Anderson, Craig (ca846m) 0064db95fe image-builder refactor
The original image-builder approach had an entirely containerized
approach for building target images.

This approach was flawed because:
1. There are a number of debian packages which will not install without
/sys, /proc, /dev, or /dev/pts mountpoints, and
2. Container build process does not support building with privileges
needed to bind-mount these directories into the chroot build space
3. It is a requirement for all packages to be installed in the container
image in order to avoid deployment risk of missing mirror resources

This patchset addresses this problem by performing necessary privileged
steps outside of a containerized build process. At the end of this
process, the root filesystem is packaged into a docker container when
elevated permissions are no longer required.

Change-Id: I5f8dc972f67c5649bf5f9403a5a512d06c948720
2021-02-17 09:21:13 -08:00

111 lines
2.8 KiB
YAML

rootfs_root: build
kernel:
modules:
load:
- name: 8021q
- name: bonding
- name: ip_vs
- name: ip_vs_rr
- name: ip_vs_wrr
- name: ip_vs_sh
- name: br_netfilter
blacklist:
- name: krbd
banners:
login: |
Airship Node \l: \n.\o
Kernel: \s \m \r \v
IP address: \4
motd: |
#!/bin/sh
. /etc/lsb-release
printf "Airship Node, based on: %s (%s %s %s)\n" "$DISTRIB_DESCRIPTION" "$(uname -o)" "$(uname -r)" "$(uname -m)"
kubelet:
# Add only image-builder appropriate kubelet args here.
# Add all others to kubeadmcontrolplane.yaml
extra_systemd_args: []
#- name: reserved-cpus
# value: '0-3'
grub:
GRUB_TIMEOUT: 5
GRUB_CMDLINE_LINUX_DEFAULT:
- name: cgroup_disable
value: 'hugetlb'
limits:
- name: core_dump
domain: '0:'
type: 'hard'
item: 'core'
value: 0
sysctl:
- name: net.bridge.bridge-nf-call-ip6tables
value: '1'
- name: net.bridge.bridge-nf-call-iptables
value: '1'
# Any directories to create on disk can be defined here
directories:
# Full path to file to create
- name: /testdir
permissions: '0755'
owner: root
group: root
# Any files to write to disk can be defined here
files:
# Full path to file to create
- name: /testdir/test.sh
file_content: |
#!/bin/bash
echo hello world
permissions: '0755'
owner: root
group: root
systemd:
# Full name, including systemd suffix. sample.service. sample.mount, sample.timer, etc.
- name: sample.service
file_content: |
[Unit]
Description=sample service
After=network.target
[Service]
ExecStart=/bin/sleep infinity
[Install]
WantedBy=multi-user.target
# whether the target image should run this service on boot
enabled: yes
# whether to override existing symlinks (e.g. name collision).
# Use only if you are intenting to overwrite an existing systemd unit
force: no
# If any custom shell scripts are needed for qcow building for image building,
# they can be added here.
buildtime_user_scripts:
- file_content: |
#!/bin/bash
echo "custom container buildtime script"
# Custom user shell scripts to be run during container execution, right before
# final QCOW image is created.
runtime_user_scripts:
- file_content: |
#!/bin/bash
echo "custom container buildtime script"
# Any other adjustments to file or directory permissions, for files that already exist.
file_permissions:
# Full path to file to create
- name: /testdir/test.sh
permissions: '0700'
owner: root
group: root