Dockerization of physical Fedora 31 servers

Instructions to convert an existing physical Linux Fedora server to a docker container. Why? Useful to test an existing rsync backup and to actually replace a physical server in case of an emergency. This procedure enables systemd suport inside the docker image, so that it can run multiple processes inside the container.

Step 1: Full backup of existing server

Plug and mount external storage to the physical server with enough capacity to hold all non-free space of the server.

Run the following command as root to create a full copy of the filesystem:

rsync -aAXv --info=progress2 --delete / --exclude={"/dev/*","/proc/*","/sys/*","/run/*","/mnt/*","/media/*","/lost+found"} /mnt/external-backup

Where /mnt/external-backup is where the external disk is mounted.

The advantages of using rsync to do a full backup are plenty:

• rsync preserves file permissions and symlinks
• rsync allows incremental backups
• rsync will just backup the used space instead of the full partition size like other methods of image creation like dd
• There is no need that the system where the backup is restored have the same partition layout

Step 2: Import as a Docker base image

Plug and mount the external drive with the backup to another computer. Export a variable containing the mount point:

export MOUNT_DIR=/path/to/backup

The command below will import the backup files as a docker base image:

tar --exclude='home/data/*' --exclude='*/chainstate' -cC $MOUNT_DIR . | docker import - base_image 

The several 'exclude' params should be used in case there is not enough space in the destination computer to run the full backup. In this case, it is possible to create a base image comprising of only the base OS files and leave aside large directories that will be mounted as volumes later.

Step 3: Create the image

The Dockerfile can be found at:

git clone https://github.com/gustavonalle/physical2docker.git && cd physical2docker

Check Docker version

Make sure to use a recent Docker version. Fedora 28 ships with an older version that has issues with privileged and systemd enabled containers. Follow https://docs.docker.com/install/linux/docker-ce/fedora/ to install the latest stable version.

Check Docker disk space

Before proceeding, check if docker has enough space to hold the image. It usually store images in /var/lib/docker/ which usually resides in the / partition that will probably be small. If not enough space is available:

• Stop the docker service
• Move the contents of /var/lib/docker/ to another partition
• Create a symlink to the new location: ln -s /path/to/new/docker/folder /var/lib/docker/
• Start docker

(Optional) Adjust ports

The Dockerfile EXPOSE directive should be tailored for specific needs.

Build the image

docker build -t image_name  .

Final step

Before running the container, disable SELinux. Make sure SELINUX=disabled is in /etc/selinux/config, and reboot.

Note: It is possible to run a systemd enabled container with Selinux enabled; I prefer not to waste my time tinkering with it though.

Run the container:

docker run -p 9130:9130 -p 8332:8332 --name container-name --privileged --tmpfs /run -v /sys/fs/cgroup:/sys/fs/cgroup:ro -v /path/to/large_dir:/path/inside/container -td image_name

The last part -v /path/to/large_dir:/path/inside/container is optional, it's used when some of the backup files were ignored at build time, to be mounted later as a volume. Also, don't forget to expose the ports in case needes with the several '-p' arguments illustrated above.

The container will run on the background, with support for systemd.

To attach to it, run:

docker exec -it container-name zsh

Presto! A container with an exact copy of the physical server. Although this procedure is for Fedora 31, it should work fine for other distros, provided the Dockerfile extends the same base image as before.

If this procedure help you, consider donating any amount to bc1qf5ndupggeqtpk5gf4leh4vcxecxakw067hel0z as a token of appreciation!