Containers & Docker

Challenges in Deploying Software

For simple applications, we are often targetting a single installation of that program running on a standalone computer. Deployment in this case consists of a user running the installer, and having the application installed and configured for that single machine. If something goes wrong, the user can typically just uninstall and reinstall the application. This type of packaging and deployment model makes sense:

• The software targets a single user, so there is rarely more than a single person affected.
• We can accurately predict the hardware demands of our software ahead of time, and we can inform the user of our requirements.
• Our application requirements rarely change after the software is installed. If the application runs fine after installation, we can be fairly confident that it will continue to run properly.

However, when we move to deploying software to a server, we are faced with additional challenges.

• A single installation may be used by hundreds or thousands of users. It is difficult to just uninstall-and-reinstall.
• Our hardware requirements may change based on the number of users that are using our software, and we need the ability to adjust to that changes in demand. For example, we may launch with a minimal hardware configuration that works for a small number of users, but which is insufficient for hundreds or thousands of users. Sometimes we can meet increased demand by adding resources, but at other times we may need to move to more capable hardware (e.g. it may be practical to increase memory on a machine from 64 GB to 128 GB, but impractical to add a second or third processor).
• Downtime becomes a consideration when your system is in continuous use (much more likely on a shared system).

How do we address this?

Standalone: One or more applications are installed on the same hardware, under the same operating system.

• Application share resources, which the OS has to allocate and manage.
• Security concerns with applications installed together.

Virtualization: Multiple virtual machines can be run on the same hardware. Each one is an abstraction of a physical machine, with its own resources and dependencies.

• Each virtual machine is running a complete OS. Can be resource intensive, since each VM is allocated its own memory, CPU cycles etc.
• Provides the ability to adjust how physical resources are shared across VMs (e.g. if we had 128 GB of RAM, we could split it among VMs in any way that made sense).
• Provides isolation of each application into its own OS instance.

Container: an isolated environment for running an application.

• Run applications (not OS) in isolation.
• Containers are processes that use the OS of the host to run an image containing the application/
• Very lightweight, fast to startup containers compared to a VM.

There are significant advantages to using containers:

• Containers are significantly smaller than virtual machines, and use fewer hardware resources.
• You can deploy containers anywhere, on any physical and virtual machines and even on the cloud.
• Containers are pretty lightweight and easy to start/stop and scale out.

What is Docker?

Docker is a containerization platform. We can use Docker to create a deployment container that contains the complete runtime environment, which can then be run anywhere that has Docker installed.

Installing Docker software provides you with the Containerization Runtime (above), plus the tools to create and deploy your own containers.

Installing Docker

Download and install directly from the Docker website, or your favorite package manager. Make sure to install the correct version for your system architecture (I’m looking at you, Apple ARM).

Check that it’s installed and available on your path.

$ docker version
Client: Docker Engine - Community
 Version:           20.10.21
 API version:       1.41
 Go version:        go1.19.3
 Git commit:        baeda1f82a
 Built:             Tue Oct 25 17:53:02 2022
 OS/Arch:           darwin/arm64
 Context:           default
 Experimental:      true

Server: Docker Desktop 4.15.0 (93002)
 Engine:
  Version:          20.10.21
  API version:      1.41 (minimum version 1.12)
  Go version:       go1.18.7
  Git commit:       3056208
  Built:            Tue Oct 25 17:59:41 2022
  OS/Arch:          linux/arm64
  Experimental:     false
 containerd:
  Version:          1.6.10
  GitCommit:        770bd0108c32f3fb5c73ae1264f7e503fe7b2661
 runc:
  Version:          1.1.4
  GitCommit:        v1.1.4-0-g5fd4c4d
 docker-init:
  Version:          0.19.0
  GitCommit:        de40ad0

Development Workflow

1. Create an image, which includes both your application and a configuration file describing how to run it.
2. Tell Docker to run this image in a container if you wish to run it locally.
3. Upload the image to the Docker registry, which allows someone else to download and run it on a different system.

Creating an Image

Let’s build a simple application, and then turn it into a Docker image. e.g.

fun main() {
  println("Hello Docker!")
}

$ kotlinc Hello.kt -include-runtime -d Hello.jar
$ java -jar Hello.jar 
Hello Docker!

To bundle this application, create a Dockerfile i.e. a configuration file for your image that describes how to execute it.

# Dockerfile
FROM openjdk:17
COPY Hello.jar /app
WORKDIR /app
CMD java -jar Hello.jar

You can find suitable Docker images on https://hub.docker.com. In this case, we’re using Temurin JDK as our base image (Linux/Java installation).

Package your application

To create the image:

$ docker build -t hello-docker .

To see the image that we’ve created:

$ docker images
REPOSITORY     TAG       IMAGE ID       CREATED         SIZE
hello-docker   latest    a615e715b56d   7 seconds ago   455MB

To run our image:

$ docker run hello-docker
Hello Docker!

Publish your image

To make this image available to other systems, you can publish it to the Docker Hub, and make it available to download. See Docker repos documentation for more details.

1. Create an account on Docker Hub if you haven’t already. Login.
2. Create a repository to hold your images.
3. Tag your local image with your username/repository.
4. Push your local image to that repository.

