Table of Contents
Introduction to Docker and Containers
Docker defines a container as a "standard unit of software". Container images package an application and all of its dependencies enabling it to run quickly and reliably across platforms. A container is isolated from processes and other containers running on the same platform unless interfaces are explicitly defined. Docker containers provide a standard, lightweight and secure virtualization solution when a full hypervisor is overkill. Unlike hypervisors that virtualize the hardware, containers are lightweight because only the OS is virtualized. Containers rely on kernel features such as namespaces, cgroups and unionfs.
Docker provides:
- Docker Engine for container runtime
- Docker tools for creating, deploying and managing containers
- Docker Hub repository for hosting container images
This wiki assumes you have a working knowledge of Yocto. It will walk you through how to build and deploy Docker on Zynq Ultrascale+ with a Yocto flow in the following four configurations:
- systemd on a flash filesystem
- systemd on a ramdisk
- sysvinit on a flash filesystem
- sysvinit on a ramdisk
WTS Support
This is an unsupported WTS flow. Please do your due diligence when implementing Docker with Yocto and regression test against your system requirements.
Requirements
- ZCU102 Eval Board
- Linux host machine or VM
- Internet
Testing
Releases
rel-v2018.3 (Rocko)
Platforms
ZCU102
meta-virtualization Layer
The Docker recipe is included in the meta-virtualization layer. The Xilinx Yocto manifest instructs repo to automatically clone meta-virtualization, so there is no need to clone it manually unless you are using an unsupported Yocto flow. You can verify that the virtualization layer is installed in the sources directory and that it's included in the bblayers.conf. From this layer we will be building and installing the docker and docker-contrib packages. We will add this to a Docker machine configuration later.
IMAGE_INSTALL_append = " docker docker-contrib"
meta-xilinx-docker Layer
Throughout this wiki we will be populating a custom layer, meta-xilinx-docker, that reflects the listing below. Creating a Custom Xilinx Yocto Layer shows how to create a base layer using the Yocto scripts.
meta-xilinx-docker
├── conf
│ ├── distro
│ │ └── petalinux-systemd.conf
│ ├── layer.conf
│ └── machine
│ ├── docker-systemd-zcu102-zynqmp.conf
│ └── docker-zcu102-zynqmp.conf
├── COPYING.MIT
├── README
├── recipes-bsp
│ └── hdf
│ └── external-hdf.bbappend
├── recipes-containers
│ └── docker
│ └── docker_git.bbappend
└── recipes-kernel
└── linux
├── cfg
│ └── docker.cfg
└── linux-xlnx_%.bbappend
Linux Configuration
The default ZCU102 kernel configuration does not have all the required CONFIG options for Docker, so we will need to turn them on through a configuration fragment. If you don't already have a kernel recipe directory in your layer, create the directory structure as shown below in your custom layer.
recipes-kernel/
└── linux
├── cfg
│ └── docker.cfg
└── linux-xlnx_%.bbappend
Next edit the linux-xlnx_%.bbappend file as shown below and add the docker.cfg. Note the wildcard "%" is used to match any kernel version.
FILESEXTRAPATHS_prepend := "${THISDIR}/cfg:"
SRC_URI_append = " file://docker.cfg"
Add a docker.cfg file in the cfg directory and add the CONFIG requirements shown in the listing below. The script from the Moby Project was used to determine these (more on this later).
Machine Configuration
Next we will configure the machine for installation of the Docker package(s) in the image. Create the machine directory under the conf directory in your layer and add the docker-systemd-zcu102-zynqmp.conf file. Note that we have defined separate machine configuration allows you to keep both sysvinit and systemd Docker builds in your deploy directory after baking. The distro configuration will be covered in the next section.
conf/
└── machine
├── docker-systemd-zcu102-zynqmp.conf
└── docker-zcu102-zynqmp.conf
Populate the docker-systemd-zcu102-zynqmp.conf file as shown below. The docker-contrib is an optional supporting package which contains a script that we will use to verify the kernel configuration.
