EC2 instances are available in a variety of different sizes with different specifications but all of them currently run on one of two types of virtualization technology: paravirtualization (PV) or hardware virtual machine (HVM).

PV instances boot with a special boot loader called PV-GRUB, which starts the boot cycle and then chain loads the kernel specified in the menu.lst file on your image. Paravirtual guests can run on host hardware that does not have explicit support for virtualization, but they cannot take advantage of special hardware extensions such as enhanced networking or GPU processing.

HVM instances are presented with a fully virtualized set of hardware and boot by executing the master boot record of the root block device of your image. This virtualization type provides the ability to run an operating system directly on top of a virtual machine without any modification, as if it were run on the bare-metal hardware. The Amazon EC2 host system emulates some or all of the underlying hardware that is presented to the guest. (from: Linux AMI Virtualization Types).

Only older instance types support PV instances, all new instance types use an HVM hypervisor. For the purposes of setting up Alpine Linux on EC2 this mainly impacts the way the bootloader is configured.

EC2 instances are created from templates called Amazon Machine Images (AMI). AMIs are built using existing running instances and then are turned into AMIs using a snapshot process. Note that an AMI is region specific so you must either follow this guide in each region you plan to deploy Alpine Linux instances or you must copy the AMI using either the console or the AWS command line client to each region you plan to use.

All instance types are 64-bit. 32-bit images will not work.

This guide will create the smallest possible (1GB) EBS-backed image which acts as a virtual USB stick that will boot and run Alpine Linux.

Initial Setup

All AMI creation tasks follow the same set of setup steps which require a running Linux instance and an attached EBS volume. The Linux instance will only be used to set up the root filesystem for your Alpine Linux system so any distro works, Amazon Linux is the default so, where relevant, this guide will use that distro.

Create an EC2 instance, t2.micro running Amazon Linux should suffice
Create a 1GB EBS volume and attach it to the instance as /dev/sdf (this will appear on an HVM instance as /dev/xvdf)
Format the disk as ext4, there is no need to partition the drive:
mkfs.ext4 /dev/xvdf
Mount the drive to the host:
mkdir /mnt/target && mount /dev/xvdf /mnt/target
Fetch a copy of the latest virt release:
curl -o /tmp/alpine-release.iso http://dl-4.alpinelinux.org/alpine/v3.7/releases/x86_64/alpine-virt-3.7.0-x86_64.iso
Mount the release iso:
mkdir /mnt/source && mount -o loop /tmp/alpine-release.iso /mnt/source
Copy the files to the target
cp -av /mnt/source/boot /mnt/target

The sudo package is not included in the on-disk repository for the virt image which means that it will not be accessible for installation at system boot time; to enable this you will need a copy of the repository from the extended image. Download this image and copy the apks folder to the target.

Fetch a copy of the latest virt release:
curl -o /tmp/alpine-extended-release.iso http://dl-4.alpinelinux.org/alpine/v3.7/releases/x86_64/alpine-extended-3.7.0-x86_64.iso
Mount the release iso:
mkdir /mnt/source-extended && mount -o loop /tmp/alpine-extended-release.iso /mnt/source-extended
Copy the files to the target
cp -av /mnt/source-extended/apks /mnt/target

Because the packages on Amazon Linux are a little older than the ones that ship with Alpine and because you will be creating an apkovl file for initial system configuration, you will either need an existing Alpine Linux system or a copy of the minirootfs image into which you can chroot. To set up a chroot:

Fetch a copy of the latest minirootfs release:
curl -o /tmp/alpine-minirootfs.tar.gz http://dl-4.alpinelinux.org/alpine/v3.7/releases/x86_64/alpine-minirootfs-3.7.0-x86_64.tar.gz
Unpack the root fs:
mkdir /tmp/alpine-chroot && tar -C /tmp/alpine-chroot -xvf /tmp/alpine-minirootfs.tar.gz
Copy the system resolv.conf into the chroot so you will have internet access:
cp /etc/resolv.conf /tmp/alpine-chroot/etc

Create an Apkovl File

The top level of the EBS volume will be scanned during boot for files named *.apkovl.tar.gz and the first one that is found will be unpacked and overlayed onto the file system. The goal of this archive is to configure the system so that you will be able to login after the system boots. At a minimum this means:

