alpine-awscli2

A pythonic method for installing awscli v2 on alpine.

Features:

• Simple Dockerfile with standard git clone, python deployment commands
• Uses dockerhub container to minimize dev system config requirements
• Reads current release version from official dockerhub aws-cli image
• Clones aws-cli source from AWS github repo with no code modification
• Builds using official dockerhub alpine image
• Optional: includes https://github.com/pypiserver/pypiserver to run as local repo
• Offers multiple methods for installing awscli on alpine containers
• No root requirement for awscli install
• Flexible, pythonic module install: user, venv, pipx

Installation Requirements

On Alpine Linux, in a container, VM, bare metal, one needs:

Resource	Reference
alpine	https://www.alpinelinux.org
python3	apk add python3 py3-pip
docker	https://wiki.alpinelinux.org/wiki/Docker#Installation
make	apk add make (development only; not needed to install from dockerhub)

Quick Start (fast and convenient)

Security Note: While this doesn't use sudo, it's still rather too trusting. It's never a good idea to run unknown commands with eval or by piping curl results into sudo or bash. To output the commands being passed to eval, just run the command inside the parenthesis. You can also capture this output to a file to make your own shell script.

Install awscli using docker and pip:

eval $(docker run rstms/alpine-awscli:latest install)

Install awscli using docker and pipx:

eval $(docker run rstms/alpine-awscli:latest pipx)

Development

To fork your own, you'll need github and dockerhub accounts.

git clone [email protected]:rstms/alpine-awscli2
cd alpine-awscli2
make build && make test && make publish 

Make Targets:

Target	Description
build	build a docker image, pull module source from github; build wheel files
wheels	copy the wheels from the image to local wheels directory
tarball	create awscli-wheels.tgz containing the wheels in a package directory
test	run the tests (examples)
publish	tag with alpine and awscli versions and push to dockerhub

Detailed Description:

The build/Dockerfile installs prerequisite software, configures the system to build the compiled modules, clones the the AWS github repo aws/aws-cli, selects the v2 branch, then prepares a distribution package using pip wheel

This results in a set of the wheel files required for awscli which can be installed to an alpine system with pip or any compatible alternative.

The container build then continues from the pypiserver/pypiserver dockerhub image, which can be run as a local PyPi-compatible module repository.

The resulting image contains the wheel files for all modules required for awscli in the directory /data/packages. The wheel files may be extracted as a tarball, copied to a directory, or the build image may be run as a repo and used with the command:

pip install -i http://locahost:8080 awscli

See the examples for various alternative installation methods.

Use the source, Luke.

Why does this project exist?

While the awscli package on PyPi just works for python2 users, those who wish to use current python versions or require the expanded V2 features may discover to their dismay that version 2 is not available using the expected idiom: pip install awscli==2.X.X

AWS has chosen not to distribute aws-cli V2.X.X and its botocore dependency on PyPi.

Their suggested installation is documented here:

• https://docs.aws.amazon.com/cli/latest/userguide/install-cliv2.html

While this method works on the list of supported environments, it fails on Alpine Linux due to limitations of the non-standard installation mechanism. One problem involves glibc dependency, but this isn't a requirement for the awscli module or for any of its dependencies despite what many have written. With proper system configuration, unmodified AWS sources build and install without issue using just the standard Python tools on Alpine.

For our uses, the software seems to be performing correctly. As time permits, we intend to utilize the test suite to verify complete functionality of the program. Careful study is required to determine if any tests make use of private AWS resources. (Help would be welcome here!)

The dockerhub image provides a minimal configuration single command installation. This workaround respects the PyPi license terms, which disallow distribution of unmodified clones of a project's packages when the owner has chosen not to publish on PyPi. Otherwise we could all just use PyPi and pip.

This situation has resulted in a fair amount of discussion:

• aws/aws-cli#4947 (Note how this thead ends)

This project is a humble attempt to solve a problem for one developer, released in the hope that it may be of use to others.