FutStats
A high level view of the solution is given beow.
The application is based on Phoenix Framework which internally uses Plug and Cowboy. So the app is a set of models, controllers and their respective views.
The API routes have been defined in the Router.ex file. Every incoming request to the server goes through a plug pipeline. The response is rendered according to the views associated with each controller. Views are defined in fut_stats_web/views while the controllers reside in fut_stats_web/controllers. The API is capable of sending JSON as well as protbuf responses.
The application uses ecto to communicate with a Postgres Database. Ecto supports schemas which directly map to database tables. The schema for the football data resides in fut_stats/models.
Before we can start seeding or querying the database we need to run the migrations. ecto.migrate maps a schema to a migration script which is executed to create a database table. The database is seeded using the seed.exs script. This process is done inside a separate ephemeral container called a seeder to keep the architecture simple.
Logging and Metrics are done using Elixir's native logger and the Instruments package. StatsD is used for the collection of the time series metrics data. In Application.ex, several important probes have been defined to measure Beam performance and resource usage. API response times and request rates are calculated by a custom plug Plugs.Metrics. Ecto query performance measured by a custom ecto logger Repo.Metrics which writes the query count and response times to the StatsD server.
Swagger has been used as the way to document the HTTP APIs and ex_doc has been used to make the documentation
The application comes bundled with a .Dockerfile and a docker-compose.yml file for easier deployments.
Make sure you have Docker and Docker Compose installed.
Initialize a Docker swarm on your machine if you haven't so already.
docker swarm init
(In case if the command prompts you for specifying an IP run docker-machine ls and input the default IP)
Then execute the following command to run a shell script. (Make sure it is executable, if not run chmod +x run.sh)
./run.sh
This script does the following
- delete any existing host volumes for the docker-compose.yml and create new ones
- delete any existing stack with the name
futstat - build the docker-compose.yml
- deploy a stack named
futstatusing the docker-compose.yml
These steps are only required the first time you are deploying the stack, in the subsequent turns you could just do the deploy stack step and it would work fine. And feel free to map the host ports to different one's if any of them are occupied by existing services.
You may have to wait for a few seconds until all the containers have successfully started.(The seeder service seeds the database with the test data)
- The FutStats API would be live at
http://<Host-Address>/api/ - Graphite console at
http://<Host-Address>:3100 - Grafana console at
http://<Host-Address>:3000 - Swagger API console at
http://<Host-Address>/swagger/
Host-Address could be localhost if running locally or an external IP if running in a cloud VM.