9.0 KiB

Raw Blame History

Alpine Linux Cloud Image Builder

This repository contains the code and and configs for the build system used to create official Alpine Linux images for various cloud providers, in various configurations. This build system is flexible, enabling others to build their own customized images.

Pre-Built Offical Cloud Images

To get started with offical pre-built Alpine Linux cloud images, visit https://alpinelinux.org/cloud. Currently, we build official images for the following cloud platforms...

...we are working on also publishing offical images to other major cloud providers.

Each published image's name contains the Alpine version release, architecture, firmware, bootstrap, and image revision. These details (and more) are also tagged on the images...

Tag	Description / Values
name	`alpine-``release``-``arch``-``firmware``-``bootstrap``-r``revision`
project	`https://alpinelinux.org/cloud`
image_key	`release``-``arch``-``firmware``-``bootstrap``-``cloud`
version	Alpine version (`x.y` or `edge`)
release	Alpine release (`x.y.z` or `YYYYMMDD` for edge)
arch	architecture (`aarch64` or `x86_64`)
firmware	boot mode (`bios` or `uefi`)
bootstrap	initial bootstrap system (`tiny` = Tiny Cloud)
cloud	provider short name (`aws`)
revision	image revision number
imported	image import timestamp
import_id	imported image id
import_region	imported image region
published	image publication timestamp
description	image description

Although AWS does not allow cross-account filtering by tags, the image name can still be used to filter images. For example, to get a list of available Alpine 3.x aarch64 images in AWS eu-west-2...

aws ec2 describe-images \
  --region eu-west-2 \
  --owners 538276064493 \
  --filters \
    Name=name,Values='alpine-3.*-aarch64-*' \
    Name=state,Values=available \
  --output text \
  --query 'reverse(sort_by(Images, &CreationDate))[].[ImageId,Name,CreationDate]'

To get just the most recent matching image, use...

  --query 'max_by(Image, &CreationDate).[ImageId,Name,CreationDate]'

Build System

The build system consists of a number of components:

the primary build script
the configs/ directory, defining the set of images to be built
the scripts/ directory, containing scripts and related data used to set up image contents during provisioning
the Packer alpine.pkr.hcl, which orchestrates build, import, and publishing of images
the cloud_helper.py script that Packer runs in order to do cloud-specific import and publish operations

Build Requirements

Python (3.9.7 is known to work)
Packer (1.7.6 is known to work)
QEMU (6.1.0 is known to work)
cloud provider account(s)

Cloud Credentials

By default, the build system relies on the cloud providers' Python API libraries to find and use the necessary credentials, usually via configuration under the user's home directory (i.e. ~/.aws/, ~/.oci/, etc.) or or via environment variables (i.e. AWS_..., OCI_..., etc.)

The credentials' user/role needs sufficient permission to query, import, and publish images -- the exact details will vary from cloud to cloud. It is recommended that only the minimum required permissions are granted.

We manage the credentials for publishing official Alpine images with an "identity broker" service, and retrieve those credentials via the --use-broker argument of the build script.

The `build` Script

usage: build [-h] [--debug] [--clean] [--pad-uefi-bin-arch ARCH [ARCH ...]]
         [--custom DIR [DIR ...]] [--skip KEY [KEY ...]] [--only KEY [KEY ...]]
         [--revise] [--use-broker] [--no-color] [--parallel N]
         [--vars FILE [FILE ...]]
         {configs,state,rollback,local,upload,import,publish,release}

positional arguments:   (build up to and including this step)
  configs   resolve image build configuration
  state     refresh current image build state
  rollback  remove existing local/uploaded/imported images if un-published/released
  local     build images locally
  upload    upload images and metadata to storage
* import    import local images to cloud provider default region (*)
* publish   set image permissions and publish to cloud regions (*)
  release   mark images as being officially relased

(*) may not apply to or be implemented for all cloud providers

optional arguments:
  -h, --help              show this help message and exit
  --debug                 enable debug output
  --clean                 start with a clean work environment
  --pad-uefi-bin-arch ARCH [ARCH ...]
                          pad out UEFI firmware to 64 MiB ('aarch64')
  --custom DIR [DIR ...]  overlay custom directory in work environment
  --skip KEY [KEY ...]    skip variants with dimension key(s)
  --only KEY [KEY ...]    only variants with dimension key(s)
  --revise                remove existing local/uploaded/imported images if
                          un-published/released, or bump revision and rebuild
  --use-broker            use the identity broker to get credentials
  --no-color              turn off Packer color output
  --parallel N            build N images in parallel
  --vars FILE [FILE ...]  supply Packer with -vars-file(s) (default: [])

The build script will automatically create a work/ directory containing a Python virtual environment if one does not already exist. This directory also hosts other data related to building images. The --clean argument will remove everything in the work/ directory except for things related to the Python virtual environment.

If work/configs/ or work/scripts/ directories do not yet exist, they will be populated with the base configuration and scripts from configs/ and/or scripts/ directories. If any custom overlay directories are specified with the --custom argument, their configs/ and scripts/ subdirectories are also added to work/configs/ and work/scripts/.

The "build step" positional argument deterimines the last step the build script should execute -- all steps before this targeted step may also be executed. That is, build local will first execute the configs step (if necessary) and then the state step (always) before proceeding to the local step.

The configs step resolves configuration for all buildable images, and writes it to work/images.yaml, if it does not already exist.

The state step always checks the current state of the image builds, determines what actions need to be taken, and updates work/images.yaml. A subset of image builds can be targeted by using the --skip and --only arguments.

The rollback step, when used with --revise argument indicates that any unpublished and unreleased local, imported, or uploaded images should be removed and rebuilt.

As published and released images can't be removed, --revise can be used with configs or state to increment the revision value to rebuild newly revised images.

local, upload, import, publish, and release steps are orchestrated by Packer. By default, each image will be processed serially; providing the --parallel argument with a value greater than 1 will parallelize operations. The degree to which you can parallelze local image builds will depend on the local build hardware -- as QEMU virtual machines are launched for each image being built. Image upload, import, publish, and release steps are much more lightweight, and can support higher parallelism.

The local step builds local images with QEMU, for those that are not already built locally or have already been imported. Images are converted to formats amenable for import into the cloud provider (if necessary) and checksums are generated.

The upload step uploads the local image, checksum, and metadata to the defined storage_url. The import, publish, and release steps will also upload updated image metadata.

The import step imports the local images into the cloud providers' default regions, unless they've already been imported. At this point the images are not available publicly, allowing for additional testing prior to publishing.

The publish step copies the image from the default region to other regions, if they haven't already been copied there. This step will always update image permissions, descriptions, tags, and deprecation date (if applicable) in all regions where the image has been published.

NOTE: The import and publish steps are skipped for those cloud providers where this does not make sense (i.e. NoCloud) or for those which it has not yet been coded.

The release step marks the images as being fully released.

The `cloud_helper.py` Script

This script is meant to be called only by Packer from its post-processor block.

Build Configuration

For more in-depth information about how the build system configuration works, how to create custom config overlays, and details about individual config settings, see CONFIGURATION.md.

9.0 KiB Raw Blame History