AWS: Template for jessie image

Based on the official debian image, with the following changes:

* Switched extlinux -> grub, because we need to change kernel options
to enable the memory cgroup controller, and extlinux is harder and has
reboot problems
* Added packages that would otherwise be installed as part of the boot
(just an optimization)
* Also add the cloud-initramfs-growroot package; with it the root
volume will resize.
* We add panic=10 & oops=panic to kernel options
* We install the packages as per the base image, except we install
  awscli from pip, because the repo version is really old.
pull/6/head
Justin Santa Barbara 2016-02-14 11:19:21 -05:00
parent 81c8494307
commit f117b26c07
2 changed files with 130 additions and 0 deletions

View File

@ -0,0 +1,40 @@
## Kubernetes-optimized images
This directory contains manifests for building Kubernetes-optimized images for
various clouds (currently just AWS). It is currently highly experimental, and
these images are not used by default (though you can pass `AWS_IMAGE` to the
AWS kube-up script if you're feeling brave).
Advantages of an optimized image:
* We can preinstall packages that would otherwise require a download. Great
for speed, and also for reliability (in case the source repository is down)
* We can make kernel configuration changes that might otherwise require a
reboot, or even apply kernel patches if we really want to. For example,
Debian requires a kernel boot parameter to enable the cgroup memory
controller, which we require.
* The more configuration we can do in advance, the easier it is for people that
don't want to use kube-up to get a cluster up and running.
Advantages of a harmonized image:
* All the platforms can test with the same versions of software, rather than
relying on whatever image happens to be optimal on that cloud.
## bootstrap-vz
Currently images are built using
(bootstrap-vz)[https://github.com/andsens/bootstrap-vz], because this is
default builder for the official Debian images, and because it supports
multiple clouds including AWS, Azure & GCE. It also supports KVM, which should
support OpenStack.
## Building an image
A go program/script to build images in (in
progress)[https://github.com/kubernetes/contrib/pull/486], in the contrib
project.
[![Analytics](https://kubernetes-site.appspot.com/UA-36037335-10/GitHub/cluster/cloudimages/README.md?pixel)]()

View File

@ -0,0 +1,90 @@
---
name: k8s-1.2-debian-{system.release}-{system.architecture}-{provider.virtualization}-{%Y}-{%m}-{%d}-ebs
provider:
name: ec2
virtualization: hvm
enhanced_networking: simple
description: Kubernetes 1.2 Base Image - Debian {system.release} {system.architecture}
bootstrapper:
workspace: /target
# tarball speeds up development, but for prod builds we want to be 100% sure...
# tarball: true
system:
release: jessie
architecture: amd64
# We use grub, not extlinux.
# See https://github.com/andsens/bootstrap-vz/issues/182
# extlinux makes it harder to modify boot args, and may have reboot problems
# bootloader: extlinux
bootloader: grub
charmap: UTF-8
locale: en_US
timezone: UTC
volume:
backing: ebs
partitions:
type: msdos
root:
filesystem: ext4
size: 8GiB
packages:
mirror: http://cloudfront.debian.net/debian
install:
# these packages are included in the official image
- python-boto
- python3-boto
- apt-transport-https
- lvm2
- ncurses-term
- parted
- bootlogd
- cloud-init
- cloud-utils
- gdisk
- sysvinit
- systemd
- systemd-sysv
# these packages are included in the official image, but we remove them
# awscli : we install from pip instead
# cloud-initramfs-growroot will resize the master partition on boot
- cloud-initramfs-growroot
# These packages would otherwise be installed during first boot
- aufs-tools
- curl
- python-yaml
- git
- nfs-common
- bridge-utils
- logrotate
- socat
- python-apt
- apt-transport-https
- unattended-upgrades
- lvm2
- btrfs-tools
# So we can install the latest awscli
- python-pip
plugins:
cloud_init:
metadata_sources: Ec2
username: admin
ntp: {}
commands:
commands:
# Install python-pip
- [ 'chroot', '{root}', 'pip', 'install', 'awscli' ]
# We perform a full replacement of some grub conf variables:
# GRUB_CMDLINE_LINUX_DEFAULT (add memory cgroup)
# GRUB_TIMEOUT (remove boot delay)
# (but leave the old versions commented out for people to see)
- [ 'chroot', '{root}', 'touch', '/etc/default/grub' ]
- [ 'chroot', '{root}', 'sed', '-i', 's/^GRUB_CMDLINE_LINUX_DEFAULT=/#GRUB_CMDLINE_LINUX_DEFAULT=/g', '/etc/default/grub' ]
- [ 'chroot', '{root}', 'sed', '-i', 's/^GRUB_TIMEOUT=/#GRUB_TIMEOUT=/g', '/etc/default/grub' ]
- [ 'chroot', '{root}', '/bin/sh', '-c', 'echo "# kubernetes image changes" >> /etc/default/grub' ]
- [ 'chroot', '{root}', '/bin/sh', '-c', 'echo "GRUB_CMDLINE_LINUX_DEFAULT=\"cgroup_enable=memory oops=panic panic=10 console=ttyS0\"" >> /etc/default/grub' ]
- [ 'chroot', '{root}', '/bin/sh', '-c', 'echo "GRUB_TIMEOUT=0" >> /etc/default/grub' ]
- [ 'chroot', '{root}', 'update-grub2' ]