Author Archives: alienbob

Slackware Cloud Server Series, Episode 1: Managing your Docker infrastructure

Hi all!
This is the first installment of a series of articles I intend to write in early 2022. With Slackware 15.0 around the corner, I think it is a good time to show people that Slackware is as strong as ever a server platform. The core of this series is not about setting up a mail, print or web server – those are pretty well-documented already. I’m going to show what is possible with Slackware as your personal cloud platform.

A lot of the work that went into developing Slackware between the 14.2 and 15.0 releases was focusing on desktop usage. The distro is equipped with the latest and greatest KDE and XFCE desktops, a low-latency preemptive kernel, Pipewire multimedia framework supporting capture and playback of audio and video with minimal latency, et cetera. I have been enjoying Slackware-current as a desktop / laptop powerhouse for many years and built a Digital Audio Workstation and a persistent encrypted Live Distro with SecureBoot support out of it.

Slackware Cloud Server Series – Summary

The imminent release of Slackware 15.0 gives me fresh energy to look into ‘uncharted’ territory. And therefore I am going to write about setting up collaborative web-based services. Whether you will do this for the members of your family, your friends, your company or any group of people that you interact with a lot, this year you will hopefully learn how to become less dependent on the big ‘cloud’ players like Google, Microsoft, Zoom, Dropbox. What I will show you, is how to setup your own collaboration platform on a server or servers that you own and control. Think of file-sharing, video-conferencing, collaborative document editing, syncing files from your desktop and your phone to your cloud server. If the series is received well, I may additionally write about expanding your cloud server to a private platform for watching movies, listening to network audio, and so on.

I have a good notion of what I will write about, but I am sure that as this series grows and I tell you stories, you will get triggered about subjects that I may not have considered yet. So here’s a challenge for you, the reader: let me know what you think would be a good addition to the series. What makes Slackware great as a “personal cloud” platform to bind people together.

Topics you can expect in this series:

  • Episode 1 (this article): Managing your Docker Infrastructure
    Here I will show you how to setup and manage Docker containers on Slackware, considering graphical management and firewalling. This is the foundation for what we will be doing in subsequent episodes.

    • Docker architecture
      • Examples
    • Docker Compose
    • Building a Docker image and running a container based on it
    • Help on commands
    • Obtain information about images and containers
    • Stop and remove a container
    • Refresh a container with the latest image from the vendor
    • Managing  containers, images, volumes, networks etc
    • Difference between load/save and import/export
    • Limiting the log size for your containers
    • Graphical management tool
    • Docker’s effect on iptables firewall
      • Generic iptables
      • UFW (Uncomplicated Firewall)
      • Firewalld
      • Alien’s Easy Firewall Generator (EFG)
    • Thanks
    • Attribution
  • Episode 2: Identity and Access management
    Setting up Keycloak for Identity and Access Management (IAM) to provide people with a single user account for all the Slackware Server Services we will be creating.
  • Episode 3: Video Conferencing
    Setting up Jitsi Meet . This makes us independent of cloud conferencing services like MS Teams, Zoom or Google Meet. The Jitsi login is again offloaded to our Keycloak IAM provider.
  • Episode 4: Productivity Platform
    Setting up NextCloud as the productivity hub where you can store and read your documents, host your photo library, hold video meetings, manage your activities, read and send emails, keep track of where your smartphone went, chat with people and a lot more. I will also show how to integrate the Jitsi video conferencing solution which was the topic of the previous episode into our NextCloud platform.
    NextCloud will be setup on your ‘bare metal’ server. User authentication to the platform will be handled by our Keycloak IAM server.
  • Episode 5: Collaborative document editing
    Integrating Collabora Online Development Edition (CODE) with NextCloud and move from reading/displaying your documents to collaborative real-time document editing with friends, family or colleagues in your self-hosted LibreOffice Online server.
  • Episode 6: Docker Registry
    Setting up a Docker Registry. A local index and registry of Docker images allows us to become independent of Docker’s commercially driven limitations on images you create and want to share with the world or use privately. I will for instance use the public Docker Hub for all the services that are discussed in this series of articles, and I will point out the limitations of doing so.
    I am not decided on the actual implementation of the Registry. The Docker Hub is an implementation of the aforementioned Registry product and it is Open Source but it does not do authentication and it has no user interface. Docker (the company) adds a lot to this service that is not Open Source to make the Hub the popular platform that it became.
    So perhaps I will combine this Registry with Portus, which is an Open Source authorization service and user interface for the Docker Registry. It is created/maintained by the OpenSuse team. It can use Keycloak as the OIDC (OpenID Connect) provider, and so our Keycloak IAM server will do the Single Sign On.
    Or I will use Quay, which is Red Hat’s solution for self-hosted Docker Registry including a web-based user interface that looks a lot like the Docker Hub. Quay can also offload the authentication/authorization to an OIDC provider like Keycloak. In addition, it allows scanning your hosted images for known security vulnerabilities using Clair, another Red Hat product and part of Project Quay.

Managing your Docker Infrastructure

In a previous article, I have shared my Docker related packages with you for Slackware 14.2 and 15.0 (well, -current formally since we are still waiting for 15.0). I assume you are running the newest Slackware almost-at-version 15.0 and have already installed those four Docker packages (containerd, runc, docker and docker-compose), you have added your user account to the ‘docker‘ group and have at least once rebooted your computer or manually ran “/etc/rc.d/rc.docker start“.
I also assume that you have a basic understanding of containers and how they differ from Virtual Machines. A brief summary:

Containers run directly on the host kernel and the ‘bare metal’. The Docker engine hides the host operating system particulars from the applications that run in the containerized environment. Docker uses cgroups and Linux kernel capabilities to shield container processes from other containers and the host, but using a “ps” command on the host you can easily see all the container processes.
A Virtual Machine on the other hand, is a virtualized machine hardware environment created by a hypervisor program like QEMU, VMWare, VirtualBox etc. Inside that virtual machine you can run a full Operating System which thinks it is running on real hardware. It will be unaware that it is in fact running inside a VM on a host computer which  probably runs a completely different OS.
From the host user perspective, nothing that goes on in the VM is actually visible; the user merely sees the hypervisor running.
And from a Docker container as well as VM user perspective, anything that happens outside that environment is invisible. The host OS can not be reached directly from within the guest.

