TheLazy.Dev
Portfolio GitHub

Blog

docker |tutorial |networking

Docker network and network namespaces in practice

rimelek

Ákos Takács

Posted on June 21, 2023

Docker network and network namespaces in practice

Introduction

This tutorial was originally posted on learn-docker.it-sziget.hu, which is hosted on Read the Docs. I share it here, on dev.to as well, because as a current Docker forum moderator, I can see how many problems people can have because of the container networks. I believe it has to reach a wider audience since networking is hard even without containers and containers can make it more complicated. You can also learn about some really useful tricks to test untrusted applications in a more secure environment.

As usually, I also have a video about this topic.

Table of content

Linux Kernel Namespaces in general

» Back to table of contents «

Slide 1.1

Let's say, you have a running PHP process on your Linux machine. That PHP process will be able to see your entire host, including files, other running processes and all the network interfaces. It will not just see the interfaces, it will be able to listen on all the IP addresses, so you have to configure PHP to listen on a specific IP address. PHP will also have internet access which is usually what you want, but not always.

This is when Linux kernel namespaces can help us.A kernel namespace is like a magic wall between a running process and the rest of the host. This magic wall will only hide some parts of the host from a specific point of view. In this tutorial we will mainly cover the three best known point of views.

  • Filesystem (Mount namespace)
  • Network (Network namespace)
  • Process (PID namespace, which means "Process ID" namespace)

The first namespace is mount namespace. You could also remember it as "Jail" or "chroot". It means the process will see only a folder on the host and not the root filesystem. That folder could be empty, but it is very often a folder that contains very similar files as the root filesystem does. This way PHP will "think" it is running on a different host and it won't even know if there is anything outside that folder.

Slide 1.2

This is not always enough. Sometimes you don't want a specific process to see other processes on the host. In this case you can use the PID namespace so the PHP process will not be able to see other processes on the host, only the processes in the same PID namespace. Since on Linux there is always a process with the process ID 1 (PID 1), for any process in the network namespace, PHP will have PID 1. For any process outside of that PID namespace, PHP will have a larger process ID like 2324.

Slide 1.3

And finally we arrived to the network namespace. Network namespace can hide network interfaces from processes in the namespace. The network namespace can have its own IP address, but it will not necessarily have one. Thanks to the network namespace, PHP will only be able to listen its own IP address in a Docker network.

Slide 1.4

These namespaces are independent and not owned by the PHP process. Any other process could be added to these namespaces, and you don't need to run those processes in each namespace.

Slide 1.5

You can choose the network namespace even without the mount namespace, so you can use an application (web browser, curl) on the host and run it in the network namespace of the PHP process, so even if the PHP is not available outside of the container, running the browser in the network namespace of the container will allow you to access your website.

Network traffic between a container and the outside world

» Back to table of contents «

Your host machine usually has a network interface which is connected to the internet or at least to a local network. We will call it "the outside world".

Slide 2.1

This interface could be "eth0", although recently it is more likely to be something like "enp1s0" or "eno1". The point is that traffic routed through this interface can leave the host machine. The container needs its own network interface which is connected to another outside of the container on the host. The name of the interface on the host will start with "veth" followed by a hash.

Slide 2.2

The veth interface will not have IP address. It could have, but Docker uses a different way so containers in the same Docker network can communicate with each other. There will be a bridge network between the veth interface and eth0. The the bridge of the default Docker network is "docker0".

Slide 2.3

This bridge will have an IP address which will also be the gateway for each container in the default Docker network. Since this is on the host outside of the network namespace, any process running on the host could listen on this IP address so processes running inside the container could use this IP to access a webservice listening on it.

Sometimes you don't want a containerized process to access the internet, because you don't trust an application and you want to test or run it without internet access for security reasons indefinitely. This is when "internal networks" can help.

Slide 2.4

Containers don't accept port forwards on IP addresses in internal networks so it is not just rejecting outgoing traffic to the outside world, but also rejecting incoming requests from other networks.

