Most of my experience with web applications is with monoliths deployed to a PaaS or using configuration management tools to traditional servers. Last week, I submerged myself in a different paradigm, the one of microservices. In this post, I’m going to share what I learned by deploying a Go application on top of Kubernetes.
To follow along, you’ll need to have Go, Docker, and Kubernetes installed in your system. I’d recommend using something like K3d and K3s to install Kubernetes on your machine if you don’t have access to one already.
The hello server
First, we’ll start by writing an elementary microservice. Let’s create a web server that responds with Hello and whatever we passed as the path of the URL.
In a clean directory, we initialize and tidy Go. You can replace my GitHub username, mauromorales
, for yours.
go mod init github.com/mauromorales/hello-server
go mod tidy
Create the file main.go
which will describe our microservice.
//A simple web server that responds with "Hello " and whatever you pass as the
//path of the URL. It also logs the requests
package main
import (
"fmt"
"log"
"net/http"
)
func Log(handler http.Handler) http.Handler {
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
log.Printf("%s %s %s", r.RemoteAddr, r.Method, r.URL)
handler.ServeHTTP(w, r)
})
}
func main() {
http.HandleFunc("/", HelloServer)
http.ListenAndServe(":8080", Log(http.DefaultServeMux))
}
func HelloServer(w http.ResponseWriter, r *http.Request) {
fmt.Fprintf(w, "Hello, %s!\n", r.URL.Path[1:])
}
To test everything is working properly, run the microservice with
go run main.go
and on a different terminal make a web request
% curl http://localhost:8080/Go
Hello, Go!
And you should also see a log for the request on the terminal running the hello-server
% go run main.go
2022/12/30 11:42:42 127.0.0.1:59068 GET /Go
A container version
Kubernetes might be written in Go, but it wouldn’t understand how to deal with our hello-server. The minimal unit of processing in K8s is the container, so let’s put our code into one. In a file called Dockerfile
add the following content:
FROM golang as builder
WORKDIR /app/
COPY . .
RUN CGO_ENABLED=0 go build -o hello-server /app/main.go
FROM alpine:latest
WORKDIR /app
COPY --from=builder /app/ /app/
EXPOSE 8080
CMD ./hello-server
Let’s first build an image
docker build -t docker.io/mauromorales/hello-server:0.1.1 .
When it’s done building, it will show up as one of your images
% docker images
REPOSITORY TAG IMAGE ID CREATED SIZE
mauromorales/hello-server 0.1.1 3960783c4afe 36 seconds ago 19.8MB
So let’s run it
docker run --rm -p 8080:8080 mauromorales/hello-server:0.1.1
And in a different terminal test that it’s still working as expected
% curl http://localhost:8080/Docker
Hello, Docker!
Looking back at the running container, we see that again our request was logged
% docker run --rm -p 8080:8080 mauromorales/hello-server:0.1.1
2022/12/30 10:48:57 172.17.0.1:58986 GET /Docker
Deploying hello-server to Kubernetes
Let’s begin by uploading the image we built in the last step, to the docker hub. For which you need an account.
docker login --username mauromorales
And once logged in, you can push the image
docker push mauromorales/hello-server:0.1.1
This process will be in iterations to understand the different components in K8S
Pods
Initially, we will only deploy a pod (a grouping mechanism for containers) of one single container. To achieve this, we create a file called pod.yaml
add the following definition:
apiVersion: v1
kind: Pod
metadata:
name: hello-server
labels:
app: hello-server
spec:
containers:
- name: hello-server
image: mauromorales/hello-server:0.1.1
imagePullPolicy: Always
ports:
- name: http
containerPort: 5000
protocol: TCP
And apply it:
% kubectl apply -f pod.yaml
pod/hello-server created
You should now see it listed:
% kubectl get pods
NAME READY STATUS RESTARTS AGE
hello-server 1/1 Running 0 111s
While the pod is running in the background, we need to forward the port to access it:
kubectl port-forward pod/hello-server 8080:8080
Now you can test again that it is working, by curling to it from a different terminal.
% curl http://localhost:8080/Pod
Hello, Pod!
But if you go back to the port forwarding, you will not see any logs. All you see are the logs of the port-forward
command.
% kubectl port-forward pod/hello-server 8080:8080
Forwarding from 127.0.0.1:8080 -> 8080
Forwarding from [::1]:8080 -> 8080
Handling connection for 8080
To read the pod logs, we require kubectl
.
% kubectl logs pod/hello-server
2022/12/30 10:59:56 127.0.0.1:51866 GET /Pod
Services
So far so good, but unfortunately a pod is not really implementing the microservice pattern. If the pod is restarted, it might lose its IP. For our little example this is not a problem, but if we were to have more than one microservice talking to each other, we would need to find a way to share the new IPs between the different microservices. Thankfully, Kubernetes comes with a solution to this issue, services.
Let’s write one inside service.yaml
apiVersion: v1
kind: Service
metadata:
name: hello-server-svc
spec:
selector:
app: hello-server
ports:
- protocol: TCP
port: 8080
targetPort: 8080
Now, apply the service:
% kubectl apply -f service.yaml
service/hello-server-svc created
And as usual, we do some port forwarding, but this time to the service instead of the pod:
kubectl port-forward service/hello-server-svc 8080:8080
Let’s test it in the second terminal
% curl http://localhost:8080/Service
Hello, Service!