What I am not going to discuss in the scope of this article is orchestration tools like Kubernetes (K8s) or Docker Swarm, that make it possible or at least a lot more convenient to manage a cluster of servers running Docker containers with multiple micro-services that are able to scale on-demand.

Docker architecture

You create and run a Docker container from (layers of) pre-existing images that inherit from each other, using the ‘docker run‘ command. You can save a container that you created, ‘flattening’ the layers the container consists of into one new image using ‘docker commit‘.

You start and run a container by referring the Docker image it should be based on, and if that image is not available locally then Docker will download it from a Docker Registry (by default hub.docker.com but it can also be a private Registry you run and control yourself). You can also manually download an image using the ‘docker pull‘ command in advance of the ‘docker run‘ command which is going to use that image so that you don’t have to wait for an image download at the moment suprême of starting your container.

Building an image yourself using the ‘docker build‘ command is essentially ruled by a Dockerfile and a context (files on the local filesystem which are used in the creation of the image). The container which is based on that image typically runs an application or performs a task.

Examples

An example: running an application which is not provided by your Slackware OS. A container can run a Postgres database server which you then use on your host computer or in other containers. This Postgres container will typically be started on boot and keeps running until shutdown of the host computer. It’s as trivial as running:

$ docker run -d --restart unless-stopped --name slack-postgres -e POSTGRES_PASSWORD=**** postgres

Another example: building Slackware packages reliably and without introducing unwanted dependencies. In this case, you use a container based from a Slackware Docker image, to download and compile a Slackware package, and then export the resulting package from within the container to the host computer’s local filesystem. Such a Docker container based on a Slackware image will be able to offer the same OS with the same configuration, every time it starts and functions independently of the host. You could be compiling Slackware packages on a Debian host for instance.
Exercise left for the reader: you will have to create that Slackware Docker image yourself… it is not available online. 

Docker Compose

Also created by the Docker company, docker-compose is a separate application which uses the capabilities of Docker. The 2.x release of Docker Compose which I offer in my repository is written in the Go language which makes it fully self-contained. The previous 1.x releases that you can obtain from SlackBuilds.org were heavily dependent on Python and for that version of docker-compose you had to install 13 dependent packages.

Docker Compose expands on the Docker concept of a single container with a single purpose. Compose is a tool for defining and running multi-container Docker applications. The individual Docker applications themselves are defined by a Dockerfile, which makes their deployment reproducible in any Docker environment.
You write a Compose file (typically called docker-compose.yml) which defines the container services that you want to run. This Compose file is written in YAML and according to the Compose file specification . The file describes an application service which is built out of multiple Docker containers  as well as their shared resources and private communication channels. Docker-compose reads the yaml file and creates data volumes and internal networks, starts the defined containers and connects them to the just-created volumes and networks. There’s a lot more you can do in a Compose file, check out the reference.

The result of the ‘docker-compose up -d‘ command is the start-up of a complex service running multiple inter-dependent applications (you’ll see those referred to as micro-services) that usually have a single point of interaction with its users (an IP address/port or a https URL). Essentially a black-box where all the internal workings are hidden inside that collection of containers. As an administrator there’s little to nothing to configure and as an user you don’t have to know anything about that complexity.
And again you could be trivially running applications in Docker containers on your Slackware host that you would not be able to run on the bare host OS without a lot of weeping and gnashing of teeth .

Docker Compose offers a lot of versatility that regular Docker does not. Whether you use one or the other depends on the complexity of that what you want to achieve. In future episodes of this article series, I will give examples of both.

Building a Docker image and running a container based on it

The ‘docker build’  command has a lot of parameters but it accepts a single argument which is either a directory containing a Dockerfile and the required context (local files to be used in the image creation process), or a URL the content of which it will download and interpret as a Dockerfile. You can also use the “-f” switch to specify a filename if it’s not called “Dockerfile” literally, or pipe a Dockerfile content into the ‘docker build‘ command via STDIN. It’s good practice to ‘tag’ the resulting image so that you’ll understand its purpose later on. You can then upload or ‘push’ the image to a Docker Registry, either the public one or a private registry that you or your company control.
Let me give a comprehensive example. We will build a new Slackware image which is based on my Slackware ‘base‘ image on the public Docker Hub (docker pull liveslak/slackware:latest). The Dockerfile will add a couple of packages, create a user account called “alien” and start a login shell if needed.
Note that in this example I will write the Dockerfile on the fly using ‘cat‘ and a ‘here-document‘ to pipe it into the ‘docker build‘ command. I will tag the resulting image as ‘slacktest‘ using the “-t” switch of the ‘docker build‘ command:

