Untropy

Untropy is a flexible command line tool to manage full stack polyglot projects.

Features

• Centralized project configuration with multiple environments and deployment tiers.
• Powerful plugin system to wrap and configure external tools.
• Cookiecutter based templates auto-configured from the project configuration.
• ...

Sequence of commands

> untropy init
> untropy plugin add untropy-externaltool
> untropy plugin create untropy-localtool
> untropy cookie project_template -o awesome_microsvc
> untropy wrapper command ...
...

Rationale

Modern software development solutions are made up a fair amount of technologies and the number of tools involved can get huge. To cite some that are very common:

• Terraform
• Packer
• Helm
• Kustomize
• Kubectl
• Yarn
• Pip
• Docker
• Vault
• Ansible
• ...

TODO Give some examples (Mattermost ?)

The configuration and execution of all these tools generates a lot a entropy, understood as the quantity of information needed to completely describe the system:

• Configuration files in different formats.
• Bits of information duplicated across tool configuration.
• Mixed scripting languages (shell, python, ...).

To manage this complexity, development teams sometimes code several script based tools to ease different use cases (CI, command line, ...). Guided by the DRY principle, they group those disparate tools into one command, managed into one project.

Soon, this swiss army knife tool becomes a monolith the whole team depends on but don't wan't to touch. While they are doing independent microservices otherwise, they need to manage them with one big CI/CD monolith.

Developers and Devops that got on board early continue to add value to the tool while newbies and haters try at all cost to avoid it.

Untropy is an attempt to have the benefits of sharing a common set of tools and good practices while avoiding the maintenance nightmare.

Each tool/wrapper can be developped indepently from the others and integrated in the projects that make use of them.

Untropy leverages the power and maturity of the python ecosystem and some of its most interesting libraries and tools:

• pip For package and dependency management.
• pipx for keeping the Untropy Python Virtual environment separate from your development environment. This is useful when python is used in the project to avoid the dependencies version nightmare.
• Pydantic for configuration management.
• Click and Typer for command execution. Adding an existing typer or click based command to untropy is done with one command.

History

Untropy is inspired by Yinn, a tool I created while working at Unowhy. It started as an Ansible wrapper in shell and became a click based python CLI. More CI/CD related tools were then integrated into it. The python package ended containing the Ansible deployment roles of the common services used by different solutions of the company. As the package was multi-purpose and monolithic, it was facing a lot of releases. It became really hard to track the changes and to avoid regressions on client projects.

As it was based on Ansible, it had its versatility and power, in particular in templating files and wrapping existing tools. But it also had its flaws, like not being able to run or even being installed on Windows.

Credits

Inspiration comes from:

Terraform

Terraform is a declarative tools, not an imperative one as untropy is but it defines a common interface for providers and data sources. With terraform, as providers are Go compiled programs, if you want to have your own custom provider, you need to have the Go runtime installed, build your provider and upload it to hashicorp public registry or hashicorp cloud private one. It's uneasy to have a custom provider along your project. With Untropy, as python is interpreted, a custom plugin living in your project can be run from your project source code. You also keep the ability to publish your plugins to the cheese chop or your private repository.

The drawbacks are of course the need for the python runtime and the possible uncompatibility of libraries used by different plugins.

Composition with Terraform is also uneasy, needing some external tools like Terragrunt.

Docker compose

The similarity of untropy with docker compose is that you can compose Off the Shelf components with custom ones, each of them configured according to the use you make of them.

In Untropy, the Off The Shelf components would be the Pypi plugins that you use, and the custom ones are the ones you have in source code form in your project or the ones you download from a private package registry. As with docker-compose, you can configure each component (plugin) to your need.

Docker is both the build tool and the runtime execution tools. With untropy and python, you have the same way or working.