I'm diving into Docker because I think it's a valuable skill for job hunting! I'm working on a small web app and I have a bunch of questions about how a realistic setup would be in the industry. Specifically, I'm curious about how Docker is utilized—does it play a role in development, or is it mainly just for production? What do I need to do for hot reloading and debugging? Also, do companies usually separate the frontend and backend? And how do they typically communicate? Is it a matter of targeting specific ports, or is that managed by something like Nginx?
3 Answers
In my team, we primarily avoid using Docker for development, unless it’s for testing dependencies such as Postgres or Redis. We create Docker images for the .NET apps during the build process and then deploy those images onto Kubernetes. I’d recommend using a dev server proxy (like what Vite offers) during development, and then setting up an Nginx container for production situations. This helps manage requests from the React app to the .NET backend effectively.
Using a tool like Aspire can simplify a lot of this for you. It kind of manages things like API requests through an Nginx reverse proxy for you. I generally prefer Vite for my setups; it has built-in proxy support and hot-reloading, so you can easily get started with minimal hassle. As for Docker, it's indeed used in development mainly for external dependencies like databases, while production is where you create the actual containers.
In our setup, we use Docker mainly for dependencies during development. For production, a possible workflow involves pushing code to the main branch and tagging it to build a Docker image, which is then pushed to our registry. From there, we deploy it to AWS ECS and route traffic through a load balancer to the right container port. We keep the frontend and backend separate to allow flexible scaling and avoid dependency issues. It's crucial to maintain a good environment setup without hardcoding configurations into the containers.
Thanks for breaking that down! I'm trying to wrap my head around your point about not embedding settings into containers for different environments. I thought having multiple images for each environment could be a good approach. Are you suggesting that I should use something like mounted volumes for environment-specific settings instead?