Awesome-Compose: Application samples for project development kickoff

Software systems have become quite complex nowadays. A system may consist of several distributed services, each one providing a specific functionality and being updated independently. Starting the development of a project of such complexity is sometimes time consuming, in particular when you are not already familiar with the software stack you are going to work on. This may be because, most of the time, we need to follow rigorous steps to put together the entire project and if we make a mistake in-between, we may have to start all over again.

As a developer, getting a quick understanding of how all the stack is wired together and having an easy to manage project skeleton may be a very good incentive to use that particular stack for future projects. 

Furthermore, there are plenty of open-source software stacks that developers can use for their own setups. Providing them with a straightforward way to deploy these software stacks in order to check them out is an important matter when looking to simplify software development and allow developers to explore different options. 

To tackle this, we have put together a Github repository with application samples that can be easily deployed with Docker Compose. The repository name is awesome-compose and it contains a curated list of Compose application samples which can provide a good starting point for how to integrate different services using a Compose file and to manage their deployment with Docker Compose. 

The awesome-compose repository was created to provide a quick and simple way for developers to experience Compose-based setups with different software stacks.  Moreover, we hope that you are willing to share your best examples of compose files for different stacks that are not already in the repository or to improve the current samples for everybody to use.

The setups provided currently in the awesome-compose repository fit mainly in two categories:

Application skeletons: useful for kicking off project development. We can find different application skeletons with multiple services already wired together and ready to be launched with docker-compose;Setups with different open-source software stacks. These are not production ready, they are intended mostly for personal/home use or simply for developers to get familiar with them within their local dev environment. 

All samples in the repository are provided as they are, everybody can customize each sample according to their needs. 

We will discuss further each category and what are the benefits they can provide.

Kickoff a project with an application sample

To be able to run the samples from the repository, make sure you have already installed:

Windows or macOS: Install Docker DesktopLinux: Install Docker and then Docker Compose

Then, either git clone or download one or more samples from the awesome-compose repository.

$ git clone https://github.com/docker/awesome-compose.git$ cd awesome-compose

At the root of each sample there is the docker-compose.yml. It contains the definition and the structure of the application and instructions on how to wire the components of the application.

Identify the sample that matches your requirements and open its directory. Sample’s directory names follow a very simple pattern consisting of component names separated by ‘-’. This allows us to quickly identify the sample we need for our project. For this exercise, let us use the nginx-flask-mysql sample.

There are a few simple steps to follow to get the application skeleton up and running and be able to modify it.

Deploy the application sample

Open the sample directory and run it with docker-compose:

$ cd nginx-flask-mysql$ docker-compose up -dCreating volume “nginx-flask-mysql_db-data” with default driverBuilding backendStep 1/8 : FROM python:3.8-alpine3.8-alpine: Pulling from library/python……Creating nginx-flask-mysql_db_1      … doneCreating nginx-flask-mysql_proxy_1   … doneCreating nginx-flask-mysql_backend_1 … done

Check there are three containers running, one for each service:

$ docker-compose psName                          Command        State        Ports—————————————————————————————————————————————————————————————————————————————————————————————nginx-flask-mysql_backend_1   /bin/sh -c flask run –hos …   Up   0.0.0.0:5000->5000/tcp
nginx-flask-mysql_db_1        docker-entrypoint.sh –def …   Up   3306/tcp, 33060/tcp
nginx-flask-mysql_proxy_1     nginx -g daemon off;          Up   0.0.0.0:80->80/tcp

Query port 80 of the proxy container with curl or in a web browser to have the backend pass the data from the DB:

$ curl localhost:80<div>Blog post #1</div><div>Blog post #2</div><div>Blog post #3</div><div>Blog post #4</div>

Modify and update the application sample

Let us assume that we have to change the application server, in this case being the backend service which is implemented in python using the Flask framework. The method returning the message we queried previously can be found below.

@server.route(‘/’)def listBlog():   global conn   if not conn:       conn = DBManager(password_file=‘/run/secrets/db-password’)       conn.populate_db()   rec = conn.query_titles()  response = ”  for c in rec:
    response = response + ‘<div> ‘ + c + ‘</div>’
  return response

Assume we change this method to remove the html tags:

@server.route(‘/’)def listBlog():  …  for c in rec:
    response = response + ‘ ‘ + c + ‘ ‘
  return response

As all our containers are already running, the logical workflow would be to stop the backend service, rebuild its image and re-run to get our change in. Doing this would be very inefficient during development.

