This article will give a glimpse of how to handle resources within a Kubernetes cluster maturely.

This article of the series will allow readers to understand EMR and Ranger integration solutions against Scenario 1: OpenLDAP + EMR-Native Ranger.

In this blog, you will learn some Docker best practices mainly focussed on Java applications. This is not only a theoretical exercise, but you will learn how to apply the best practices to your Dockerfiles. Enjoy! 1. Introduction Writing Dockerfiles seems easy: just pick an example from the internet and customize it to fit your needs. However, many examples are good for a development environment but are not production worthy. A production environment has more strict requirements especially concerning security. Besides that, Docker also provides guidelines for writing good Dockerfiles. It is just like writing code: you may know the syntax, but that does not mean you can write clean and good code in that specific programming language. The same applies to Dockerfiles. With this blog, you will learn some best practices, guidelines you can apply when writing Dockerfiles. The previous sentence deliberately says can apply and not must apply. It all depends on your use case. The example Dockerfile which often can be found when searching for Dockerfile for Java applications, is the following: Dockerfile FROM eclipse-temurin:17 RUN mkdir /opt/app ARG JAR_FILE ADD target/${JAR_FILE} /opt/app/app.jar CMD ["java", "-jar", "/opt/app/app.jar"] This Dockerfile is doing the following: FROM: Take eclipse-temurin:17 Java Docker image as base image; RUN: create a directory for the application jar file; ARG: provide an argument JAR_FILE so that you do not have to hard code the jar file name into the Dockerfile; ADD: add the jar file to the Docker image; CMD: the command that has to be executed when running the container, in this case, just start the Java application. In the next sections, you will change this Dockerfile to adhere best practices. The resulting Dockerfile of each paragraph is available in the git repository in directory Dockerfiles. At the end of each paragraph the name of the corresponding final Dockerfile will be mentioned where applicable. This post is inspired by the CIS Docker Benchmarks, the blog 10 best practices to containerize Java applications with Docker by Brian Vermeer and my own experiences. Code being used in this blog is available at GitHub. 2. Prerequisites The following prerequisites apply to this blog: Basic Linux knowlegde; Basic Java and Spring Boot knowledge; Basic Docker knowlegde. 3. Sample Application A sample application is needed in order to demonstrate the best practices. Therefore, a basic Spring Boot application is created containing the Spring Web dependency. The application can be run by invoking the following command from within the root of the repository: Shell $ mvn spring-boot:run For building the Docker image, a fork of the dockerfile-maven-plugin of Spotify will be used. The following snippet is therefore added to the pom file. XML com.xenoamess.docker dockerfile-maven-plugin 1.4.25 mydeveloperplanet/dockerbestpractices ${project.version} ${project.build.finalName}.jar The advantage of using this plugin is that you can easily reuse the configuration. Creating the Docker image can be done by a single Maven command. Building the jar file is done by invoking the following command: Shell $ mvn clean verify Building the Docker image can be done by invoking the following command: Shell $ mvn dockerfile:build Run the Docker image: Shell $ docker run --name dockerbestpractices mydeveloperplanet/dockerbestpractices:0.0.1-SNAPSHOT Find the IP-address of the running container: Shell $ docker inspect dockerbestpractices | grep IPAddress "SecondaryIPAddresses": null, "IPAddress": "172.17.0.3", "IPAddress": "172.17.0.3" In the above example, the IP-address is 172.17.0.3. The application also contains a HelloController which just responds with a hello message. The Hello endpoint can be invoked as follows: Shell $ curl http://172.17.0.3:8080/hello Hello Docker! Everything is now explained to get started! 4. Best Practices 4.1 Which Image to Use The image used in the Dockerfile is eclipse-temurin:17. What kind of image is this exactly? Therefore, you need to check how this image is built. Navigate to DockerHub; Search for ‘eclipse-temurin’; Navigate to the Tags tab; Search for 17; Sort by A-Z; Click the tag 17. This will bring you to the page where the layers are listed. If you look closely to the details of every layer and compare this to the tag 17-jre, you will notice that the tag 17 contains a complete JDK and tag 17-jre only contains the JRE. The latter is enough for running a Java application and you do not need the whole JDK for running applications in production. It is even a security issue when the JDK is used because the development tools could be misused. Besides that, the compressed size of the tag 17 image is almost 235MB and for the 17-jre it is only 89MB. In order to reduce the size of the image even further, you can use a slimmed image. The 17-jre-alpine image is such a slimmed image. The compressed size of this image is 59MB and reduces the compressed size with 30MB compared to the 17-jre. The advantage is that it will be faster to distribute the image because of its reduced size. Be explicit in the image you use. The above used tags are general tags which point to the latest version. This might be ok in a development environment, but for production it is better to be explicit about the version being used. The tag being used in this case will be 17.0.5_8-jre-alpine. And if you want to be even more secure, you add the SHA256 hash to the image version. The SHA256 hash can be found at the page containing the layers. When the SHA256 hash does not correspond to the one you defined in your Dockerfile, building the Docker image will fail. The first line of the Dockerfile was: Dockerfile FROM eclipse-temurin:17 With the above knowledge, you change this line into: Dockerfile FROM eclipse-temurin:17.0.5_8-jre-alpine@sha256:02c04793fa49ad5cd193c961403223755f9209a67894622e05438598b32f210e Build the Docker image and you will notice that the (uncompressed) size of the image is drastically reduced. It was 475MB and now it is 188MB. Shell $ docker images REPOSITORY TAG IMAGE ID CREATED SIZE mydeveloperplanet/dockerbestpractices 0.0.1-SNAPSHOT 0b8d89616602 3 seconds ago 188MB The resulting Dockerfile is available in the git repository with name 1-Dockerfile-specific-image. 