i uruchom go:

curl -s //get.docker.com | sudo sh

Po zakończeniu instalacji skryptu zostanie wyświetlona uwaga podobna do poniższej, zawierająca szczegółowe informacje na temat wersji Dockera, zarówno klienta, jak i serwera:

Zwróć uwagę na szczegóły u dołu dotyczące dodawania użytkowników innych niż root do platformy Docker. Jest to wygodne, ale jeśli to zrobisz, zaleca się utworzenie użytkownika innego niż root specjalnie do pracy z Dockerem i bez żadnej innej funkcji. Jednak ze względu na ten samouczek trzymam się używania sudodo uruchamiania Dockera przez nieuprzywilejowanego użytkownika.

Teraz możesz przetestować podstawowy kontener Docker:

$ sudo docker run -i -t ubuntu /bin/bash

To polecenie pobiera ogólny obraz Docker Ubuntu (zgodnie z ubuntuparametrem) i uruchamia /bin/bashpolecenie w tym kontenerze. -iI -topcje otworzyć standardowe wejście i pseudo TTY odpowiednio. 

Jeśli się powiedzie, nazwa hosta w wierszu poleceń powinna zmienić się na coś podobnego [email protected]:/#, co wskazuje numer identyfikacyjny (i nazwę hosta) nowego działającego kontenera. Aby wyjść, wpisz exittak samo, jak przy opuszczaniu dowolnej sesji powłoki.

Powinieneś teraz mieć funkcjonalną instalację Dockera na swoim serwerze. Możesz to przetestować i uzyskać podstawowe informacje za pomocą docker infopolecenia:

$ sudo docker info

The output of the docker info command shows the number of containers and images, among other pertinent information. Note that it may be quite lengthy; this example shows only the last of two pages.

One last change you will need to make if you’re running Ubuntu’s UFW firewall is to allow for packet forwarding. You can check whether UFW is running by entering the following:

$ sudo ufw status

If the command returns a status of inactive, you can skip this next step. Otherwise you will need to edit the UFW configuration file /etc/default/ufw and change the policy for forwarding from DROP to ACCEPT. To do this using the Nano editor, enter the following:

$ sudo nano /etc/default/ufw

And change this line:

DEFAULT_FORWARD_POLICY="DROP"

To this:

DEFAULT_FORWARD_POLICY="ACCEPT"

Save the file, then run:

$ sudo ufw reload

Work with Docker images and Docker containers

Docker containers are much more efficient than virtual machines. When a container is not running a process, it is completely dormant. You might think of Docker containers as self-contained processes—when they’re not actively running, they consume no resources apart from storage.

You can view active and inactive containers using the docker ps command:

# This command will show ALL containers on the system

$ sudo docker ps  -a

# This will show only RUNNING containers

$ sudo docker ps       

You can view all available commands by simply entering docker. For an up-to-date rundown of all commands, their options, and full descriptions, consult the official command-line client documentation.

When I ran docker run earlier, that command automatically pulled an Ubuntu container image from the Docker Hub registry service. Most of the time, though, you’ll want to pull container images into the local cache ahead of time, rather than do that on demand. To do so, use docker pull, like this:

$ sudo docker pull ubuntu

A full, searchable list of images and repositories is available on the Docker Hub.

Docker images vs. containers

Something worth spelling out at this point is how images, containers, and the pull/push process all work together.

Docker containers are built from images, which are essentially shells of operating systems that contain the necessary binaries and libraries to run applications in a container.

Images are labeled with tags, essentially metadata, that make it easy to store and pull different versions of an image. Naturally, a single image can be associated with multiple tags: ubuntu:16.04, ubuntu:xenial-20171201, ubuntu:xenial, ubuntu:latest.

When I typed docker pull ubuntu earlier, I pulled the default Ubuntu image from the Ubuntu repository, which is the image tagged latest. In other words, the command docker pull ubuntu is equivalent to docker pull ubuntu:latest and (at the time of this writing) docker pull ubuntu:xenial

Note that if I had typed: 

$ sudo docker pull -a ubuntu

I would have puledl all images (the -a flag) in the Ubuntu repository into my local system. Most of the time, though, you will want either the default image or a specific version. For example, if you want the image for Ubuntu Saucy Salamander, you’d use docker pull -a ubuntu:saucy to fetch the image with that particular tag from that repo.

The same logic behind repos and tags applies to other manipulations of images. If you pulled saucy as per the above example, you would run it by typing sudo docker run -i -t ubuntu:saucy /bin/bash. If you type sudo docker image rm ubuntu, to remove the ubuntu image, it will remove only the image tagged latest . To remove images other than the default, such as Ubuntu Saucy, you must include the appropriate tag:

sudo docker image rm ubuntu:saucy

Docker image and container workflow

Back to working with images. Once you’ve pulled an image, whatever it may be, you create a live container from it (as I’ve shown) by executing the docker run command. After you have added software and changed any settings inside the container, you can create a new image from those changes by using the docker commit command.