$ cat <<EOT | docker build -t slacktest -
FROM liveslak/slackware:latest 
MAINTAINER Eric Hameleers <alien@slackware.com> 
ARG SL_UID="1000" 
ARG SL_GID="users" 
ARG SL_USER="alien" 
# Install compiler toolchain and supporting tools. 
RUN rm -f /etc/slackpkg/templates/compilertoolchain.template
RUN for PKG in \ 
ca-certificates \ 
curl \ 
cyrus-sasl \ 
gc \ 
gcc \ 
git \ 
glibc \ 
glibc-profile \ 
glibc-zoneinfo \ 
guile \ 
intltool \ 
kernel-headers \ 
libmpc \ 
libffi \ 
libtasn1 \ 
make \ 
mpfr \ 
nettle \ 
p11-kit \ 
perl \ 
; do echo $PKG >> /etc/slackpkg/templates/compilertoolchain.template ; done 
RUN slackpkg -batch=on -default_answer=y update gpg 
RUN slackpkg -batch=on -default_answer=y update 
RUN slackpkg -batch=on -default_answer=y install-template compilertoolchain 
# Refresh SSL certificates: 
RUN /usr/sbin/update-ca-certificates -f 
# Create the user to switch to: 
RUN useradd -m -u "${SL_UID}" -g "${SL_GID}" -G wheel "${SL_USER}" && \ 
sed -ri 's/^# (%wheel.*NOPASSWD.*)$/\1/' /etc/sudoers 
USER "${SL_USER}" 
ENV HOME /home/"${SL_USER}" 
WORKDIR /home/"${SL_USER}" 
# Start a bash shell if the container user does not provide a command: 
CMD bash -l
EOT

The output of this ‘docker build’ command shows that the process starts with downloading (pulling) the slackware base image:

Sending build context to Docker daemon  3.072kB 
Step 1/16 : FROM liveslak/slackware:base_x64_14.2 
base_x64_14.2: Pulling from liveslak/slackware 
Digest: sha256:352219d8d91416519e2425a13938f94600b50cc9334fc45d56caa62f7a193748 
Status: Downloaded newer image for liveslak/slackware:base_x64_14.2 
---> 3a9e2b677e58 
Step 2/16 : MAINTAINER Eric Hameleers <alien@slackware.com> 
---> Running in d7e0e17f68e6 
Removing intermediate container d7e0e17f68e6 
---> 75b0ae363daf
...(lot of output snipped) ...
Successfully built e296afb023af
Successfully tagged slacktest:latest
$ docker images |grep slacktest
slacktest                latest          e296afb023af   2 minutes ago   387MB

Here you see how a Dockerfile uses ‘FROM’ to download a pre-existing image from the Hub and then RUNs several commands which it applies to the base image which was downloaded from Docker Hub, creating a new image and then naming it “slacktest” with a tag “latest” and an ID of “e296afb023af“. In that process, several intermediate layers are created, basically every command in the Dockerfile which you would normally execute on the command-line will create one. The end result is an image which consists of multiple layers. You can use the ‘history’ command so see how these were built:

$ docker history e296afb023af 
IMAGE          CREATED          CREATED BY                                      SIZE      COMMENT 
e296afb023af   49 minutes ago   /bin/sh -c #(nop)  CMD ["/bin/sh" "-c" "bash…   0B 
e1224d15abac   49 minutes ago   /bin/sh -c #(nop) WORKDIR /home/alien           0B 
4ce31fa86cca   49 minutes ago   /bin/sh -c #(nop)  ENV HOME=/home/alien         0B 
365bffd67f07   49 minutes ago   /bin/sh -c #(nop)  USER alien                   0B 
05d08fe94b47   49 minutes ago   |3 SL_GID=users SL_UID=1000 SL_USER=alien /bi…  336kB 
3a300b93311f   49 minutes ago   |3 SL_GID=users SL_UID=1000 SL_USER=alien /bi…  218kB 
7a14f55146dd   50 minutes ago   |3 SL_GID=users SL_UID=1000 SL_USER=alien /bi…  233MB 
a556359ec6d3   51 minutes ago   |3 SL_GID=users SL_UID=1000 SL_USER=alien /bi…  6.21MB 
da19c35bed72   52 minutes ago   |3 SL_GID=users SL_UID=1000 SL_USER=alien /bi…  2.13kB 
604fece72366   52 minutes ago   |3 SL_GID=users SL_UID=1000 SL_USER=alien /bi…  161B 
1c72e034ef06   52 minutes ago   |3 SL_GID=users SL_UID=1000 SL_USER=alien /bi…  0B 
3f020d0c2024   52 minutes ago   /bin/sh -c #(nop)  ARG SL_USER=alien            0B 
91a09ad8693a   52 minutes ago   /bin/sh -c #(nop)  ARG SL_GID=users             0B 
96d433e0162d   52 minutes ago   /bin/sh -c #(nop)  ARG SL_UID=1000              0B 
75b0ae363daf   52 minutes ago   /bin/sh -c #(nop)  MAINTAINER Eric Hameleers…   0B 
3a9e2b677e58   11 days ago                                                      148MB     Imported from -

That last line shows the image “3a9e2b677e58” which is the slackware ‘base‘ image which I downloaded from Docker Hub already 11 days ago.

If we now use ‘docker run’  to create a container based on this image we would expect it to give us a bash prompt as user ‘alien’. Let’s try it and run an interactive container connecting it to a pseudo terminal using the “-ti” switch:

$ docker run -ti slacktest 
alien@88081b465a98:~$ id 
uid=1000(alien) gid=100(users) groups=100(users),10(wheel) 
alien@88081b465a98:~$ exit
logout
$ docker ps -a | grep slacktest
88081b465a98   slacktest                   "/bin/sh -c 'bash -l'"   2 minutes ago   Exited (0) 16 seconds ago
$ docker rm 88081b465a98 
88081b465a98
$

Indeed we ended up as user “alien” in a machine whose hostname is the ID of the container (88081b465a98). And with the final command, we could remove the container without first stopping it, because it was already stopped when we issued the “exit” command.

