Developing and testing APIs locally connected to a database is no joke. The database often becomes a pain point. However, using Docker, the process becomes easier and much simpler, making replication easy. In this blog, we will see how we can Dockerize a Golang API with a MySQL Database, also making it Docker Compose ready
For the demo, I created this RESTful Golang API. We can perform CRUD operations like, create, delete, and edit schedules to MySQL Database. We can learn more about the endpoints, methods, etc., in the project README. We will not go too deep into how the API works, as our primary goal is to focus on the Dockerization part.
For Dockerzing an app we need to create a Dockerfile. Let me tell you there are 100s of ways to write a Dockerfile, and there is nothing wrong or right, every individual/company has its own set of practices and way of writing. In our case, we will follow four of the best practices in our Dockerfile to get a better and optimized image that is smaller and more secure. Let's understand all 4 practices and why we implement them before jumping into writing a Dockerfile.
Using lighter base images: For almost every language there is a lighter version of the image. By lighter I don't mean a couple of Megabytes smaller, it can be 10x smaller, the reason is lighter is that it doesn't contain unnecessary dependencies which makes it smaller and more secure. Yes, more dependencies come with more security risk. we will have a
bullseyeandalpineversion for the node, Golang, etc.Multi-Stage Builds: They are one of the superpowers of Docker, they allow us to run the build steps in parallel and also let us create a final image by copying necessary files and things from other steps and having only items that are needed to run our program.
Create a Binary: Many languages support creating binary out of source code, making it small and much easier to run because we don't need to handle complete source code. Also, we can run in any environment irrespective of language barrier.
Breaking down layers: As we know every instruction in a Dockerfile is a layer, and breaking down layers is a good way to make our build faster. For example, if we are copying all the files (along with the dependencies file to install) from the source and then installing the dependencies, even if we didn’t make any changes to the dependencies, every time we rebuild an image, it will copy all the files and install dependencies. To overcome this, we break them into several layers and can copy the dependency file in one step and install it. And in the next step, we can copy all the files. Now, when we make a change to the code, and rebuild this time only the layer where we are copying all the files will be rebuilt, leaving the dependency step (which is cached as there are no changes). We can see an example in the below Dockerfile as well. We write Dockerfile in such a way that things that step that change less, like base image, dependencies, etc., will be on the top-to-bottom approach.
So, here is the Dockerfile we created for our Go API.
# Build Stage
FROM golang:alpine3.20 AS builder
WORKDIR /build
COPY go.mod go.sum ./
RUN go mod download
COPY . .
RUN go build -o /app .
# Final Stage
FROM alpine:3.20
COPY --from=builder /app /app
CMD ["/app"]
In the FROM we can see we used a golang:alpine version as the base image instead of using a full-blown golang one and naming the steps builder, the name/label will help us in copying the files from one stage to another. After that, we created a work directory. Then instead of copying all the files together we just copied the go.mod and go.sum and install dependencies (I have explained the reason for that above in the Breaking down layers point).
Now once the dependencies are installed we copy the remaining files. Then we create a binary from our source code by running go build and naming the binary with the -o output flag to the current location.
Now, here things get interesting, in the final stage we don't need a golang image or such, we can use an Alpine base image because we now have a binary, and we can run on any Linux system irrespective of programming language specifics. In the next step, we can copy the binary from the builder step to the final stage and run the binary.
That's it. That's how Dockerzie our app and we can further improve the Dockerfile by introducing the best practices like, creating a user and running as a non-root, etc. Now we can build an image with the Dockerfile and run and then connect to the remote or local MySQL server by providing the credentials it needs and then hit those API endpoints.
But, we will not stop here we will take a step further, and we will also run a MySQL server in a container and connect with our app. But, one point to note here is we can run a spin of a MySQL container and connect our API container to that, but there is so much manual work and long commands to type in the terminal, and things can go wrong. To overcome this we will use Docker Compose instead to make our life easier.
Let's create a file called compose.yml and use the blow config.
services:
app:
container_name: go-api
build:
context: .
dockerfile: Dockerfile
image: go-api
ports:
- 8080:8080
environment:
- DB_HOST=mysql
- DB_PORT=3306
- DB_USER=user
- DB_PASSWORD=password
- DB_NAME=my-database
depends_on:
mysql:
condition: service_healthy
networks:
- go-network
mysql:
container_name: go-mysql
image: mysql:9.0
environment:
- MYSQL_ROOT_PASSWORD=password
- MYSQL_USER=user
- MYSQL_PASSWORD=password
volumes:
- dbdata:/var/lib/mysql
networks:
- go-network
healthcheck:
test: ["CMD", "mysqladmin", "ping", "-h", "localhost"]
interval: 10s
timeout: 5s
retries: 3
volumes:
dbdata:
networks:
go-network:
driver: bridge
It's a pretty basic configuration, but a couple of key things I want to mention is that if we see DB_HOST is mysql, there is no localhost or ip because in compose services communicate with other services by the service name. This is out-of-the-box networking provided by Docker Compose.
Another point, it often happens when working with Docker Compose where we have two services: an app and a database, and the app service starts first and crashes because the database isn't ready when the app tries to connect. To overcome this, we set up a healthcheck for the database to confirm its readiness. Then, in our app service, we can use depends_on with a condition to ensure the app only starts when the database service is healthy.
Now when we do Docker compose for 1st time, we might encounter an error saying permission denied because it doesn't have permission as well as a database with the name my_database, so we need to both by execing into the container.
Even though our app has crashed the DB is still up and running. We can check by doing docker ps.
Now exec into the container by doing docker exec -it <container-id> sh. Container ID can be copied from the output after executing docker ps. Once we exec into the container, now log into the mysql by the below command:
mysql -u root -p
It will ask for the password, enter the password you mentioned in the compose.yml file. Once we log in, we can create a database. Create a database with the same name specified in the compose file. In my case, it's my_database. Execute the below command:
CREATE DATABASE my_database;
Now to give the right privileges and flush it execute the below command.
GRANT ALL PRIVILEGES ON my_database.* TO 'user'@'%';
FLUSH PRIVILEGES;
Once we are done, we need to stop the running compose service and restart again by doing docker compose up
That's it for this blog. I'm glad you're still reading and made it to the end—thank you so much for your support and reading. I sometimes share tips on Golang on Twitter. You can connect with me there.
Top comments (13)
I love these kind of posts!
Thank you so much, Francesco!
you know I love this sort of stuff for real ahah. I am planning another one but it's still in the process
I KNOW! I can't wait to hear more about it!
Good one, @pradumnasaraf brother. I am also starting with Golang these days. Really helpful. 🤍
Thnak you for reading, Shrijal
Thank for sharing
Thank you for reading, Mohamed
As I'm learning Docker this helped me alot💖
Thnak you for reading, Luv
Good job!!!
Thnak you for reading, Felipe
For go, since everything is compiled into the binary, you should consider using distroless images as your base.
They're smaller than Alpine, but should have all the necessary files (like timezone files, and ssl certs) that you should need.
Just change:
To:
That should save you a bit of space on your final image without sacrificing functionality.
You could also, if you really want to strip it down, do
FROM scratch, and copy over SSL certs and timezone files from alpine. That should get you what you need for full functionality, and an absolute minimum image size. But, there might be some other OS dependencies that aren't obvious, which is why I usually use distroless, just in case.Here's another really good approach that also uses
FROM scratch.