We could in turn, hack the docker-compose file and add under the backend service the following setting:

backend:    build: backend    restart: always    volumes:
    – ./backend:/code    …

This would instruct Docker Compose to mount the backend source code to the container path from where it is being executed on container start.

Now, all we have to do is to restart the backend to have the changed code executed.

$ docker-compose restart backendRestarting nginx-flask-mysql_backend_1 … done

Querying again the proxy, we can observe the change:

$ curl localhost:80Blog post #1 Blog post #2 Blog post #3 Blog post #4

Cleanup deployment  and data

To remove all containers run:

$ docker-compose downStopping nginx-flask-mysql_backend_1 … doneStopping nginx-flask-mysql_db_1 … doneStopping nginx-flask-mysql_proxy_1 … doneRemoving nginx-flask-mysql_backend_1 … doneRemoving nginx-flask-mysql_db_1 … doneRemoving nginx-flask-mysql_proxy_1 … doneRemoving network nginx-flask-mysql_backnetRemoving network nginx-flask-mysql_frontnet

Adding the -v parameter to the down command ensures all data hosted by the db service is being deleted:

$ docker-compose down -v…Removing volume nginx-flask-mysql_db-data

To conclude this part, the samples provided in the awesome-compose repository may help developers to put together all the components for their project in a matter of minutes. This is most beneficial in particular for beginners in the development with containerized applications that can be managed with docker-compose.

Setups for different software stacks

The second type of samples that awesome-compose repository contains are compose files for setting up different platforms such as Nextcloud, WordPress, Gitea etc. These samples consist mostly of a Compose file defining a basic setup for each of the components. The purpose for these is to provide developers an easier introduction to different software stacks they could take a quick view at what they offer and tinker with. 

Let us consider the Nextcloud setup for the next exercise. Nextcloud is an open source file sharing platform that anyone can install for their private use. The setups in the awesome-compose repository are pieced together according to the instructions on the Nextcloud’s official image page in Docker Hub.

To deploy it, select the directory of the nextcloud sample you prefer:

$ cd nextcloud-postgres/$ lsdocker-compose.yaml  README.md

And run it with docker compose:

$ docker-compose up -dCreating network “nextcloud-postgres_default” with the default driver
Creating volume “nextcloud-postgres_db_data” with default driver
Creating volume “nextcloud-postgres_nc_data” with default driverPulling nc (nextcloud:apache)…
apache: Pulling from library/nextcloud…Creating nextcloud-postgres_nc_1 … doneCreating nextcloud-postgres_db_1 … done

Check that containers are running:

$ docker-compose psName                      Command                         State     Ports—————————————————————————————————————————————————————————————————————————————————————–nextcloud-postgres_db_1   docker-entrypoint.sh postgres   Up        5432/tcp
nextcloud-postgres_nc_1   /entrypoint.sh apache2-for …    Up        0.0.0.0:80->80/tcp$ docker psCONTAINER ID   IMAGE              COMMAND                 CREATED   STATUS              PORTS                NAMESa1381bcf5b1c   nextcloud:apache   “/entrypoint.sh apac…”  14 minutes ago   Up About a minute   0.0.0.0:80->80/tcp   nextcloud-postgres_nc_1ec66a5aff8ac  postgres:alpine “docker-entrypoint.s…”    14 minutes ago    Up About a minute   5432/tcp             nextcloud-postgres_db_1

In only a few minutes (depends on the internet connection) we get a nextcloud platform up and running on our local machine. Opening a browser window and going to localhost:80 should land us on the Nextcloud’s initialization page:

Similarly to the first exercise, to remove all containers run: 

$ docker-compose downStopping nextcloud-postgres_nc_1 … doneStopping nextcloud-postgres_db_1 … doneRemoving nextcloud-postgres_nc_1 … doneRemoving nextcloud-postgres_db_1 … doneRemoving network nextcloud-postgres_default

Use the -v parameter to delete the volumes where the nextcloud data gets stored. Identical steps should be followed for all other samples of useful software stacks.

Summary

Lowering the barrier in deploying different software stacks, enables more and more developers to have a look at them and potentially use them for their projects.

The awesome-compose repository was created with the purpose of aggregating compose files and application samples that may be useful to anyone interested in containerized applications.

Call for contributions

All developers already familiar with compose and that have already set up interesting stacks that may be useful to others, are highly encouraged to add them to the repository and share them.  Improvements on the current samples are also very much appreciated!
The post Awesome-Compose: Application samples for project development kickoff appeared first on Docker Blog.
Quelle: https://blog.docker.com/feed/

Published by