You can create a user-defined Docker network which will have a new bridge. If you also define that network as "internal" using the following command for example

docker network create secure_net --internal
Enter fullscreen mode Exit fullscreen mode

the network traffic will not be forwarded from the bridge to eth0 so PHP will only be able to access services running on the host or in the container.

There is another very important interface called "lo" better known as "localhost" which usually has an IP address like 127.0.0.1. This is however not the only IP address that is bound to this interface. Every part of the IP could be changed after 127. and still pointing to the same interface. It is important to know that every network namespace has its own localhost.

Let's see what it means.

Slide 2.6

If the web application is listening on port 80 on localhost, a web browser outside of the container will not be able to access it, since it has a different localhost. The same is true when for example a database server is running on the host listening on port 3306 on localhost. The PHP process inside the container will not be able to reach it.

Slide 2.7

Since the reason is the network namespace, you could just run the container in host network mode

docker run -d --name php-hostnet --network host rimelek/phar-examples:1.0
Enter fullscreen mode Exit fullscreen mode

which means you just don't get the network isolation. The host network mode does not mean that you are using a special Docker network. It only means you don't want the container to have its own network namespace.

Slide 2.8

Of course we wanted to have the network isolation and we want to keep it. The other solution is running another container which will use the same network namespace.

Slide 2.9

Manipulating network namespaces

» Back to table of contents «

Docker is not the only tool to manipulate namespaces. I will show you the following tools.

  • Container engines (Docker)
  • "ip" command
  • "nsenter" command
  • "unshare" command

Two containers using the same network namespace

» Back to table of contents «

Of course the first we have to talk about is still Docker. The following commands will start a PHP demo application and run a bash container using the same network namespace as the PHP container so we can see the network interfaces inside the PHP container.

docker run -d --name php rimelek/phar-examples:1.0
docker run --rm -it --network container:php bash:5.1 ip addr
Enter fullscreen mode Exit fullscreen mode

There is a much easier solution of course. We can just use docker exec to execute a command in all of the namespaces of the PHP container.

docker exec php ip addr
Enter fullscreen mode Exit fullscreen mode

This command works only because "ip", which is part of the "iproute2" package is installed inside the PHP container, so it wouldn't work with every base image and especially not with every command.

"nsenter": run commands in any namespace

» Back to table of contents «

The "nsenter" (namespace enter) command will let you execute commands in specific namespaces. The following command would execute ip addr in the network namespace of a process which has the process ID $pid.

sudo nsenter -n -t $pid ip addr
Enter fullscreen mode Exit fullscreen mode

We have to get the id of a process running inside a container. Remember, the process has a different ID inside and outside of the container because of the PID namespace, so we can't just run the ps aux command inside the container. We need to "inspect" the PHP container's metadata.

pid="$(docker container inspect php --format '{{ .State.Pid }}')"
Enter fullscreen mode Exit fullscreen mode

The above command will save the process ID in the environment variable called "pid". Now let's run nsenter again.

sudo nsenter -n -t $pid ip addr
sudo nsenter -n -t $pid hostname
sudo nsenter -n -u -t $pid hostname
Enter fullscreen mode Exit fullscreen mode

The first command will show us the network interfaces inside the network namespace of the PHP container. The second command will try to get the hostname of the container, but it will return the hostname of the host machine. Although the hostname is related to the network in our mind, it is not part of the network namespace. It is actually the part of the UTS namespace. Since the long name of the namespace would just confuse you, I will not share it at this point of the tutorial. The good news is that we can also use the UTS namespace of the container by adding the -u flag to the "nsenter" command, and this is what the third line does.

