Temperature sensors simulation - Client/Server (Kafka)

This is a simple client/server application that demonstrates a publisher/subscriber pattern to send a message to connected clients.

Used Kafka as a message broker and cppkafka client library to communicate with it.

The server is a simple C++ application that uses cppkafka library to communicate with Kafka. It is a simple publisher that sends a message to a topic.

The client(s) are simple subscribers that receive a message from a topic that they're subscribed to.

Architectural Decision Records (ADRs)

• vspkg for dependency management.
  • Used vcpkg for dependency management to keep package management simple and consistent across platforms.
• kafka as a message broker.
  • Kafka is used for building real-time data pipelines and streaming apps. It is scalable, fast and fault-tolerant. [1]
  • Maybe the straight forward solution would be ASIO sockets with boost or something similar, but I wanted to try Kafka with C++ seemed like a good fit for this task and a new challenge (I only used Python with ELK+Kafka before).
  • Can be easily attached to ELK stack for logging and monitoring purposes, with Kibana as a dashboard.
• cppkafka as a C++ client library for Kafka.
  • Found another client librdkafka but it was a bit more complicated to use.
  • cppkafka is basically a wrapper around librdkafka but with a simpler API, and it utilizes modern C++ features.
• docker-compose used to easily manage the containers and the environment variables, the kafka image is from wurstmeister and I only build the project image once and use it with the client and server services with different environment variables.

How to run

Docker is required to run the application, at least to run kafka and zookeeper.

Run Kafka and Zookeeper

Run the provided bash and bat scripts (depends on your OS) to run Kafka and Zookeeper and create the relevant topics. docker_script.bash or docker_script.bat

Running the server and client(s)

To run locally

To run the server and client locally you'd have to have vcpgk installed (on Windows Visual Studio is required to build) and run the following commands:

cmake . -DCMAKE_TOOLCHAIN_FILE={path_to_vcpkg}/scripts/buildsystems/vcpkg.cmake && cmake --build .

Then run the server and client executables.

To run with docker

The server and the client containers are built from the same image, only environment variables are different.

To run the server:

docker-compose run server

To run a client:

docker-compose run client

To do

• Add tests.
  • Checkout gtest or boost for tests.
• Handle client group id dynamically (group id being hard coded causes some conflicts when running multiple clients).