Flox vs. container workflows

Containers are everywhere these days. They’re the de-facto deployment method in the industry, and they’re often used for local development as well to ensure that every developer gets the same environment. Flox gives you these same benefits without many of the papercuts, but the workflow is slightly different so it’s worth exploring.

Create a new development environment

Let’s say you’ve created a new directory for your project, myproject, and entered it:

mkdir myproject && cd myproject

Flox
Containers

Create a Flox environment for the project via flox init:

flox init

This creates a .flox directory in myproject. At this point you can enter the environment, but it doesn’t provide any new packages or functionality.

Create a new Dockerfile.

touch Dockerfile

At this point the Dockerfile isn’t really useable since it doesn’t say which image to build on top of (there’s no FROM line). Let’s pick the latest Ubuntu LTS release.

echo "FROM ubuntu:noble" >> Dockerfile

This already adds some variability in that which image this refers to may change from moment to moment.

Add packages

Now let’s install some packages. In a typical project there’s usually different subsets of packages:

Packages you don’t care about too much, so the latest version will suffice.
Packages whose versions you want pinned to a recent version.
Packages that you’re behind on updating because upgrading requires significant effort.

We’ll pick a single package from each category:

Latest is fine: curl
Specific version: yarn 1.22
Behind: python3.10

Flox
Containers

This is pretty straightforward:

flox install curl yarn@1.22 python3@3.10

Adding another package at a later date is as simple as running flox install <package> again.

Add a RUN command to your Dockerfile:

RUN apt update && apt install curl npm

yarn will be installed via npm, so we include that in the RUN command. This version of Python isn’t included in the Ubuntu 24.04 repositories, so you must install it from a Personal Package Archive (PPA). Do that with another RUN command:

RUN sudo add-apt-repository ppa:deadsnakes/ppa -y && \
    sudo apt update && \
    sudo apt install python3.10

This PPA is probably reputable, but now you’re managing third-party package repositories.Adding another package at a later date requires editing the first RUN command. This requires rebuilding later layers, such as the one that installs python3.10.

Configuration

What does it look like to configure your Flox environment compared to a container? A Flox environment is configured via a declarative TOML file called a “manifest”. The manifest for the environment created above looks like this:

version = 1

[install]
curl.pkg-path = "curl"
yarn.pkg-path = "yarn"
yarn.version = "1.22"
python3.pkg-path = "python3"
python3.version = "3.10"

[options]
systems = ["aarch64-darwin", "aarch64-linux", "x86_64-darwin", "x86_64-linux"]

One thing to note here is that the manifest defines a cross-platform environment out of the box to ensure that there are no nasty surprises down the line. A container is configured via a Dockerfile, which is an imperative sequence of commands.

FROM ubuntu:noble

RUN apt update && apt install curl npm

RUN sudo add-apt-repository ppa:deadsnakes/ppa -y && \
    sudo apt update && \
    sudo apt install python3.10

The upside to a Dockerfile is that the commands are familiar (e.g. apt install), but the order of commands matters and you end up stuffing a lot of commands into a single RUN command to avoid creating extra layers.

Use the development environment

Let’s say you want to do some work in the development environment. With Flox you’re put inside a subshell. With containers you can use a shell inside the container or connect to the running container via SSH. However, with containers you also need to mount in your source code, etc.

Flox
Containers

Activate the environment:

$ flox activate
flox [myproject] $ # now you're in the environment

Notice that there wasn’t a separate “build” step. When you install, uninstall, or edit a Flox environment it’s transactionally built to ensure that it’s always working.

First build the image:

docker build -t myproject .

Then start the image (mounting your source) and create a shell inside of it:

$ docker run -v ./src:/src myproject -d --name myproject_container
$ docker exec -it myproject_container bash
$ # now you're inside the container

Once you’re inside the containers, you don’t have any of your local tools, your shell’s aliases, access to your dotfiles, or configurations for your local tools.

Tear down the development environment

Flox
Containers

This is pretty straightforward:

flox [myproject] $ exit

This is pretty straightforward:

$ exit # leave the container shell
$ docker stop myproject_container

Perform initialization

Let’s say you need to move some files around, ensure a directory exists, or some other kind of initialization before doing work inside the development environment. We’ll do a pretend version of this by simply creating a directory foo.