"ip netns" to create new network namespaces

» Back to table of contents «

"nsenter" was great for running commands in existing namespaces. If you want to create network namespaces, you can use the ip netns command, but before we create one, let's list existing network namespaces:

ip netns list
Enter fullscreen mode Exit fullscreen mode

The above command will give you nothing even if you have running containers using network namespaces. To understand why, first let's look at content of two folders

ls /run/netns
sudo ls /run/docker/netns
Enter fullscreen mode Exit fullscreen mode

The first line, used by the "ip" command will not give you anything, but the second will give you at least one file, which is the file of the network namespace of our previously
started PHP container.

As you can see, if you want to work with namespaces, you need to refer to a file or the name of the file. Docker and the "ip" command uses a different folder to store those files. These files are not the only way to refer to network namespaces and we will discuss it later.

It's time to create our first network namespace without Docker.

sudo ip netns add test
ls /run/netns
Enter fullscreen mode Exit fullscreen mode

The "ls" command isn't required here, but can show us that we indeed created a file. Let's run ip addr inside our new network namespace:

sudo ip netns exec test ip addr
Enter fullscreen mode Exit fullscreen mode

Note: You could actually use nsenter to run ip addr in a network namespace even if you don't have an existing process.

sudo nsenter --net=/run/netns/test ip addr
Enter fullscreen mode Exit fullscreen mode

The output will be

1: lo: <LOOPBACK> mtu 65536 qdisc noop state DOWN group default qlen 1000
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
Enter fullscreen mode Exit fullscreen mode

As you can see this new network namespace doesn't even have a loopback IP address so basically it doesn't have "localhost". It shows us that a network namespace does not give us a fully configured private network, it only gives us the network isolation. Now that we know it, it is not surprising that the following commands will give us error messages.

sudo ip netns exec test ping dns.google
# ping: dns.google: Temporary failure in name resolution
sudo ip netns exec test ping 8.8.8.8
# ping: connect: Network is unreachable
Enter fullscreen mode Exit fullscreen mode

Since this network namespace is useless without further configuration and configuring the network is not part of this tutorial, we can delete it:

sudo ip netns del test
Enter fullscreen mode Exit fullscreen mode

"unshare": Temporary network namespace creation

» Back to table of contents «

If you want to create a temporary network namespace and run a command inside it, you can use unshare. This command has similar parameters as nsenter but it doesn't require existing namespaces. It will create new namespaces for the commands that you want to run. IT could be useful when you just want to test an application that you it shouldn't use the network so you can run it in a safer environment.

sudo unshare -n ip addr
Enter fullscreen mode Exit fullscreen mode

It will give you the same output as our previous attempt to create a network namespace.

1: lo: <LOOPBACK> mtu 65536 qdisc noop state DOWN group default qlen 1000
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
Enter fullscreen mode Exit fullscreen mode

Working with Docker's network namespaces

Allow the "ip" command to use Docker's network namespaces

» Back to table of contents «

If you want, you could remove /run/netns and create a symbolic link instead pointing to /run/docker/netns.

sudo rm -r /run/netns
sudo ln -s /run/docker/netns /run/netns
ip netns list
Enter fullscreen mode Exit fullscreen mode

Sometimes you can get an error message saying that

rm: cannot remove '/run/netns': Device or resource busy
Enter fullscreen mode Exit fullscreen mode

Since we started to use the "ls" and "ip" commands to list namespaces, it is likely that we get this error message even though we are not actively using that folder. There could be two solutions to be able to remove this folder:

  • Exiting from current shell and opening a new one
  • Rebooting the machine

The first will not always work, and the second is obviously something that you can't do with a running production server.