Next you need to set your distro policy to the new petalinux-systemd policy. Set the DISTRO variable as shown below in your local.conf or your custom <machine>.conf. We will set it in the docker-systemd-zcu102-zynqmp.conf machine configuration file.
In order to simplify the network manager on an embedded system, connman is recommended to manage the network adapter. This is completely optional. Without connman, you will need to setup the networkd service unit configuration files which is beyond the scope of this wiki. Last the loglevel was downgraded to level 6 and audit turned off from the kernel command line to reduce console chatter.
# Inherit all the properties from the zcu102-zynqmp machine
require conf/machine/zcu102-zynqmp.conf
# Reuse the fdt for the zcu102 bsp
YAML_DT_BOARD_FLAGS_pn-device-tree = "{BOARD zcu102-rev1.0}"
# Downgrade the loglevel to reduce console chatter and disable audit
KERNEL_BOOTARGS_append_pn-u-boot-zynq-uenv = " loglevel=6 audit=0"
# Include docker and docker-contrib in the image
IMAGE_INSTALL_append = " docker docker-contrib"
# Comment these out to use sysvinit
DISTRO = "petalinux-systemd"
IMAGE_INSTALL_append = " connman"
If you want to boot with an initramfs, add the variables in the listing below to your <machine>.conf or local.conf and observe the ramdisk info note below when running Docker.
Distro Configuration (systemd)
Optional
If you are using the default sysvinit, you may skip this section.
Systemd is a replacement init system for sysvinit which is used by many advanced Linux Distributions. While Docker works with sysvinit, it works best with systemd. When systemd is enabled, systemd will start the docker daemon during boot. To enable systemd in your image, you need to patch an existing distro configuration or create a new one. We will create a new one here based on the Petalinux Distro. Create the directory structure shown below in your layer and add the petalinux-systemd.conf
conf/ └── distro └── petalinux-systemd.conf
require conf/distro/petalinux.conf DISTRO = "petalinux-systemd" DISTRO_NAME = "PetaLinux with systemd" DISTRO_FEATURES_append = " systemd" # Uncomment this to remove legacy sysvinit scripts #DISTRO_FEATURES_BACKFILL_CONSIDERED += "sysvinit" VIRTUAL-RUNTIME_init_manager = "systemd" VIRTUAL-RUNTIME_initscripts = "systemd-compat-units"
Kernel Config for Systemd
Systemd has been tested with Docker using the default kernel configs including the Docker support configs on ZCU102. If you are planning on using systemd in production, please make sure that the kernel configs meets your system requirements as recommended by freedesktop.org. Systemd configuration is beyond the scope of this wiki.
HDF
Optional
If you are using a custom HDF, you may skip this section.
One last housekeeping thing is to add a bbappend to reuse the default HDF from GitHub, so we don't need to add a custom HDF (you're welcome Mr. software developer). This script just creates a link to the base machine we want to leverage, e.g. zcu102-zynqmp, in the external-hdf working directory.
Build Docker
The bitbake command below assumes you are building a docker-systemd-zcu102-zynqmp.conf machine which includes the configurations for Docker on a ZCU102. If you are setting the configurations in your local.conf, then you may target any machine such as zcu102-zynqmp.
$ MACHINE=docker-systemd-zcu102-zynqmp bitbake petalinux-image-minimal
If you want to build Docker with sysvinit, then you may create another machine docker-zcu102-zynqmp.conf and comment out the systemd lines as noted in the docker-systemd-zcu102-zynqmp.conf listing above.
$ MACHINE=docker-zcu102-zynqmp bitbake petalinux-image-minimal
Install Docker
Docker expects to run from a non-RAM based root filesystem since it uses pivot_root to jail the container. For that reason it's recommended that you setup an SD card with VFAT and Ext4 partitions. Once your SD card is partitioned, copy the boot images to the VFAT partition and extract the rootfs to the Ext4 partition. These images are available in the deploy/images directory of the machine you built. Copy the images as shown below noting that the VFAT partition is mounted on boot and the Ext4 partition is mounted on rootfs. This example is using the dtb that is built from the kernel tree. If you are using a custom HDF, you will want to use the dtb compiled from the DTG so make sure that uEnv.txt is using the correct dtb.