Help on commands

Working with Docker containers, it is good practice that you get familiar with the command-line client (aptly name ‘docker’). You’ll find a lot of command-line examples in this article which should really become part of your hands-on toolkit.
The ‘docker’ tool has some 30 sub-commands and all of them have built-in help. To get the full list of available  docker commands run:

$ docker --help

The help per command can be found by suffixing it with “–help”. Many commands have sub-commands and these will also show help:

$ docker system --help
$ docker system prune --help

Obtain information about images and containers

The ‘docker’ command-line utility can give you a lot of information about your docker infrastructure. I will discuss a graphical management tool a bit further down, but here are some useful commands:

Get a listing of all images on your Docker host:

$ docker images

Get a listing of all running containers:

$ docker ps

Also list those containers that are in a stopped state:

$ docker ps -a

Show diskspace used by all your images, containers, data volumes and caches:

$ docker system df

Show the logs of a container:

$ docker logs CONTAINER_ID

Get the ID of a container or an image:

$ docker ps -aqf "name=containername"

The “containername” is interpreted as a regular expression. If you omit the “q” switch you will see the full info about the container, instead of only the container ID. Find an image ID using the following command (if you use a containername instead of an imagename you have a second way to get a container ID):

$ docker inspect --format="{{.Id}}" imagename

Note that in this case, “imagename” needs to be the exact string of the image (or container) as shown in the ‘docker ps‘ output, not a regular expression or part of the name.

Get the filesystem location of a data volume:

Data volumes are managed by Docker, on a Linux host the actual physical location of a Docker volume is somewhere below “/var/lib/docker/volumes/“. but on a Windows Docker host that will be something entirely different.
Using the ‘docker inspect‘ command you can find the actual filesystem location of a data volume by evaluating the “Mountpoint” value in the output. Example from my local server:

$ docker inspect volume portainer_data
[
  {
    "CreatedAt": "2021-08-01T12:44:24+02:00",
    "Driver": "local",
    "Labels": {},
    "Mountpoint": "/var/lib/docker/volumes/portainer_data/_data",
    "Name": "portainer_data",
    "Options": {},
    "Scope": "local"
  }
]

You could use this information to access data inside that volume directly, without using Docker tools.

More:

There’s a lot more but I leave it up to you to find other commands that are informational for you.

Stop and remove a container

Note that when you stop and remove a container, you do not delete the image(s) on which the container is based! So what is the usefulness of removing containers? Simple: sometimes a container crashes or gets corrupted. In such a case you simply stop the container, remove it and start it again using the exact same command which was used earlier to start it. The container will be rebuilt from scratch from the still existing images on your local file-system and the service it provides will start normally again.
As an example, create and run the ‘hello-world‘ container, then stop and remove it again:

$ docker run hello-world
Hello from Docker! 
This message shows that your installation appears to be working correctly.
...
$ docker ps -a |grep hello-world
0369206ddf2d   hello-world                 "/hello"  ...
$ CONTAINER_ID=$(docker ps -a |grep hello-world | awk '{print $1}')
$ docker stop $CONTAINER_ID
$ docker rm $CONTAINER_ID

Refresh a container with the latest image from the vendor

Suppose you are running a Docker container for some complex piece of software. You can’t simply run “slackpkg upgrade-all” to upgrade that software to its latest version or to fix some security vulnerability. Instead, what you do with a container is to refresh the Docker image which underpins that container, then stop the container, remove the container layer, remove the old vendor image, and then start the container again. On startup  of the container, it will now be based on the latest vendor image.
Configuration of a container happens either inside of data volumes that you maintain outside of the container and/or on the startup command-line of that container. Deleting the container or the image layers is not touching any of your configuration or your data.
This is the strong point of container life cycle management – it’s so easy to automate and scale up.

As an example, let’s refresh a container that I have running here at home (the Portainer container that’s discussed further down).

Download the latest Portainer Community Edition image from Docker Hub:

$ docker pull portainer/portainer-ce

Find the ID (first comumn) of the portainer container that we currently have running:

$ docker ps -af "name=portainer"

Stop and remove the container:

$ docker stop CONTAINER_ID
$ docker rm CONTAINER_ID

Find the ID of the portainer image which is oldest (look in the CREATED column) and remove it:

$ docker images | grep portainer
$ docker image rm OLDIMAGE_ID

Check the ID of the remaining image:

$ docker images | grep portainer

Start the container using the same command you would have used before to start it:

$ docker run -d --restart unless-stopped -p 127.0.0.1:9000:9000 --name portainer -v /var/run/docker.sock:/var/run/docker.sock -v /usr/share/docker/data/portainer:/data portainer/portainer-ce

And now we are running the same Portainer configuration, but with the latest available software version. A matter of seconds if you would script this.

Managing  containers, images, volumes, networks etc

As time passes and you experiment with containers, you will inevitably end up with stuff on your host computer that is no longer being used. Think of images which you wanted to try out with ‘docker run‘ or downloaded using ‘docker pull‘. Or cached data resulting from image building. The ‘docker system df‘ command I showed you earlier will give you a scare perhaps if you run it… see here the output when I run it on my server:

TYPE TOTAL ACTIVE SIZE RECLAIMABLE
Images 11 10 3.424GB 390.3MB (11%)
Containers 10 3 531.8MB 32.01MB (6%)
Local Volumes 3 1 0B 0B
Build Cache 0 0 0B 0B

The “reclaimable” percentage is data which is no longer being used by any container that Docker knows about. You can safely delete that data. The following command will do that for you:

$ docker system prune

If you want to make sure that every un-used scrap of image data is deleted, not just the dangling symlinked images, run:

$ docker system prune -a

And if you want to remove un-used data volumes, you would run:

$ docker system prune --volumes

. . . of course you execute this command only after you have verified that the affected volumes do not contain data you still need.

Difference between load/save and import/export

Docker manages images, containers, data volumes etc in an OS independent way. Meaning you are able to transport data volumes, images and containers from one host to another, or even from one type of OS (Linux) to another (MS Windows) and preserve compatibility.