A better way of handling the situation is creating symbolic links under /run/netns pointing to files under /run/docker/netns. In Docker's terminology the file is called "sandbox key". We can get the path of a container's sandbox key by using the following command:

sandboxKey=$(docker container inspect php --format '{{ .NetworkSettings.SandboxKey }}')
Enter fullscreen mode Exit fullscreen mode

The end of that path is the filename which we will need to create a link under /run/netns.

netns=$(basename "$sandboxKey")
Enter fullscreen mode Exit fullscreen mode

Using the above variables we can finally create our first symbolic link

sudo ln -s $sandboxKey /run/netns/$netns
Enter fullscreen mode Exit fullscreen mode

Finally, ip netns ls will give us an output similar to the following:

a339e5fc43f0 (id: 0)
Enter fullscreen mode Exit fullscreen mode

Name resolution issue with "ip netns exec"

» Back to table of contents «

It's time to run ip netns exec to test the network of a Docker container.

sudo ip netns exec $netns ip addr
sudo ip netns exec $netns ping 8.8.8.8
sudo ip netns exec $netns ping dns.google
Enter fullscreen mode Exit fullscreen mode

The first two lines will give the expected results, but the third line will give us the following error message.

ping: dns.google: Temporary failure in name resolution
Enter fullscreen mode Exit fullscreen mode

What happened?

Slide 3.1

We ran the ping command only in the network namespace of the container, which means the configuration files that are supposed to control how name resolution works are loaded from the host. My host was an Ubuntu 20.04 LTS virtual machine created by Multipass. By default, the IP address of the nameserver is 127.0.0.53. Remember, that this IP address belongs to the loopback interface which is different in each network namespace. In the network namespace of our PHP container there is no service listening on this IP address.

Solution 1: Change the configuration on the host

» Back to table of contents «

!!! DO NOT !!! test it in a production environment as it could also break your name resolution if you are doing something wrong.

/etc/resolv.conf is usually a symbolic link pointing one of the following files:

  • /run/systemd/resolve/stub-resolv.conf
  • /run/systemd/resolve/resolv.conf

Depending on your system it could point to an entirely different file or it could also be a regular file instead of a symbolic link. I will only discuss the above files in this tutorial.

Run the following command to get the real path of the configuration file.

readlink -f /etc/resolv.conf
Enter fullscreen mode Exit fullscreen mode

Note: Alternatively, you could also run realpath /etc/resolv.conf.

If the output is /run/systemd/resolve/stub-resolv.conf, you are using the stub resolver and the content of the file looks like this without the comments:

nameserver 127.0.0.53
options edns0 trust-ad
search .
Enter fullscreen mode Exit fullscreen mode

On the other hand, /run/systemd/resolve/resolv.conf will directly contain the nameservers:

nameserver 192.168.205.1
search .
Enter fullscreen mode Exit fullscreen mode

Now I will change the symbolic link:

sudo unlink /etc/resolv.conf
sudo ln -s /run/systemd/resolve/resolv.conf /etc/resolv.conf
Enter fullscreen mode Exit fullscreen mode

After this I will be able to successfully ping the domain name of Google's name server:

sudo ip netns exec $netns ping dns.google
Enter fullscreen mode Exit fullscreen mode

I don't want to keep this configuration, so I will restore the stub resolver:

sudo unlink /etc/resolv.conf
sudo ln -s /run/systemd/resolve/stub-resolv.conf /etc/resolv.conf
Enter fullscreen mode Exit fullscreen mode

Solution 2: Using per-namespace resolv.conf

» Back to table of contents «

We can create additional configuration files for each network namespace.
First we have to create a new folder using the name of the namespace undr /etc/netns

sudo mkdir -p /etc/netns/$netns
Enter fullscreen mode Exit fullscreen mode

After that we have to create a resolv.conf file in the new folder and add a nameserver definition like nameserver 8.8.8.8

echo "nameserver 8.8.8.8" | sudo tee /etc/netns/$netns/resolv.conf
Enter fullscreen mode Exit fullscreen mode

And finally we can ping the domain name

sudo ip netns exec $netns ping dns.google
Enter fullscreen mode Exit fullscreen mode

Solution 3: Using a custom mount namespace based on the original root filesystem

» Back to table of contents «

This is a very tricky solution which I would not recommend usually, but it could be useful to learn about the relation of different types of namespaces. The solution is based on the following facts.