$ docker image ls
REPOSITORY     TAG       IMAGE ID       CREATED         SIZE
hello-docker   latest    f81c65fd07d3   3 minutes ago   455MB

$ docker tag f81c65fd07d3 jfavery/cs346

$ docker push jfavery/cs346:latest
The push refers to repository [docker.io/jfavery/cs346]
5f70bf18a086: Pushed
8768f51fa877: Pushed
5667ad7a3f9d: Pushed
6ea5779e9620: Pushed
fb4f3c9f2631: Pushed
12dae9600498: Pushed
latest: digest: sha256:6ddd868abde318f67fa50e372a47d4a04147d29722c4cd2a59c45b97a413ea22 size: 1578

Pull your image to a new machine

To pull (download) this image to a new machine, use docker pull.

$ docker pull jfavery/cs346
Using default tag: latest
latest: Pulling from jfavery/cs346
0509fae36eb0: Pull complete
6a8d9c230ad7: Pull complete
0dffb0eed171: Pull complete
77de63931da8: Pull complete
dc36babb139f: Pull complete
4f4fb700ef54: Pull complete
Digest: sha256:6ddd868abde318f67fa50e372a47d4a04147d29722c4cd2a59c45b97a413ea22
Status: Downloaded newer image for jfavery/cs346:latest
docker.io/jfavery/cs346:latest

$ docker run hello-docker

$ docker images
REPOSITORY      TAG       IMAGE ID       CREATED          SIZE
jfavery/cs346   latest    f81c65fd07d3   10 minutes ago   455MB

$ docker run jfavery/cs346
Hello Docker!

To run a long-running program (that doesn’t halt after execution), use the -d flag.

$ docker run -d jfavery/cs346

Persisting Data Across Containers

So what is happening is that when we launch a container, it creates a new environment from the image and sets up the container with its own mutable environment and data. This works great, until you stop the container - when you restart it, the environment is recreated and you lose any previous data!

How do we avoid this? We can create a volume on the host OS, outside of the scope of the container, and then provide the container access to the volume. For example, we can create a data file that will persist after container restarts.

# create a volume on the host
# we attach it at runtime below
$ docker volume create data-storage

# data-storage is the volume we created
# /data is a container directory that maps to the volume
$ docker run -v data-storage:/data jfavery/cs346

Managing Web Services

One common use case for container is as a way to deploy server applications, including web services. These have unique requirements compared to standard applications – namely the need to manage network requests that originate from outside of the container. Docker can handle this, with some additional configuration.

The following example is a Spring Boot application from the public repo: slides/containerization/docker-spring-boot. The application listens on port 8080, and manages get and post requests for messages:

class Message(id: String, text: String)

• post will store a message that is send in the post body.
• get will return and display a list of the messages that have been stored.

Here’s an example of a Dockerfile for this web service:

# Dockerfile
FROM openjdk:17
VOLUME /tmp
EXPOSE 8080
ARG JAR_FILE=target/spring-boot-docker.jar
ADD ${JAR_FILE} app.jar
ENTRYPOINT ["java","-jar","/app.jar"]

• FROM: the starting image (Linux + JVM)
• VOLUME: mapping an external volume
• EXPOSE: port that our application we will listen on
• ARG: passing in JAR_FILE arguments pointing to our application’s JAR file
• ADD: remap our JAR file to a local/internal JAR file that will be executed
• ENTRYPOINT: how to run the JAR file

To build the Docker image:

$ docker build -t docker-spring-boot .

[+] Building 1.9s (8/8) FINISHED
 => [internal] load build definition from Dockerfile
 => => transferring dockerfile: 69B
 => [internal] load .dockerignore
 => => transferring context: 2B
 => [internal] load metadata for docker.io/library/openjdk:17
 => [auth] library/openjdk:pull token for registry-1.docker.io
 => [internal] load build context
 => => transferring context: 57.63MB
 => CACHED [1/2] FROM docker.io/library/openjdk:18@sha256:9b448de897d211c9e0ec635a485650aed6e28d4eca1efbc34940560a480b3f1f
 => [2/2] ADD build/libs/docker-spring-server-1.0.jar app.jar
 => exporting to image
 => => exporting layers
 => => writing image sha256:2593c79e75b19b36dd2b0ee16fca23753578fb6381fb6d14f5c5e44fc0162bb4
 => => naming to docker.io/library/docker-sprint-boot  

When we run the container, we need to specify that we want to map port 8080 from the outside environment into the container. We can do this using the -p command-line option:

$ docker run -p 8080:8080 docker-spring-boot
  .   ____          _            __ _ _
 /\\ / ___'_ __ _ _(_)_ __  __ _ \ \ \ \
( ( )\___ | '_ | '_| | '_ \/ _` | \ \ \ \
 \\/  ___)| |_)| | | | | || (_| |  ) ) ) )
  '  |____| .__|_| |_|_| |_\__, | / / / /
 =========|_|==============|___/=/_/_/_/
 :: Spring Boot ::                (v2.7.4)

2023-03-26 16:31:11.453  INFO 1 --- [           main] com.example.demo.DemoApplicationKt       : Starting DemoApplicationKt using Java 17.0.2 on d2a3849df55b with PID 1 (/app.jar started by root in /)
2023-03-26 16:31:11.455  INFO 1 --- [           main] com.example.demo.DemoApplicationKt       : No active profile set, falling back to 1 default profile: "default"
...
...
...

Our web service is now running in a container! We can access the web service as-if it was running locally on port 8080.