It’s important to note that Docker only stores the deltas, or changes, in images built from other images. As you build your own images, only the changes you make to the base image are stored in the new image, which links back to the base image for all its dependencies. Thus you can create images that have a virtual size of 266MB, but take up only a few megabytes on disk, due to this efficiency.

Fully configured containers can then be pushed up to a central repository to be used elsewhere in the organization or even shared publicly. In this way, an application developer can publish a public container for an app, or you can create private repositories to store all the containers used internally by your organization.

Create a new Docker image from a container

Now that you have a better understanding of how images and containers work, let’s set up a Apache web server container and make it permanent.

Start with a new Docker container

First, you need to build a new container. There are a few ways to do this, but because you have a few commands to run, start a root shell in a new container:

$ sudo docker run -i -t --name apache_web ubuntu /bin/bash

This creates a new container with a unique ID and the name apache_web. It also gives you a root shell because you specified /bin/bash as the command to run. Now install the Apache web server using apt-get:

[email protected]:/# apt-get update

[email protected]:/# apt-get install apache2

Note that you don’t need to use sudo, because your’re running as root inside the container. Note that you do need to run apt-get update, because, again, the package list inside the container is not the same as the one outside of it.

The normal apt-get output appears, and the Apache2 package is installed in your new container. Once the install has completed, start Apache, install curl, and test the installation, all from within your container:

[email protected]:/# service apache2 start

[email protected]:/# apt-get install curl

[email protected]:/# curl //localhost

Following the last command, you should see the raw HTML of the default Apache page displayed in the console. This means our Apache server is installed and running in your container.

If you were doing this in a production environment, you’d next configure Apache to your requirements and install an application for it to serve. Docker letd directories outside a container be mapped to paths inside it, so one approach is to store your web app in a directory on the host and make it visible to the container through a mapping.

Create a startup script for a Docker container

Remember that a Docker container runs only as long as its process or processes are active. So if the process you launch when you first run a container moves into the background, like a system daemon, Docker will stop the container. Therefore, you need to run Apache in the foreground when the container launches, so that the container doesn’t exit as soon as it fires up.

Create a script, startapache.sh, in /usr/local/sbin: 

# You might need to first install Nano inside the container

[email protected]:/# apt-get install nano

[email protected]:/# nano /usr/local/sbin/startapache.sh

In the startapache.sh file, add these lines:

#!/bin/bash

. /etc/apache2/envvars

/usr/sbin/apache2 -D FOREGROUND

Write the changes and save the file. Then make it executable:

[email protected]:/# chmod +x /usr/local/sbin/startapache.sh

All this small script does is bring in the appropriate environment variables for Apache and start the Apache process in the foreground.

You’re done modifying the contents of the container, so you can leave the container by typing exit. When you exit the container, the container will stop.

Commit the container to create a new Docker image

Now you need to commit the container to save the changes you’ve made:

$ sudo docker commit apache_web local:apache_web

The commit will save your container as a new image and return a unique ID. The argument local:apache_web will cause the commit to be placed in a local repository named local with a tag of apache_web.

You can see this by running the command sudo docker images:

REPOSITORY  TAG         IMAGE ID      CREATED      VIRTUAL SIZE

local       apache_web  d95238078ab0  4 minutes ago  284.1 MB

Note that the exact details of your image—the image ID, the size of the container—will be different from my example.

Docker containers are designed to be immutable. Whenever you commit changes to a container, the results are written out to an entirely new container, never to the original. If you want to swap out Apache with, say, Nginx, you would start with the original ubuntu:latest container, add Nginx to that, and save out the results as an all-new container named something like local:nginx.

Understand Docker networking basics

Now that you have our image, you can start our container and begin serving pages. Before you do, however, let me take a moment to explain how Docker handles networking.

When Docker is installed, it creates three virtual networks that can be used by Docker containers:

  • bridge: This is the network that containers connect to by default. The bridge network allows containers to talk to each other directly, but not to the host system.
  • host: This network lets containers be seen by the host directly, as if any apps within them were running as local network services.
  • none: This is essentially a null or loopback network. A container connected to none can’t see anything but itself.

When you want to launch a container and have it communicate with both other containers and the outside world, you need to manually map ports from that container to the host. For the sake of my example, you can do this on the command line when you launch your newly created container:

$ sudo docker run -d -p 8080:80 --name apache local:apache_web /usr/local/sbin/startapache.sh