Customizing Docker Images with Dockerfile: Instructions and Best Practices

This guide demonstrates how to create and use a Dockerfile to build customized Docker images, illustrating each step from initializing a project directory to defining image layers.

First, a simple Dockerfile for an Nginx image is created with two instructions: FROM nginx and RUN echo '<h1>Hello, Docker!</h1>' > /usr/share/nginx/html/index.html . The file is placed in an empty directory using mkdir mynginx && cd mynginx && touch Dockerfile .

The FROM instruction selects a base image and must be the first line; common sources include official images on Docker Hub, OS images such as ubuntu , or the virtual scratch image for completely empty layers.

The RUN instruction executes shell commands during the build. Two formats exist: shell form ( RUN <command> ) and exec form ( RUN [\"executable\",\"arg1\",\"arg2\"] ). To keep images small, related commands should be combined into a single RUN using && and line‑continuation backslashes, followed by cleanup steps that remove temporary files and package caches.

Image construction is performed with docker build -t nginx:v3 . . The trailing dot denotes the build context, which is the directory containing the Dockerfile and any files referenced by COPY or ADD . Docker packs this directory and sends it to the daemon; a .dockerignore file can exclude unnecessary files.

Advanced build sources include remote Git repositories ( docker build https://github.com/user/repo.git ), tar archives ( docker build http://server/context.tar.gz ), and reading a Dockerfile or context from standard input ( docker build - < Dockerfile ).

The article then reviews the remaining Dockerfile instructions: COPY and ADD for adding files, with ADD also handling URLs and automatic extraction of compressed archives; CMD and ENTRYPOINT for defining the container’s default command and entry point, illustrating how ENTRYPOINT allows additional arguments to be passed at runtime; ENV for setting environment variables that can be referenced in later instructions; VOLUME for declaring mount points; EXPOSE for documenting which ports the container listens on; and WORKDIR for setting the working directory for subsequent instructions.