Import and export commands – used for migration of container filesystems

Docker command ‘docker export‘ allows you to create a (compressed) tar archive containing the bare file system of a container:

$ docker export my_container | xz > mycontainer.tar.xz

This filesystem tarball can again be imported into a new ‘filesystem image‘. Also look at my example of creating a Slackware base image using ‘docker import‘.

$ xz -cd mycontainer.tar.xz | docker import - my_newcontainer

The export and import commands don’t preserve the container’s configuration and underlying image information, you are dealing with just the filesystem. This makes these commands quite suited for creating Docker base images which you can then base your future containers and images on via the “FROM” keyword in a Dockerfile.
For actual Docker migrations to another host, the load and save commands are better suited. Read on:

Load and save commands – used for host migration of images

Docker command ‘docker save‘  allows you to create a (compressed) tar archive of a Docker image, containing all its layers, tags, history and configuration. The only thing that is not added to this tarball are the data volumes.

If you want to migrate a container, you first save (commit) its changes as a new Docker image using the ‘docker commit’ command.

$ docker commit CONTAINER_ID my_important_image

… and then save that image in a OS-agnostic way into a tarball. The ‘save’ command does not compress the data so we use ‘xz’ for that:

$ docker save my_important_image | xz > my_important_image.tar.xz

This compressed tarball can be copied to another host (potentially with a different OS running on the host) and ‘docker load‘ allows you to load this image (even when compressed) with all its tag information and history into the new host’s Docker environment.

$ cat my_important_image.tar.xz | docker load

You can then use the ‘docker run‘ command as usual to start a new container based off that image.

And what to do with the data volumes? Docker manages its volumes in an OS agnostic way which means we can probably not just create a tarball from the volume directory on the host, since that may not be compatible with how a data volume is implemented on another host. If we use Docker commands to save a data volume and load it on another host, we should be covered. The Docker documentation on volumes has an example on how to migrate data volumes.

Saving a data volume to a tarball:

Suppose your ‘portainer‘ container uses a data volume called ‘portainer_data‘ which is mounted on the container’s “/data” directory, containing important data which you want to migrate to another host.
To generate a tarball of the data in this volume, you create a one-shot container based on some base image (I’ll use ‘alpine’) that does not do anything in itself, and give that container the command to create the backup of your ‘portainer_data‘ volume. The steps are:

  • Launch a new container and mount the volume from the ‘portainer‘ container
  • Mount a local host directory as /backup
  • Pass a command that tars the contents of the ‘portainer_data‘ volume to a backup.tar file inside our /backup directory.
  • When the command completes and the container stops, we are left with a backup of our ‘portainer_data‘ volume; the file “backup.tar” in our current directory.
$ docker run --rm --volumes-from portainer -v $(pwd):/backup alpine tar cvf /backup/backup.tar /data

… and on a new host, load this tarball into the data volume of the ‘portainer’ container which you have already setup there but it’s still missing its data:

$ docker run --rm --volumes-from portainer -v $(pwd):/backup alpine bash -c "cd /data && tar xvf /backup/backup.tar --strip 1"

Voilà!

Limiting the log size for your containers

Docker captures the standard & error output (stdout/stderr) for all your containers and writes these to logfiles in JSON format. By default, Docker does not impose size restrictions on its log files.
These logfiles are shown when you run a ‘docker logs CONTAINER_ID‘ command.
This behavior could potentially fill up your hard drive if you run a container which generates an extensive amount of logging data. By modifying Docker’s “/etc/docker/daemon.json” configuration file we can limit the maximum size of the container log files. This is what you need to add to the Docker daemon configuration to limit the size of any log file to 50 MB and keep at most two backups after log rotation:

{
  "log-driver": "json-file",
    "log-opts": {
    "max-size": "50m",
    "max-file": "3"
    }
}

More details to be found at https://docs.docker.com/config/containers/logging/json-file/.

Graphical management tool

There’s a nice graphical web-based administration tool for your Docker infrastructure which itself runs in a container (you can also install it on bare metal as a self-contained executable), called portainer. Portainer will listen at a TCP port of your host and you can point your web browser at http://localhost:9000/ to access the GUI. The Community Edition (portainer-ce) is open source and free and this is what we’ll install and run as a Docker container.
The Portainer developers posted a tutorial and feature walk-through of their product on Youtube:

Some considerations before we start implementing this.

Data persistence:
Portainer wants to keep some data persistent across restarts. When it is run from within a container, you want to have that persistent data outside of your container, somewhere on the local filesystem. The Docker concept for persistent data management is called “volumes”. Docker volumes are created as directories within “/var/lib/docker/volumes/” on Linux. A volume is made available in the container using the “-v” switch to the ‘docker run‘ command, which maps a host volume to an internal container directory. Docker manages its volumes in an OS agnostic way, making it possible to transparently move your containers from Linux to Windows for instance.
We will not use a Docker volume in this example but instead create the “/usr/share/docker/data/portainer” directory and pass that as the external location for the container’s internal “/data” directory.

Network security:
We will expose the Portainer network listen port to only localhost and then use an Apache reverse proxy to let clients connect. Note that port 9000 exposes the Portainer UI without SSL encryption; we will use Apache reverse proxy to handle the encryption.

Application security:
Below you will see that I am mapping the Docker management socket “/var/run/docker.sock” into the container. Normally that would be very bad practice, but Portainer is going to be our Docker management tool and therefore it needs the access to this socket.

This batch of commands will get your Portainer-CE instance up and running and it will be restarted when Docker starts up (eg. after rebooting).

