distroless VS docker-socket-proxy

lots of questions here regarding what this product is. I guess i can provide some information for the context, from a perspective of an outside contributor.

Chainguard Images is a set of hardened container images.

They were built by the original team that brought you Google's Distroless (https://github.com/GoogleContainerTools/distroless)

However, there were few problems with Distroless:

1. distroless were based on Debian - which in turn, limited to Debian's release cadence for fixing CVE.

2. distroless is using bazelbuild, which is not exactly easy to contrib, customize, etc...

3. distroless images are hard to extend.

Chainguard built a new "undistro" OS for container workload, named Wolfi, using their OSS projects like melange (for packaging pkgs) and apko (for building images).

The idea is (from my understanding) is that

1. You don't have to rely on upstream to cut a release. Chainguard will be doing that, with lots of automation & guardrails in placed. This allow them to fix vulnerabilties extremely fast.

> If you have one image based on Ubuntu in your stack, you may as well base them all on Ubuntu, because you only need to download (and store!) the common base image once

This is only true if your infrastructure is static. If your infrastructure is highly elastic, image size has an impact on your time to scale up.

Of course, there are better choices than Alpine to optimize image size. Distroless (https://github.com/GoogleContainerTools/distroless) is a good example.

The same as our code dependencies, container updates can include security patches and bug fixes and improvements. However, they can also include breaking changes and it is crucial you test them thoroughly before putting them into production. Wherever possible, I recommend using the distroless base image which will drastically reduce both your image size, your risk vector, and therefore your maintenance version going forward.

# Use a large Node.js base image to build the application and name it "build" FROM node:18-alpine as build WORKDIR /app # Copy the package.json and package-lock.json files into the working directory before copying the rest of the files # This will cache the dependencies and speed up subsequent builds if the dependencies don't change COPY package*.json /app # You might want to use yarn or pnpm instead RUN npm install COPY . /app RUN npm run build # Instead of using a node:18-alpine image, we are using a distroless image. These are provided by google: https://github.com/GoogleContainerTools/distroless FROM gcr.io/distroless/nodejs:18 as prod WORKDIR /app # Copy the built application from the "build" image into the "prod" image COPY --from=build /app/.output /app/.output # Since this image only contains node.js, we do not need to specify the node command and simply pass the path to the index.mjs file! CMD ["/app/.output/server/index.mjs"]

Lots of examples without the entire OS as other comments mention, an example would be Googles distroless[0]

[0]: https://github.com/GoogleContainerTools/distroless

Docker doesn't do this all the time. Distroless Docker containers are relatively common. https://github.com/GoogleContainerTools/distroless

Deployment: https://github.com/GoogleContainerTools/distroless

Or use distroless image as it includes one, among others. https://github.com/GoogleContainerTools/distroless/blob/main/base/README.md

I just found this the other day. You might be interested I haven't done myself yet https://github.com/Tecnativa/docker-socket-proxy

I think you could provide access to the socket using a "docker-socket-proxy" container. It allows other containers to access the docker socket, you can even control which actions are allowed and which are not. You can use a bridge network for the communication to the socket-proxy container, so the socket-proxy container does not need to map/expose any ports. In the other container you need to set the "DOCKER_HOST" env variable accordingly, e.g. "DOCKER_HOST=tcp://mydockersockerproxycontainer:2375". https://github.com/Tecnativa/docker-socket-proxy

I use the container docker socket proxy

There are Docker socket proxys (like docker-socket-proxy 😉) that are made exactly for this. You can pass only read access to the socket and even restrict what resources can be read.

So far I have Traefik set up and tested (along with some security lockdowns https://github.com/Tecnativa/docker-socket-proxy). This is working well: I can manually create containers, get a cert, dynamic hostnames, etc.

May be not necessarily: https://github.com/Tecnativa/docker-socket-proxy

Many of you might already be familiar with Tecnativa's docker-socket-proxy which says:

version: "3.7" services: traefik: image: traefik:v2.6 command: # Entrypoints configuration - --entrypoints.web.address=:80 # Docker provider configuration - --providers.docker=true # Makes sure that services have to explicitly direct Traefik to expose them - --providers.docker.exposedbydefault=false # Use the secure docker socket proxy - --providers.docker.endpoint=tcp://socket_proxy:2375 # Default docker network to use for connections to all containers - --providers.docker.network=traefik_public # Logging levels are DEBUG, PANIC, FATAL, ERROR, WARN, and INFO. - --log.level=info ports: - 80:80 networks: - traefik_public - socket_proxy restart: unless-stopped depends_on: - socket_proxy # https://github.com/traefik/whoami whoami: image: traefik/whoami:v1.7.1 labels: # Explicitly instruct Traefik to expose this service - traefik.enable=true # Router configuration ## Listen to the `web` entrypoint - traefik.http.routers.whoami_route.entrypoints=web ## Rule based on the Host of the request - traefik.http.routers.whoami_route.rule=Host(`whoami.karvounis.tutorial`) - traefik.http.routers.whoami_route.service=whoami_service # Service configuration ## 80 is the port that the whoami container is listening to - traefik.http.services.whoami_service.loadbalancer.server.port=80 networks: - traefik_public # https://github.com/Tecnativa/docker-socket-proxy # Security-enhanced proxy for the Docker Socket socket_proxy: image: tecnativa/docker-socket-proxy:latest restart: unless-stopped environment: NETWORKS: 1 SERVICES: 1 CONTAINERS: 1 TASKS: 1 volumes: - /var/run/docker.sock:/var/run/docker.sock:ro networks: - socket_proxy networks: traefik_public: external: true socket_proxy: external: true