Flox
Containers

This would be performed in the hook.on-activate script that’s run when activating your environment. You’ll add this by first running flox edit and then modifying the hook section of your manifest to look like this:

[hook]
on-activate = '''
  mkdir foo
'''

This would be performed with another RUN command, though it creates another layer in the image:

RUN mkdir foo

Suppose you work on a team and you’ve just set up the development environment. Now you want to share it with your team so you can ensure that everyone has the same environment.

Flox
Containers

Since flox init creates a .flox directory inside your project, you can simply check this directory into source control.

git add .flox && git commit -m "Add Flox environment" && git push

Anyone with Flox installed can now work on this project with two commands:

$ git clone <your repo>
$ flox activate

Any packages not locally cached would be downloaded. Since the Flox environment produces a lockfile each time it is built, every developer that does a flox activate with the same lockfile will get the same exact software down to the git revisions of the upstream source repositories.If your environment doesn’t need to be tied to this specific project you could also push the environment to FloxHub with flox push. Then your team would activate the environment as a “remote” environment:

$ # You
$ flox push
$ # Your coworker
$ flox activate -r your_user/myproject

The Dockerfile would be checked into source control, then each developer would build the image locally via docker build. However, since building the image isn’t reproducible (packages in the repositories may have new updates/bugs, base image may have been updated), each developer may have a slightly different development environment.It’s possible to ensure that all of the developers get the same software by building the image in a CI system and having developers avoid building the image locally.

Create a repository or CI rule that builds the development image every time the Dockerfile changes.
Build the image in CI and upload the image to a registry.
Developers docker pull the image when there are updates.

This is additional complexity though, and requires extra infrastructure.

Development-time services

In order to mimic the production environment you may want some services running during development (e.g. a web server, a database, etc). For this example let’s say you want a minimal Caddy server running with some environment variables set.

Flox
Containers

First install Caddy:

flox install caddy

Then edit your manifest to create a new service:

[services.server]
command = "caddy run"
vars.VAR1 = "var1"
vars.VAR2 = "var2"

Since Flox environments aren’t isolated from the host machine’s network you don’t need to map any ports.You can start this service from inside the environment with flox services start, or you can have it start automatically when entering the environment via flox activate --start-services.

First create a docker-compose.yml file that pulls in a Caddy image:

version: "3.8"

services:
  caddy:
    image: caddy:latest
    container_name: caddy_server
    ports:
      - "80:80"
      - "443:443"
    environment:
      - VAR1=var1
      - VAR2=var2

This configuration maps the container’s ports so that they’re accessible from the host machine.The service is started separately via docker-compose up, but you may also add the development container to the docker-compose.yml file so that it’s started at the same time as the server.

Run tests in CI

Say you’ve done some development and now want to run your changes through CI.

Flox
Containers

A CI system would activate the Flox environment for the repository and then run a specified command inside the environment for each step of the CI job.Since a Flox environment contains a lockfile, a CI system that runs flox activate will get exactly the same software as the developer pushing the changes. This greatly reduces the number of “it works on my machine” instances.Flox provides a number of plugins for CI providers, including GitHub Actions, CircleCI, and GitLab. See the CI/CD tutorial for more information.

Send artifacts to production

Now that you have a working development environment, you need to build a container so that it can be deployed.

Flox
Containers

You can create a container from an environment via the flox containerize command, and this image will contain the same exact software (again, down to the git revisions of the upstream source repositories) that you used for both local development and CI. If your environment already contains the programs you want to run in production, you’re in good shape.You can also produce build artifacts from a Flox environment. See the builds documentation for more information.

Typically your Dockerfile will contain multiple stages, possibly a base builder stage, a development stage that builds on builder, and a production stage that contains only the executable.You would typically build this image in CI via a docker build command that targets the production stage. CI would also upload the production image to a container registry.

Introduction

Install Flox

Quickstart

Tutorials

Imageless Kubernetes

Concepts

Languages

Customer

Flox vs. container workflows

Create a new development environment

Add packages

Configuration

Use the development environment

Tear down the development environment

Perform initialization

Development-time services

Run tests in CI

Send artifacts to production

Introduction

Install Flox

Quickstart

Tutorials

Imageless Kubernetes

Concepts

Languages

Customer

Documentation Index

​Create a new development environment

​Add packages

​Configuration

​Use the development environment

​Tear down the development environment

​Perform initialization

​Share the environment with your team

​Development-time services

​Run tests in CI

​Send artifacts to production

Create a new development environment

Add packages

Configuration

Use the development environment

Tear down the development environment

Perform initialization

Share the environment with your team

Development-time services

Run tests in CI

Send artifacts to production