# mkdir -p /usr/share/docker/data/portainer
# docker run -d --restart unless-stopped -p 127.0.0.1:9000:9000 --name portainer -v /var/run/docker.sock:/var/run/docker.sock -v /usr/share/docker/data/portainer:/data portainer/portainer-ce

Two of the three network ports that Portainer opens up are not being exposed (mapped) to the outside world:
Port 8000 exposes a SSH tunnel server for Portainer Agents, but we will not use this. Portainer Agents are installed on Docker Swarm Clusters to allow central management of your Swarm from a single Portainer Server.
Also, the Portainer container generates a self-signed SSL certificate on first startup and uses it to offer https encrypted connections on port 9443 but our reverse proxy solution allows the use of a proper Let’s Encrypt certificate.

Now that the Portainer container is running, connect a browser on your host computer to http://localhost:9000/ to perform initial setup; basically defining the password for the Portainer admin user account.

The reverse proxy configuration including adding a Let’s Encrypt certificate is a bit beyond the scope of this article, but you can read this earlier blog post to learn about securing your webserver with SSL certificates.

I will share the block that you would have to add to your Apache configuration somehow. With this configuration and proper SSL setup your Portainer would be accessible via https://your_host_name/portainer/ instead of only at http://localhost:9000/ :

# Create a reverse proxy:
<Location /portainer>
    ProxyPass http://127.0.0.1:9000
    ProxyPassReverse http://127.0.0.1:9000
</Location>

Here is a screenshot of my Docker infrastructure at home as interpreted by Portainer:

Have a look at Portainer to see if it adds something useful to your Docker management toolkit. For me the most useful feature is the real-time container logging display, which is a lot more convenient than trying to scroll back in a Linux screen session.

Docker’s effect on iptables firewall

Docker dynamically modifies your host’s iptables ruleset on Linux when it starts, in order to provide container isolation. See https://docs.docker.com/network/iptables/ for the rationale. This can sometimes interfere with pre-existing services, most prominently the Virtual Machines that you may also be running and whose virtual network has been bridged to your host’s network interface.
I use vde (virtual distributed ethernet) for instance to provide transparent bridged network connectivity for the QEMU VM’s that I use when building packages for Slackware.
With this bridged network setup, your host computer essentially acts as a router, forwarding network packets across the bridge.
When Docker starts, it will add two custom iptables chains named ‘DOCKER’ and ‘DOCKER-USER’, adds its own custom rules to the ‘DOCKER’ chain and then sets the default FORWARD policy to DROP. Thereby killing the network connectivity for the already bridged networks… Bad. But even though Docker has a configurable parameter to disable iptables, this only prevents the creation of these custom chains and rules when the Docker daemon starts. Docker confesses that there is no guarantee that later on when you start containers, your iptables rulesets will stay un-touched. So, disabling iptables management in Docker is not an option. I’ll tell you a bit on how to deal with this (all information gleaned from the Internet).

Generic iptables

Since Docker sets the policy for the FORWARD chain to DROP, for your bridged networks this will result loss of connectivity (packets are no longer forwarded across the bridge).
You can use the custom DOCKER-USER chain to ensure that this bridging is not disrupted. Docker inserts these custom chains in such a way that rules defined in the DOCKER-USER chain have precedence over rules in the DOCKER chain. You have to add an explicit ACCEPT rule to that DOCKER-USER chain:

# iptables -I DOCKER-USER -i br0 -j ACCEPT

Where ‘br0‘ is the name of your network bridge (adapt if needed).

UFW (Uncomplicated Firewall)

UFW is a tool created for Ubuntu but popular on all distros, which aims at simplifying iptables ruleset management by providing a higher-level commandline interface. There’s a package build script for it on SlackBuilds.org.

UFW allows to insert rules that are evaluated  after all other chains have been traversed. This allows the user to nullify the unwanted behavior of Docker which kills bridged networks. See this github discussion for the full details.

Append the following at the end of /etc/ufw/after.rules (replace eth0 with your external facing interface):

# Put Docker behind UFW
*filter
:DOCKER-USER - [0:0]
:ufw-user-input - [0:0]

-A DOCKER-USER -m conntrack --ctstate RELATED,ESTABLISHED -j ACCEPT
-A DOCKER-USER -m conntrack --ctstate INVALID -j DROP
-A DOCKER-USER -i eth0 -j ufw-user-input
-A DOCKER-USER -i eth0 -j DROP
COMMIT

And undo any and all of:

  • Remove “iptables”: “false” from /etc/docker/daemon.json
  • Revert to DEFAULT_FORWARD_POLICY=”DROP” in /etc/default/ufw
  • Remove any docker related changes to /etc/ufw/before.rules

Firewalld

Firewalld is a dynamically managed firewall which works with a ‘zones‘ concept and gets its triggers from a D-Bus interface. There’s a package build script available on SlackBuilds.org.

If you are running Docker version 20.10.0 or higher with firewalld on your system with and Docker has iptables support enabled, Docker automatically creates a firewalld zone called docker and inserts all the network interfaces it creates (for example, docker0) into the docker zone to allow seamless networking.

Consider running the following firewalld command to remove the docker interface from the ‘trusted‘ zone.

# Please substitute the appropriate zone and docker interface
$ firewall-cmd --zone=trusted --remove-interface=docker0 --permanent
$ firewall-cmd --reload

Restarting the docker daemon (/etc/rc.d/rc.docker restart) will then insert the interface into the docker zone.

Alien’s Easy Firewall Generator (EFG)

If you used my EFG to create a firewall configuration, you will be bitten by the fact that Docker and Docker Compose create new network devices . Permissive rules need to be be added to the INPUT and OUTPUT chains of your “rc.firewall” or else ‘dmesg’ will be full of ‘OUTPUT: packet died’.

The Docker network interface is called “docker0” and for each container, Docker creates an additional bridge interface, the name of which starts with “br-” and is followed by a hash value. Therefore, these lines need to be applied at the appropriate locations in the script:

DOCK_IFACE="docker0"
BR_IFACE="br-+"
$IPT -A INPUT -p ALL -i $DOCK_IFACE -j ACCEPT
$IPT -A INPUT -p ALL -i $BR_IFACE -j ACCEPT
$IPT -A OUTPUT -p ALL -o $DOCK_IFACE -j ACCEPT
$IPT -A OUTPUT -p ALL -o $BR_IFACE -j ACCEPT

Thanks

… for reading until the end! I hope you gained some knowledge. Feel free to leave constructive feedback in the comments section below.