  • The "nsenter" command allows us to define a custom root directory (mount namespace) instead of using an existing mount namespace
  • The "mount" command has a --bind flag which allows us to "bind mount" a folder to a new location. This is similar to what Docker does if you choose "bind" as the type of a volume. See Bind mounts | Docker
  • There are some folders that are not part of the root filesystem, so when we mount the root filesystem we don't mount those folders. /run is on tmpfs, so it is stored in memory.
  • Mounting a file over a symbolic link is not possible, but mounting over an empty file which is a target of a symbolic link works.

Slide 3.2

First we will set the variables again with an additional project_dir which you can change if you want

sandboxKey=$(docker container inspect php --format '{{ .NetworkSettings.SandboxKey }}')
pid="$(docker container inspect php --format '{{ .State.Pid }}')"

project_dir="$HOME/projects/netns"
Enter fullscreen mode Exit fullscreen mode

Then we create the our project directory

mkdir -p "$project_dir"
cd "$project_dir"
Enter fullscreen mode Exit fullscreen mode

Mount the system root to a local folder called "root".

mkdir -p root
sudo mount --bind / root
Enter fullscreen mode Exit fullscreen mode

Since "run" is on tmpfs and it wasn't mounted, we create an empty file to work as a placeholder for the target of the symbolic link at /etc/resolv.conf

sudo mkdir -p "root/run/systemd/resolve/"
sudo touch "root/run/systemd/resolve/stub-resolv.conf"
Enter fullscreen mode Exit fullscreen mode

Now we can copy the resolv.conf that contains the actual name servers and mount it over our placeholder stub-resolv.conf.

cp "/run/systemd/resolve/resolv.conf" "resolv.conf"
sudo mount --bind "resolv.conf" "root/run/systemd/resolve/stub-resolv.conf"
Enter fullscreen mode Exit fullscreen mode

And finally we can run the following nsenter command.

sudo nsenter -n --root=$PWD/root --target=$pid ping dns.google
Enter fullscreen mode Exit fullscreen mode

Now nsenter will use $PWD/root as the filesystem of the new mount namespace and use the network namespace of the PHP container to run ping.

PING dns.google (8.8.4.4) 56(84) bytes of data.
64 bytes from dns.google (8.8.4.4): icmp_seq=1 ttl=112 time=11.5 ms
64 bytes from dns.google (8.8.4.4): icmp_seq=2 ttl=112 time=12.1 ms
64 bytes from dns.google (8.8.4.4): icmp_seq=3 ttl=112 time=11.7 ms
Enter fullscreen mode Exit fullscreen mode

Debugging the Minotaur

» Back to table of contents «

Slide 4.1

I call this technique "Debugging the Minotaur" because unlike before when we ran a new container to attach it to another container's network namespace, we are still on the host and we use most of the host's namespaces and we choose to use one container's mount namespace (and only the mount namespace) and another container's network namespace (and only the network namespace). As we were creating a Minotaur where the body of the Minotaur is the mount namespace of the debugger container with all of its tools and the head is the other container's network namespace which we want to debug. To do this, we use only nsenter and nothing else.

Slide 4.2

We know that we can use an executable on the host's filesystem and run it in a network namespace. We can also choose the mount namespace and that can be the filesystem of a running container. First we want to have a running debugger container. nicolaka/netshoot is an excellent image to start a debugger container from. We need to run it in detached mode (-d) so it will run in the background (not attaching to the container's namespaces) and also in interactive mode (-i) so it will keep running instead of exiting immediately.

docker run -d -i --name debug nicolaka/netshoot:v0.9
Enter fullscreen mode Exit fullscreen mode

Now we need to get the sandbox key for the network namespace and since we want to debug the PHP container, we will get the sandbox key from it. We also need something for the mount namespace of the debugger container. This is a good time to learn that if we have an existing process, we can find all of its namespaces using a path like this:

