Tips on Writing Dockerfile, Reduce Sizes and Build Time of Docker ImageNovember 8, 2019 · 4 Min Read · 0 Comment · In Docker, Docker Image
For the past couple of years, I have been creating Dockerfile for different projects. based on those experiences I am going to share some tips on writing docker files to communication between containers.
_FYI: before reading this article, please read the article from official docker blog site to learn about best practices for creating Dockerfile._
Table of contents ︎
Look for official docker images ︎
When you are trying to run a project in Docker, its better to use official images rather than writing your own. If you need to install different packages that does not come with the official image, then you can extend the official images. For example, if you want to install Flask in your Docker Container, instead of creating an Image from Ubuntu or Alpine Linux, use Python’s official image:
FROM python:3.7-alpine RUN pip install flask
Reduce the size of docker image ︎
Reducing size of a docker image is very important. Because before you know it, your image might take 1-3 GB of space. There are different ways you can reduce it. For example:
Alpine based images ︎
For most of the official docker images, there is a alpine varient. Alpine varients are much more light weight then their ubuntu/debian counter parts. Also, alpine base image is only 5MB.
2. Delete packages after dependent applications are installed ︎
Lets say you want to install
psycopg2 python package in your docker, for that you need to add packages like
musl-dev etc. But those packages won’t be needed once
psycopg2 is installed, so its better to delete them. For that, in alpine you can use the following code:
RUN apk update \ && apk add --virtual build-deps gcc python3-dev musl-dev \ && apk add postgresql-dev \ && pip install psycopg2 \ && apk del build-deps
Explanation: here I am installing the packages using
--virtual build-deps tag, which means those packages will be installed
build-deps directory. After
psycopg2 is installed, I can simply delete it.
3. Remove cache ︎
Remove cache from alpine image by adding
--no-cache at the end of
apk add. But down side is that each time you build the image, your packages will be installed again.
Probably you can do the same in
debian based images using
-- no-install-recommends on apt-get install.
You can do the same thing with package manager, for example in
RUN pip install --no-cache-dir django
4. Reduce layers in Dockerfile ︎
Reduce multiple lines into one if possible. Because each layer takes space in image. For example, reduce the following:
RUN apk add gcc RUN apk add python3-dev
To this one:
RUN apk add gcc python3-dev
5. No need to install debug tools ︎
You don’t need debug tools like
Vim in Docker(at least not in production). Install only the necessary packages.
Reduce build time by caching layers ︎
Sometimes its better to cache layers which takes long time to install or download. For example:
COPY /PLUGIN_DIR/pom.xml . RUN mvn dependency:go-offline RUN mvn package
In this code, I use an extra layer
RUN mvn dependency:go-offline. When the image is built for the first time, this layer executes to resolve dependencies. But consecutive builds after that, they don’t execute this layer as its alredy cached. It reduces build time significantly.
Here is another example:
RUN pip3 install \ pandas==0.25.2 \ numpy==1.17.3 \ psycopg2==2.8.4 \ gensim==3.8.1 RUN pip3 install -r requirements.pip
Here you can see I am using extra layer to install pandas, numpy, scipy etc, and that layer will be cached after first build.
In conclusion ︎
Thats it for now. I will add more stuff in future in this article or future articles. So stay tuned 😄 . Finally, feel free to share your feedback in comment section below.
Relevent posts ︎
- Tips on using Docker Compose, Build Argument and Environment Variable
- Tips on Using Docker Volume, Entrypoint and .dockerignore file
Last updated: May 27, 2020
For the past couple of days, I am testing out Manjaro linux. Most of the time, I was …
Imagine, you have some microservices and you want to use API from one service to another, …