Cheers, Eric


Attribution

Most of the firewall-related information comes from docker.com and github.com. Please inform me about errors.
All images in this article were taken from docs.docker.com website.

My Docker packages for Slackware-current

I have been using Docker for a while now, it’s being used to provide services to friends and family.
I was always intimidated by the large amount of packages that were needed to get Docker and docker-compose up and running, and I did not have experience with Docker at the time (almost two years ago) so I decided to go the easy route and use the SlackBuilds.org scripts when I first needed to run a Docker container. I wrote a blog about that even, it explained how to run an Outline server to allow journalists to do their work in repressive countries but the article also shares the details how to build the Docker packages and run the daemon.

If you want to read some background information about Docker’s strength and what its use-cases are, I encourage you to start reading here: https://docs.docker.com/get-started/overview/ .
Essentially, Docker uses Linux kernel and filesystem capabilities to isolate an application and its dependencies from the host computer it is being executed on. Docker provides powerful means to connect multiple containers via internal (virtual) networking and can expose ports to the network outside of your container. It enables you to run applications reliably without having to worry about the underlying Operating system. You can even run Docker on a MS Windows computer but your containerized application running inside Docker will not be aware of that.
This is sometimes called ‘light-weight virtualization’ because unlike real virtualization solutions like QEMU, Virtual Box or VMWare, the containerized application still runs on your host’s kernel. This is why you can run a 32-bit container image on a 64-bit (Linux 64-bit kernel has that capability to  execute 32-bit binaries) host but you cannot run  a 64-bit image on a 32-bit host kernel.

Now that I am more familiar with Docker, have been running multiple services in containers for more than a year, and have created and published my own images (more about that later) I decided to create my own set of Docker packages. Having pre-built packages will make it a lot easier for people to start exploring the usefulness of Docker containers.

One thing upfront: I have significantly decreased the total amount of packages you need to run Docker.
I have combined the SlackBuilds.org packages ‘docker’, ‘docker-cli’, ‘docker-proxy’ and ‘tini’ into a single package called ‘docker’ and also added ‘docker-buildx’ to that docker package. Also, the re-write of docker-compose from Python to Go has the benefit that the run-time package dependencies for ‘docker-compose’ have been reduced from thirteen to zero.

Starting with my Docker packages

As stated in the subject: the packages I created are for Slackware-current only. If you want to compile this yourself on Slackware 14.2, I cannot guarantee success since I did not try compiling them there myself – but in any case you’ll have to build and install libseccomp from SlackBuilds.org, this is part of -current but not 14.2.

What you need from my repository to run Docker is: runc, containerd, docker and docker-compose. Four packages – that’s it.
If you want to be able to (re-)compile these packages, you will additionally need google-go-lang. After installing google-go-lang you need to logoff and login again (or run the command “source /etc/profile.d/go.sh” in your terminal) to give Google’s version of Go preference over the GCC version of Go that’s probably already installed on your computer.

The ‘docker’ package installation script will add a couple of lines to “/etc/rc.d/rc.local” and “/etc/rc.d/rc.local_shutdown” to make Docker start on boot and properly stop during shutdown of the computer. The docker rc script “/etc/rc.d/rc.docker” will initially be installed without the execute bit, so if you actually want to start using Docker you have to make the script executable. This is a one-time action. Future package upgrades will honor the executable status of that script.

# chmod +x /etc/rc.d/rc.docker

You can start the Docker daemon now if you don’t want to waste time with a reboot:

# /etc/rc.d/rc.docker start

The package installation will also trigger the creation of a new group called ‘docker’. If you want to be able to run and manage your Docker images and containers as your own non-root user account, you need to add your user account to this ‘docker’ group and then logoff/login again, and restart the docker deamon. Otherwise, all Docker operations can only be executed by the root user.

# gpasswd -a <your_useraccount> docker
# /etc/rc.d/rc.docker restart

After doing all the prep work, your account added to the ‘docker’ group and the daemon running, it’s time for a first test. Run the following command:

$ docker run hello-world

You’ll see the following output:

Unable to find image 'hello-world:latest' locally 
latest: Pulling from library/hello-world 
2db29710123e: Pull complete 
Digest: sha256:2498fce14358aa50ead0cc6c19990fc6ff866ce72aeb5546e1d59caac3d0d60f 
Status: Downloaded newer image for hello-world:latest 

Hello from Docker! 
This message shows that your installation appears to be working correctly
...

What you will have validated here is the proper functioning of your Docker installation:

  1. The Docker command-line client contacted the Docker daemon.
  2. The Docker daemon downloaded (or ‘pulled‘) the “hello-world” image from the Docker Hub.
  3. The Docker daemon created a new container from that image which runs the executable that produces the output you could read on your terminal just now.
  4. The Docker daemon streamed that output to the Docker client, which sent it to your terminal.

The “hello-world” image is of course trivial, but there are many more with real-life use-cases which you can find on  Docker Hub, https://hub.docker.com/. You can download these images freely but if you want to upload (or ‘push’) an image which you created yourself, you’ll have to create an account on Docker Hub. Free acccounts allow the creation of one single private repository and unlimited public repositories.