4.2 Do Not Run As Root By default, the application runs as user root inside the container. This exposes many vulnerability risks and is not something you must want. Therefore, it is better to define a system user for your application. You can see in the first log line when starting the container that the application is started by root. Shell 2022-11-26 09:03:41.210 INFO 1 --- [ main] m.MyDockerBestPracticesPlanetApplication : Starting MyDockerBestPracticesPlanetApplication v0.0.1-SNAPSHOT using Java 17.0.5 on 3b06feee6c65 with PID 1 (/opt/app/app.jar started by root in /) Creating a system user can be done by adding a group javauser and a user javauser to the Dockerfile. The javauser is a system user which cannot login. This is achieved by adding the following instruction to the Dockerfile. Notice that creating the group and user are combined in one line by means of the ampersand signs in order to create only one layer. Dockerfile RUN addgroup --system javauser && adduser -S -s /usr/sbin/nologin -G javauser javauser The complete list of arguments which can be used for adduser are the following: -h DIR Home directory -g GECOS GECOS field -s SHELL Login shell -G GRP Group -S Create a system user -D Don’t assign a password -H Don’t create home directory -u UID User id -k SKEL Skeleton directory (/etc/skel) You will also need to change the owner of the directory /opt/apt to this new javauser, otherwise the javauser will not be able to access this directory. This can be achieved by adding the following line: Dockerfile RUN chown -R javauser:javauser /opt/app And lastly, you need to ensure that the javauser is actually used in the container by means of the USER command. The complete Dockerfile is the following: Dockerfile FROM eclipse-temurin:17.0.5_8-jre-alpine@sha256:02c04793fa49ad5cd193c961403223755f9209a67894622e05438598b32f210e RUN mkdir /opt/app RUN addgroup --system javauser && adduser -S -s /usr/sbin/nologin -G javauser javauser ARG JAR_FILE ADD target/${JAR_FILE} /opt/app/app.jar RUN chown -R javauser:javauser /opt/app USER javauser CMD ["java", "-jar", "/opt/app/app.jar"] In order to test this new image, you first need to stop and remove the running container. You can do so with the following commands: Shell $ docker stop dockerbestpractices $ docker rm dockerbestpractices Build and run the container again. The first log line mentions now that the application is started by javauser. Before, it stated that it was started by root. Shell 2022-11-26 09:06:45.227 INFO 1 --- [ main] m.MyDockerBestPracticesPlanetApplication : Starting MyDockerBestPracticesPlanetApplication v0.0.1-SNAPSHOT using Java 17.0.5 on ab1bcd38dff7 with PID 1 (/opt/app/app.jar started by javauser in /) The resulting Dockerfile is available in the git repository with name 2-Dockerfile-do-not-run-as-root. 4.3 Use WORKDIR In the Dockerfile you are using, a directory /opt/app is created. After that, the directory is several times repeated, because this is actually your working directory. However, Docker has the WORKDIR instruction for this purpose. When the WORKDIR does not exist, it will be created for you. Every instruction after the WORKDIR instruction will be executed inside the specified WORKDIR. So, you do not have to repeat the path every time. The second line contains the RUN instruction: Dockerfile RUN mkdir /opt/app Change this with using the WORKDIR instruction. Dockerfile WORKDIR /opt/app Now you can also remove every /opt/app reference, because the WORKDIR instruction ensures that you are in this directory. The new Dockerfile is the following: Dockerfile FROM eclipse-temurin:17.0.5_8-jre-alpine@sha256:02c04793fa49ad5cd193c961403223755f9209a67894622e05438598b32f210e WORKDIR /opt/app RUN addgroup --system javauser && adduser -S -s /usr/sbin/nologin -G javauser javauser ARG JAR_FILE ADD target/${JAR_FILE} app.jar RUN chown -R javauser:javauser . USER javauser CMD ["java", "-jar", "app.jar"] Build and run the container. As you can see in the logging, the jar file is still executed from within directory /opt/app: Shell 2022-11-26 16:07:18.503 INFO 1 --- [ main] m.MyDockerBestPracticesPlanetApplication : Starting MyDockerBestPracticesPlanetApplication v0.0.1-SNAPSHOT using Java 17.0.5 on fe5cf9223143 with PID 1 (/opt/app/app.jar started by javauser in /opt/app) The resulting Dockerfile is available in the git repository with name 3-Dockerfile-use-workdir. 4.4 Use ENTRYPOINT There exists a difference between the CMD instruction and the ENTRYPOINT instruction. More detailed information can be found in this blog. In short, use: ENTRYPOINT: when you build an executable Docker image using commands that always need to be executed. You can append arguments to the command if you like to; CMD: when you want to provide a default set of arguments but which are allowed to be overridden by the command line when the container runs. So, in the case for running a Java application, it is better to use ENTRYPOINT. The last line of the Dockerfile is: Dockerfile CMD ["java", "-jar", "app.jar"] Change it into the following: Dockerfile ENTRYPOINT ["java", "-jar", "app.jar"] Build and run the container. You will not notice any specific difference, the container just runs as it did before. The resulting Dockerfile is available in the git repository with name 4-Dockerfile-use-entrypoint. 4.5 Use COPY instead of ADD The COPY and ADD instructions seem to be similar. However, COPY is preferred above ADD. COPY does what it says, it just copies the file into the image. ADD has some extra features, like adding a file from a remote resource. The line in the Dockerfile with the ADD command: Dockerfile ADD target/${JAR_FILE} app.jar Change it by using the COPY command: Dockerfile COPY target/${JAR_FILE} app.jar Build and run the container again. You will not say a big change, besides that in the build log the COPY command is shown now instead of the ADD command. The resulting Dockerfile is available in the git repository with name 5-Dockerfile-use-copy-instead-of-add. 4.6 Use .dockerignore In order to prevent from accidentily adding files to your Docker image, you can use a .dockerignore file. With a .dockerignore file, you can specify which files may be sent to the Docker daemon or may be used in your image. A good practice is to ignore all files and to add explicitely the files you allow. This can be achieved by adding an asterisk pattern to the .dockerignore file which excludes all subdirectories and files. However, you do need the jar file into the build context. The jar file can be excluded from being ignored by means of an exclamation mark. The .dockerignore file looks as follows. You add it to the directory where you run the Docker commands from. In this case, you add it to the root of the git repository. Plain Text **/** !target/*.jar Build and run the container. Again, you will not notice a big change, but when you are developing with npm, you will notice that creating the Docker image will be much faster because the node_modules directory is not copied anymore into the Docker build context. The .dockerignore file is available in the git repository Dockerfiles directory. 4.7 Run Docker Daemon Rootless The Docker daemon runs as root by default. However, this causes some security issues as you can imagine. Since Docker v20.10, it is also possible to run the Docker daemon as a non-root user. More information how this can be achieved can be found here. An alternative way to accomplish this, is to make use of Podman. Podman is a daemonless container engine and runs by default as non-root. However, although you will read that Podman is a drop-in replacement for Docker, there are some major differences. One of them is how you mount volumes in the container. More information about this topic can be read here. 5. Conclusion In this blog, some best practices for writing Dockerfiles and running containers are covered. Writing Dockerfiles seems to be easy, but do take the effort in learning how to write them properly. Understand the instructions and when to use them.