Configure networking to use dhcp and start when the system boots
Install an SSH daemon
Configure a user with your SSH keys
Enable sudo so that you can gain root access

Set up the chroot and chroot into it, then add any desired packages. The busybox-initscripts package provides the required files for udhcpc which will allow DHCP to be enabled. The syslinux package is required to install the boot-loader but is not required on the final system.

mkdir /tmp/alpine-chroot/mnt/target
mount -o bind /mnt/target /tmp/alpine-chroot/mnt/target
mount -t proc none /tmp/alpine-chroot/proc
mount -t devtmpfs none /tmp/alpine-chroot/dev
mount -t sysfs none /tmp/alpine-chroot/sys

chroot /tmp/alpine-chroot sh

apk update
apk add alpine-conf sudo openssh busybox-initscripts syslinux

The Alpine init system will start up the core required services if the file /etc/.default_boot_services exists, so create that.

touch etc/.default_boot_services

Set up networking:

cat > /etc/network/interfaces <<EOF
auto lo
iface lo inet loopback

auto eth0
iface eth0 inet dhcp
EOF

ln -s /etc/init.d/networking /etc/runlevels/default/
ln -s /etc/init.d/sshd /etc/runlevels/default/

Add a user, set an SSH key, and ensure that this user can sudo:

adduser -D -s /bin/sh alpine
addgroup alpine wheel

mkdir /home/alpine/.ssh
chmod 700 /home/alpine/.ssh
cat > /home/alpine/.ssh/authorized_keys <<EOF
... your ssh public key(s) ...
EOF
chmod 600 /home/alpine/.ssh/authorized_keys

sed -i '/%wheel/s/^# //' /etc/sudoers

Exclude backup files and any files that are going to exist in the unpacked image. Then create an apkovl bundle.

lbu exclude \
    etc/group- etc/passwd- etc/shadow- \
    etc/apk/keys etc/apk/arch etc/apk/repositories \
    etc/os-release \
    etc/issue \
    etc/alpine-release
lbu include /home/alpine/.ssh
lbu package amazon.apkovl.tar.gz

Verify that everything you expect to be in the apovl file is there and if not you can modify the apkovl.

Finally, copy the apkovl to /mnt/target.

cp amazon.apkovl.tar.gz /mnt/target

EBS Backed HVM AMI

For HVM instances you will need to install a boot-loader. Any bootloader should do the trick but this guide will use syslinux because it's fast and small and you don't have to support any odd hardware or system layout.

First you will need to ensure that the ext4 driver is loaded and may as well shorten the timeout to speed instance boot time.

sed -i \ -e '/usb-storage/s//usb-storage,ext4/' \ -e '/^TIMEOUT/s/20/2/' \ /mnt/target/boot/syslinux/syslinux.cfg

Finally, install the bootloader:

extlinux -i /mnt/target/boot

EBS Backed PV AMI

For PV AMIs PV-GRUB will attempt to read /boot/grub/menu.lst to load the kernel. First create a /boot/grub/grub.conf:

mkdir -p /mnt/target/boot/grub

cat > target/boot/grub/grub.conf <<EOF 
default=0
timeout=2
hiddenmenu

title Alpine Linux
root (hd0)
kernel /boot/vmlinuz-virthardened alpine_dev=xvda1:ext4 modules=loop,squashfs,sd-mod,ext4 console=hvc0 pax_nouderef BOOT_IMAGE=/boot/vmlinuz-virthardened
initrd /boot/initramfs-virthardened
EOF

ln -s /mnt/target/boot/grub/grub.conf  /mnt/target/boot/grub/menu.lst

Create the AMI

The following will need to be done in the AWS console.

Detach the new volume (make note of the volume ID)
Launch a new instance, use the defaults; you are going to cannibalize it in a moment
Once the instance starts, stop but do not terminate it
Under EBS, detach the existing volume, and attach the Alpine Linux volume as /dev/xvda (for HVM) or /dev/sda1 (for PV)
Restart the instance
Log in and make sure it works
Do any final cleanups necessary. Only make changes appropriate for an AMI, you are going to snapshot this instance and use it as the base for the AMI.
Stop but do not terminate the instance
Right click the stopped instance and choose 'Create Image (EBS AMI)'
Once the AMI creation finishes you can cleanup the instances and EBS volumes used