A gentle introduction to containers

Zach Lipp (he/him/his)

S2 ‘24

2024-04-03

Goals

  • Introduce high level container concepts
  • Provide an overview of the Docker CLI
  • Encourage new adopters to use containers

Roadmap

  • Reviewing container concepts
  • Introducing docker
  • Unpacking common docker commands
  • Introducing docker compose

Reviewing container concepts

What are containers?

Containers run application code in an isolated, reproducible, and portable environment

Why containers?

  1. Isolate and reproduce an application you are developing
  2. Standardizing your build requirements
  3. Giving you a simpler entrypoint to someone else’s application

Containers and images

Containers are running instances of images.

An image is a compressed collection of files and dependencies

Images consists of layers

Images are stored in registries

Containers and images

Containers are almost always Linux

Linux containers work by sharing the host’s kernel

🙋: You aren’t using Linux but you are using Docker
  • Docker Desktop (Mac, Windows) runs an entire Linux VM
  • Apple silicon needs ARM-based images

Introducing Docker

Demo: Running an image

What the heck happened?

  • Let’s walk through it!

docker pull

docker pull hides a lot

docker pull hello-world

docker pull hello-world:latest

docker pull docker.io/hello-world:latest

docker pull docker.io/library/hello-world:latest

docker pull docker.io/library/hello-world:latest@sha256:a26bff933ddc26d5cdf7faa98b4ae1e3ec20c4985e6f87ac0973052224d24302

docker pull broken down

docker pull broken down more

A full docker pull example

docker pull gcr.io/distroless/java17-debian12:latest@sha256:bf9b3faf63d2d74f28920711067a56ce1caa8000b619dbcf0bcccb9b48d3d155

docker run

Common docker run arguments

  • --name: name the container
  • positional: the name (and tag) of the image to run
docker run --name hello-world hello-world
  • --rm: delete the container after the entrypoint exits
docker run --rm hello-world:latest
  • -v: mount a volume into the container
docker run \
    -v $(pwd)/these/host/files:/this/container/dir \
    hello-world

Common docker run arguments

  • -p: forward a port from the container to the host
docker run \
    -p 8000:80 # host_port:container_port \
    nginx
  • -it: interactive teletype
  • --entrypoint: set a new entrypoint
docker run \
  -it \
  --entrypoint bash \ # try sh if bash isn't there
  nginx

Common docker run arguments

  • -e: set an environment variable in the container
docker run \
  -e WORLD=world \
  --entrypoint env \
  --rm \
  nginx

docker build

Introducing Dockerfiles

  • docker build works on a file format called a Dockerfile

Example Dockerfile

Source

FROM python:3.11-slim

WORKDIR home

COPY requirements.txt .

RUN pip install -r requirements.txt

COPY src .

COPY tests tests

ENTRYPOINT uvicorn --host 0.0.0.0 --port 1337 main:app
# ENTRYPOINT pytest

Example Dockerfile

FROM: Specifies the base image, usually the first layer

FROM python:3.11-slim

WORKDIR home

COPY requirements.txt .

RUN pip install -r requirements.txt

COPY src .

COPY tests tests

ENTRYPOINT uvicorn --host 0.0.0.0 --port 1337 main:app
# ENTRYPOINT pytest

Example Dockerfile

WORKDIR: Changes the directory in the image

FROM python:3.11-slim

WORKDIR home

COPY requirements.txt .

RUN pip install -r requirements.txt

COPY src .

COPY tests tests

ENTRYPOINT uvicorn --host 0.0.0.0 --port 1337 main:app
# ENTRYPOINT pytest

Example Dockerfile

COPY: Copies files from the host to the image

FROM python:3.11-slim

WORKDIR home

COPY requirements.txt .

RUN pip install -r requirements.txt

COPY src .

COPY tests tests

ENTRYPOINT uvicorn --host 0.0.0.0 --port 1337 main:app
# ENTRYPOINT pytest

Example Dockerfile

RUN: Run a command while building the image

FROM python:3.11-slim

WORKDIR home

COPY requirements.txt .

RUN pip install -r requirements.txt

COPY src .

COPY tests tests

ENTRYPOINT uvicorn --host 0.0.0.0 --port 1337 main:app
# ENTRYPOINT pytest

Example Dockerfile

ENTRYPOINT: the container’s default command

FROM python:3.11-slim

WORKDIR home

COPY requirements.txt .

RUN pip install -r requirements.txt

COPY src .

COPY tests tests

ENTRYPOINT uvicorn --host 0.0.0.0 --port 1337 main:app
# ENTRYPOINT pytest

Each command is a layer

…and each layer is cached

FROM python:3.11-slim

WORKDIR home

COPY requirements.txt .

RUN pip install -r requirements.txt

COPY src .

COPY tests tests

ENTRYPOINT uvicorn --host 0.0.0.0 --port 1337 main:app
# ENTRYPOINT pytest

Common docker build arguments

  • -t: A tag for the image, practically necessary
  • positional arg: The build context
    • COPY commands are relative to this path
    • COPY commands cannot look above the build context
    • Usually the project root
  • -f: The path to the Dockerfile. If not set, looks for build_context/Dockerfile

Example docker build command

docker build \
    -t REGISTRY/ORGANIZATION/IMAGE_NAME:IMAGE_TAG \
    .

Other docker commands

  • docker push: Push an image to a registry
  • docker exec: Start a shell in a running container
  • docker logs: View stdout/stderr for a container
  • docker rm/rmi: Remove stopped containers/images

DEEP BREATH

Why are there so many commands? And why are they so weird?

💡

We don’t have to use the base docker client

Introducing docker compose

Introducing docker compose

docker compose provides a more standard experience for working with Dockerfiles

docker compose lets us codify a lot of the specifics the docker CLI expects in arguments

Example service

FizzBuzz as a Service (FBaaS)

Source

Example docker-compose.yaml

Source

version: "3.9"
services:
  hub:
    build: example-services/hub
    ports:
      - "6000:6000"
    environment:
      FIZZER_PORT: 6001
      BUZZER_PORT: 6002
  fizzer:
    build: example-services/fizzer
  buzzer:
    build: example-services/buzzer
  jaeger:
    image: jaegertracing/all-in-one:1.42.0
    profiles:
      - tracing
    ports:
      # UI
      - "16686:16686"
    environment:
      COLLECTOR_OTLP_ENABLED: true
      LOG_LEVEL: info

Demo

Summary

  • We discussed container concepts and built up our vocabulary
  • We introduced the docker tool and CLI commands
  • We simplified the docker CLI with docker compose

Closing note: docker opens doors

You don’t have to be an expert sysadmin to try out a database (or cache, or distributed tracing visualizer…)

Closing note: Go nuts

dockerhub’s most pulled images

Thanks!

lippingoff.netlify.app/talks/docker_intro/