This article explains the SSL Certificate configuration for secure communication over the HTTPS using .NET Core Web API and Docker.

PlatformCon 2022, the first-ever conference by and for platform engineers, dove into the latest trends in best practices. Here are the highlights you won’t want to miss.

This article compares the different geo-redundancy backup solutions and explains why cloud services are the best choice.

An analysis of Progressive Delivery options in the Cloud Native landscape will be done to explore how this enhancement can be added in a Kubernetes environment.

Comparison and trade-offs of Amazon MSK Serverless with other fully-managed SaaS Kafka cloud services like Confluent Cloud.

This article introduces the role of each connection parameter in MQTT and helps developers take their first steps in using MQTT.

In this article, we will explore different types of container-based deployment options, the benefits of each, and considerations for taking each approach.

Kubernetes is powerful — but complex. It’s easy to leave a cluster vulnerable to attack. The first line of defense are automated security checks.

This article will explain what Amazon EC2 is and how you can do it better!

Accidental Kubernetes PVC or namespace delete can Persistent Volume loss. Portworx team shares with us how to solve that.

Adapt your company to the newest technologies and trends with a loosely coupled architecture and the best DevOps practices.

In this article we are going to look at how we can use Kafka’s JDBC Source Connector to query a large table/view in batches both for new records and updated ones and publish them to a topic using Avro Schema.

As a programmer, much of our time at work is spent developing code that we usually spend a lot of time configuring the toolset that we use on a daily basis.

We have a microservices environment with Kubernetes(k8s), in which we develop a service that requires a database. We decided to do the migration management of the database using Liquibase.

This post tells how a GitHub data insight website with a distributed database reduced online serving latency from 1.11s to 123.6ms.

How can an attacker exploit leaked credentials? In this new series, we try to answer this question by imagining plausible attack scenarios. Fourth case: secrets are stolen with a malicious GitHub action.

Containerization technology can be adopted in all industries. Here are the top 15 ways to implement it using Docker.