/proc/<PID>/ns/<NAMESPACE>
Enter fullscreen mode Exit fullscreen mode

where <PID> is the process id and <NAMESPACE> in case of the discussed best known namespaces is one of the followings: mnt, net, pid. We could use /proc/$pid/ns/net instead of the sandbox key, but in this example I will keep it to demonstrate that you can do both.

php_sandbox_key=$(docker container inspect php --format '{{ .NetworkSettings.SandboxKey }}')
debug_pid=$(docker container inspect debug --format '{{ .State.Pid }}')
Enter fullscreen mode Exit fullscreen mode

Now that we have the variables, let's use nsenter a new way. So far we used the sandbox key only to help the ip command to recognize the network namespaces. Now we have to refer to it directly and nsenter can do that.

sudo nsenter --net=$php_sandbox_key --mount=/proc/$debug_pid/ns/mnt ping dns.google
Enter fullscreen mode Exit fullscreen mode

This way we have a ping command running, but sometimes we need to do more debugging. The ping command is almost always available on Linux systems, although you can use tshark or tcpdump to see the network packets, but I prefer to use tshark. The following command will show us packets going through the debugger container's eth0 interface so you can actually see the source of everything before those packets are reaching the veth* interface on the host. Since you can use tshark from the debugger container, you don't have to install it. In case you have a more advanced debugger script which for some reason needs to access other namespaces on the host, you can do that too.

sudo nsenter --net=$php_sandbox_key --mount=/proc/$debug_pid/ns/mnt tshark -i eth0
Enter fullscreen mode Exit fullscreen mode

As a final step, open a new terminal and generate some traffic on the container network. Get the ip address of the container and use curl to get the main page of the website in the container.

ip=$(docker container inspect php --format '{{ .NetworkSettings.IPAddress }}')
curl "$ip"
Enter fullscreen mode Exit fullscreen mode

As a result, in the previous terminal window you should see the request packets and the response.

Testing a web-based application without internet in a container

Running a web browser in a net namespace on Linux (Docker CE)

» Back to table of contents «

If you are running Docker CE on Linux (not Docker Desktop), you can just use a web browser on your host operating system and run it in the network namespace of a container. If the application inside is listening on localhost, you can access it from the web browser in the same network namespace.

Slide 5.1

docker run -d --name php-networkless --network none rimelek/phar-examples:1.0
pid_networkless=$(docker container inspect php --format '{{ .State.Pid }}')
sudo nsenter --net=/proc/$pid_networkless/ns/net curl localhost
Enter fullscreen mode Exit fullscreen mode

Or sometimes you know that the frontend is safe to use, so you only want to test the backend.

Slide 5.2

In that case you can run the container with network, but only with an "internal" network, so the host and the container can communicate, but no traffic will be forwarded to the internet from the Docker bridge. This way you can run your browser "on the host" and use the container's ip address instead of "localhost".

Note: Actually everything is running on the host. Only the isolated processes will see it differently.

You need to

  • create an internal network,
  • run the container using the internal network
  • get the ip address of the container
  • open your web browser or use curl to access the website 
docker network create internal --internal
docker run -d --name php-internal --network internal rimelek/phar-examples:1.0
ip_internal=$(docker container inspect php-internal --format '{{ .NetworkSettings.Networks.internal.IPAddress }}')
curl "$ip_internal"
Enter fullscreen mode Exit fullscreen mode

Since curl will not execute javascript, you can even check the generated source code, but nothing in the container will
be able to send request to the outside world except the host machine:

docker exec php-internal ping 8.8.8.8
Enter fullscreen mode Exit fullscreen mode

Running a web browser in a net namespace in a VM (Docker Desktop)

» Back to table of contents «

When you start to use Docker Desktop, one of the most important facts is that your containers will run in a virtual machine even on Linux (See: Getting Started: Docker Desktop). It means your actual host, the virtual machine and the container's will have their own "localhost".

Slide 6.1

The network namespaces will be in that virtual machine, so you can't just run your web browser on your host operating system inside the network namespace.