$ cd tmp/deploy/images/docker-systemd-zcu102-zynqmp $ cp BOOT-docker-systemd-zcu102-zynqmp.bin /media/<user>/boot/boot.bin $ cp Image /media/<user>/boot/ $ cp uEnv.txt /media/<user>/boot/ $ cp Image-zynqmp-zcu102-rev1.0.dtb /media/<user>/boot/ $ sudo tar xzf petalinux-image-minimal-docker-systemd-zcu102-zynqmp.tar.gz -C /media/<user>/rootfs/
Run Docker
Insert the SD card into you ZCU102, connect the Ethernet to an internet connected router and boot the board. To verify that the kernel is configured properly, you can run the Moby script check-config.sh. Your output should look similar to the listing below. Note that AUFS_FS and zfs are not enabled which is expected.
systemd
If you are running Docker with systemd and connman, then systemd will start dockerd and your network adapter will come up automatically during boot.
Ramdisk
If you want to run Docker with systmed on a ramdisk, then you need the bbappend below in order to set the DOCKER_RAMDISK variable during boot.
sysvinit
If you want to continue running your system with sysvinit, then you will need to manually start the dockerd in the background.
# dockerd &
Ramdisk
If you want to run Docker with sysvinit on a ramdisk, then you must export the DOCKER_RAMDISK environment variable and set it to true before launching dockerd.
# export DOCKER_RAMDISK=true; dockerd &
You can verify that the Docker daemon is running by issuing the "ps" command. Next, test that you can connect to it by issuing "docker info".
With the ZCU102 booted and connected to the internet, run the standard hello-world container image to test your embedded Docker installation. Your console should look similar to the hello-world listing. Since this is the first time you are running the container, the image must be pulled from the Docker Hub repository. Any subsequent runs will pull the image from the local repository.
root@docker-systemd-zcu102-zynqmp:~# docker run --rm hello-world
Unable to find image 'hello-world:latest' locally
WARN[0277] Could not get operating system name: Error opening /usr/lib/os-release: open /usr/lib/os-release: no such file or directory
WARN[0277] failed to retrieve docker-init version: exec: "docker-init": executable file not found in $PATH
latest: Pulling from library/hello-world
3b4173355427: Pull complete
Digest: sha256:2557e3c07ed1e38f26e389462d03ed943586f744621577a99efb77324b0fe535
Status: Downloaded newer image for hello-world:latest
[ 354.686277] cgroup: docker-runc (2485) created nested cgroup for controller "memory" which has incomplete hierarchy support. Nested cgroups may change behavior in the future.
[ 354.701848] cgroup: "memory" requires setting use_hierarchy to 1 on the root
Hello from Docker!
This message shows that your installation appears to be working correctly.
To generate this message, Docker took the following steps:
1. The Docker client contacted the Docker daemon.
2. The Docker daemon pulled the "hello-world" image from the Docker Hub.
(arm64v8)
3. The Docker daemon created a new container from that image which runs the
executable that produces the output you are currently reading.
4. The Docker daemon streamed that output to the Docker client, which sent it
to your terminal.
To try something more ambitious, you can run an Ubuntu container with:
$ docker run -it ubuntu bash
Share images, automate workflows, and more with a free Docker ID:
https://hub.docker.com/
For more examples and ideas, visit:
https://docs.docker.com/get-started/
Once you have verified that Docker is working with hello-world, you can try running more advanced containers such as ubuntu, centos, gcc, etc.
root@docker-systemd-zcu102-zynqmp:~# docker run --rm -it ubuntu bash
Now you should be in a Ubuntu bash shell. You can now install applications with apt-get. Type exit to quit the container.
Related Links
Creating a Custom Xilinx Yocto Layer
Adding an HDF to a Xilinx Yocto Layer
Customizing Device Trees in Xilinx Yocto
Xilinx Yocto Builds without an Internet Connection