Slackware Docker images

I have a repository on Docker Hub where I share my base images for stable Slackware versions (that’s the 14.2 release right now). Go get them at https://hub.docker.com/r/liveslak/slackware/tags . These “base images” are roughly 55 MB compressed, which means they are really basic. They are created using a script (create_slackware_docker_image.sh) which takes its inspiration both from the liveslak scripts and of Vincent Batts‘ work on Docker images for Slackware. Essentially, the script installs some packages into a package root, uses tar to put all of that in a tarball and loads the tarfile into Docker which will then make it into an actual image for you. That image wille be called “slackware:base_x64_14.2” (unless you specified a different architecture and release of course) and that’s what I uploaded to the Docker Hub.
My tiny Slackware docker-images are not really meant to be used as-is. Instead, they can act as the foundation for Slackware-based Docker images that you might want to build yourself – any Docker image starts with some existing base image and adds new layers on top. Base images like mine above don’t inherit from a lower-level image and are a special case, read more here: https://docs.docker.com/develop/develop-images/baseimages/

Now let’s pull that Slackware base image from the Hub and peek inside!

$ docker run -ti liveslak/slackware:latest /bin/bash -l

The “-ti” parameters tell docker to assign a pseudo-tty and run an interactive session on your terminal.
The convention for the image name is “username/imagename:tag” which shows that “liveslak” is the user who hosts the image on Docker Hub; “slackware” is the name of the image and it has a tag “latest” which means: just give me the latest version that I can get. For the slackware image, this means you get the 64-bit variant (I also have a 32-bit image) of Slackware 14.2.
The “/bin/bash -l” at the end is the command which Docker should run after bringing the container online. Remember, the base image contains nothing but a small amount of installed Slackware packages and does not start any application by itself. More complex Docker images may run all kinds of applications, some in the background and some meant to be interacted with such as the bash shell.

Running the above command yields this result:

Unable to find image 'liveslak/slackware:latest' locally 
latest: Pulling from liveslak/slackware 
6c721e5d29bd: Pull complete 
Digest: sha256:352219d8d91416519e2425a13938f94600b50cc9334fc45d56caa62f7a193748 
Status: Downloaded newer image for liveslak/slackware:latest 
root@b0264b9e59ff:/#

And we end up at the command prompt of our running container. The container user is ‘root’, it’s the only user in that base image. You do not have to enter a password.
Let’s play a bit:

root@b0264b9e59ff:/# ls /var/log/packages/ | wc -l
77
root@b0264b9e59ff:/# cat /etc/os-release 
NAME=Slackware 
VERSION="14.2" 
ID=slackware 
VERSION_ID=14.2 
PRETTY_NAME="Slackware 14.2" 
ANSI_COLOR="0;34" 
CPE_NAME="cpe:/o:slackware:slackware_linux:14.2" 
HOME_URL="http://slackware.com/" 
SUPPORT_URL="http://www.linuxquestions.org/questions/slackware-14/" 
BUG_REPORT_URL="http://www.linuxquestions.org/questions/slackware-14/"
root@b0264b9e59ff:/# exit
logout
$

Suppose you want want to download and use the 32-bit Slackware 14.2 base image instead. Then you would run:

$ docker run -ti liveslak/slackware:base_ia32_14.2 /bin/bash -l

Here ends my very brief introduction to Docker on Slackware. Let me know what you think of this! Is there anything you would like to see explained in more detail?
Eric

Update 2022-jan-13: I have added packages for Slackware 14.2 (32bit and 64bit) to my repository.

LibreOffice 7.2.5 packages for Slackware-current

LibreOffice Community Edition 7.2.5 was released yesterday and I have uploaded a new set packages for Slackware-current.

This is the fifth iteration in the 7.2 release cycle with two more to come in the next three months. Since this is a minor upgrade, the focus is on bug fixing and improving the stability.

The new packages can be found in my repository (https://slackware.nl/people/alien/slackbuilds/libreoffice/) or any mirror if you wait a day, for instance https://slackware.uk/people/alien/slackbuilds/libreoffice/ .


Enjoy the new release – Eric

Chromium 97 added to my repository; fixes a critical vulnerability

A couple of days ago Google released the sources for Chromium 97.0.4692.71. I am still waiting for an update to the chromium-ungoogled sources but I have already uploaded Slackware packages for chromium (targeting 14.2 and -current) to my repository.

This release addresses one ‘critical’ vulnerability (CVE-2022-0096, ‘Use after free in Storage‘) and brings a total of 37 security fixes. From a user perspective, I could not find particularly exciting new functionality in the release notes. The security fixes still make this a recommended upgrade.

You can get my chromium packages from slackware.nl or its mirrors.

Eric

Update 08-jan-2022: packages for chromium-ungoogled 97.0.4692.71 are now also available in my repository.

Repository purge coming up soon-ish

After I built a fresh Avidemux (see previous post) I realized how many old packages I still have in my repositories. They are taking up space on many server mirrors.

I have decided that I will start a cleansing process, a purge if you want, of all the older stuff. The reason is not just disk space of course. It’s my realization that there may be vulnerabilities in these old packages that I never addressed; and I really hope that people have migrated their machines to Slackware 14.2 (servers or conservative desktop users) or went with -current (modern desktop users, let’s call those).

From time to time, you need to clean house. I myself am infamous for not throwing away anything… just take a look at my attic. So these packages will be gone from online servers, but live on in my own local package archive.

Because of the time involved (manual checks to ensure that I don’t remove too much by mistake) this will be done in several ‘waves’. My repositories already don’t have packages for Slackware releases older than 13.37 and tomorrow I will purge all packages targeting Slackware 13.37. Later, that will be followed by a purge of all packages for Slackware 14.0, and finally those that were meant for Slackware 14.1 will also be removed. I will stick with package support for Slackware 14.2 (and 15.0 when that gets released) and always -current.

In case you still run one of these older versions of Slackware, you would want to make local copies of the following repositories ASAP:

And also these:

I will leave the multilib repositories (http://www.slackware.com/~alien/multilib/ and its mirrors) untouched.

Eric