Slide 6.2

You can't even run the web browser in the virtual machine (in case of Docker Desktop) since that is just a server based on LinuxKit without GUI inside so you can't simply just use an internal network and connect to the IP address from the browser.

Slide 6.3

We need a much more complex solution which requires everything that we have learnt so far and more.

  • We know that our PHP app has to run in a container without internet access
  • We also know that we can achieve that by using internal networks or no network at all except loopback interface.
  • Since Docker Desktop runs containers in a virtual machine, we definitely need network in the PHP container so we can access it from the outside. It means we obviously need to forward a port from the host to Docker Desktop's virtual machine, but we have also learnt that internal networks do not accept forwarded ports.
  • We can however run a container with only an internal network and a proxy container with an internal and a public network which will be accessible from the outside. This container will forward all traffic to another container in the internal network.
  • There is a way to run a web browser in a container and you can run this container in the PHP container's network namespace. The problem is that you need to access the graphical interface inside the container.
  • Fortunately there is also a sponsored OSS project called linuxserver/firefox. This project let's you run Firefox and a remote desktop server in the container.

How will this all look like? The following diagram illustrates it.

Slide 6.4

  • You will use a web browser on the host as a remote desktop client to access the forwarded port of the proxy server on the IP address of the public network.
  • The PHP container will have an internal network
  • The Firefox container with the remote desktop client will use the network namespace of the PHP container so Firefox will not have internet access.
  • The proxy server (with both internal and public network) will forward your request to the PHP container's network namespace to access the remote desktop server.
  • The remote desktop server will stream back the screen only through the proxy server so the graphical interface of the containerized Firefox will appear in the web browser running on your host. If a harmful application tries to use javaScript to access another website it won't be able to since all you can see is a picture of a web browser running in an isolated environment.

I have created a compose file which we can use to create this whole environment.

Create a project folder anywhere you like. This is mine:

project_dir="$HOME/Data/projects/testprojects/netns"
mkdir -p "$project_dir"
cd "$project_dir"
Enter fullscreen mode Exit fullscreen mode

Download the compose file from GitHub

curl --output compose.yml \ 
     https://gist.githubusercontent.com/rimelek/91702f6e9c9e0ae75a72a42211099b63/raw/339beaf0c50790e86ab8a011ed298c250da3b7ec/compose.yml
Enter fullscreen mode Exit fullscreen mode

Compose file content:

networks:
  default:
    internal: true
  public:

services:
  php:
    image: rimelek/phar-examples:1.0

  firefox:
    network_mode: service:php
    environment:
      PUID: 1000
      PGID: 1000
      TZ: Europe/London
    shm_size: "1gb"
    image: lscr.io/linuxserver/firefox:101.0.1

  proxy:
    image: alpine/socat:1.7.4.4-r0
    command: "TCP-LISTEN:3000,fork,reuseaddr TCP:php:3000"
    ports:
      - 3000:3000
    networks:
      - default
      - public
Enter fullscreen mode Exit fullscreen mode

Start the containers:

docker compose up -d
Enter fullscreen mode Exit fullscreen mode

If you open localhost:3000 in your browser, you will see the containerized browser and the demo application without CSS and JavaScript since those files would be loaded from an external source and they are not available.

Slide 6.5

Now that you know it is trying to load CSS and some harmless JavaScripts, you can run it with a public network

docker run -d --name php-internet -p 8080:80 rimelek/phar-examples:1.0
Enter fullscreen mode Exit fullscreen mode

and open it in an other tab on port 8080.

Slide 6.6

Used sources

» Back to table of contents «

Recommended similar tutorials

» Back to table of contents «

💖 💪 🙅 🚩
rimelek
Ákos Takács

Posted on June 21, 2023

Join Our Newsletter. No Spam, Only the good stuff.

Sign up to receive the latest update from our blog.

Related

Docker network and network namespaces in practice
docker Docker network and network namespaces in practice

June 21, 2023