And look at the service logs
% kubectl logs service/hello-server-svc
2022/12/30 11:15:00 127.0.0.1:39346 GET /Service
Deployments
This is looking much better now, if I wanted to scale this service, all I’d need to do is to create another pod, with the hello-server
label. But this would be very tedious and error-prone. Thankfully, Kubernetes gives us deployments, which handle that for us and gives us deployment strategies. Let us then create a deployment with three replicas.
First, we need to delete the pod we created.
% kubectl delete pod/hello-server
pod "hello-server" deleted
And in a file called deployment.yaml
add the following description:
apiVersion: apps/v1
kind: Deployment
metadata:
name: hello-server
labels:
app: hello-server
spec:
replicas: 3
selector:
matchLabels:
app: hello-server
template:
metadata:
labels:
app: hello-server
spec:
containers:
- name: hello-server
imagePullPolicy: Always
image: mauromorales/hello-server:0.1.1
ports:
- containerPort: 8080
and apply it
kubectl apply -f deployment.yaml
When finished you should be able to list the generated pods
% kubectl get pods
NAME READY STATUS RESTARTS AGE
hello-server 1/1 Running 0 7m13s
hello-server-5c7c6f798f-hp99p 1/1 Running 0 13s
hello-server-5c7c6f798f-f2b4c 1/1 Running 0 13s
hello-server-5c7c6f798f-2fxdm 1/1 Running 0 13s
The first pod, is the one we created manually, and the next three are the ones the deployment created for us.
We start forwarding traffic to the deployment:
kubectl port-forward service/hello-server-svc 8080:8080
And test it out
% curl http://localhost:8080/Deployment
Hello, Deployment!
Let us also check the logs:
% kubectl logs deployment/hello-server
Found 3 pods, using pod/hello-server-5c7c6f798f-hp99p
2022/12/30 11:29:47 127.0.0.1:46420 GET /Deployment
Port forwarding is nice, but at its current state will only map to one replica, which is less than ideal. In order to load-balance our service, we need to add an ingress rule.
Create a file ingres.yaml
with the following content
apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
name: hello-server
annotations:
ingress.kubernetes.io/rewrite-target: /
kubernetes.io/ingress.class: traefik
spec:
rules:
- http:
paths:
- path: /
pathType: Prefix
backend:
service:
name: hello-server-svc
port:
number: 8080
And apply it
% kubectl apply -f ingress.yaml
ingress.networking.k8s.io/hello-server created
As you probably expect it, we need to forward traffic, however this time instead of forwarding to our service, we forward to the traefik service (served on port 80):
kubectl port-forward -n kube-system service/traefik 8080:80
Let’s test it out by sending 20 requests this time
% for i in `seq 1 20`; do curl http://localhost:8080/Traeffic; done
...
Hello, Traeffic!
And have a look at the logs
kubectl logs deployment/hello-server
Found 3 pods, using pod/hello-server-5c7c6f798f-hp99p
2022/12/30 11:43:14 10.42.0.22:43070 GET /Traeffic
2022/12/30 11:43:14 10.42.0.22:43070 GET /Traeffic
2022/12/30 11:43:14 10.42.0.22:43070 GET /Traeffic
2022/12/30 11:43:14 10.42.0.22:43070 GET /Traeffic
2022/12/30 11:43:14 10.42.0.22:43070 GET /Traeffic
2022/12/30 11:43:14 10.42.0.22:43070 GET /Traeffic
2022/12/30 11:43:14 10.42.0.22:43070 GET /Traeffic
And we can also check the individual logs of each pod
% kubectl logs pod/hello-server-5c7c6f798f-hp99p --since=4m
2022/12/30 11:43:14 10.42.0.22:43070 GET /Traeffic
2022/12/30 11:43:14 10.42.0.22:43070 GET /Traeffic
2022/12/30 11:43:14 10.42.0.22:43070 GET /Traeffic
2022/12/30 11:43:14 10.42.0.22:43070 GET /Traeffic
2022/12/30 11:43:14 10.42.0.22:43070 GET /Traeffic
2022/12/30 11:43:14 10.42.0.22:43070 GET /Traeffic
2022/12/30 11:43:14 10.42.0.22:43070 GET /Traeffic
seneca% kubectl logs pod/hello-server-5c7c6f798f-f2b4c --since=4m
2022/12/30 11:43:14 10.42.0.22:59596 GET /Traeffic
2022/12/30 11:43:14 10.42.0.22:59596 GET /Traeffic
2022/12/30 11:43:14 10.42.0.22:59596 GET /Traeffic
2022/12/30 11:43:14 10.42.0.22:59596 GET /Traeffic
2022/12/30 11:43:14 10.42.0.22:59596 GET /Traeffic
2022/12/30 11:43:14 10.42.0.22:59596 GET /Traeffic
seneca% kubectl logs pod/hello-server-5c7c6f798f-2fxdm --since=4m
2022/12/30 11:43:14 10.42.0.22:40840 GET /Traeffic
2022/12/30 11:43:14 10.42.0.22:40840 GET /Traeffic
2022/12/30 11:43:14 10.42.0.22:40840 GET /Traeffic
2022/12/30 11:43:14 10.42.0.22:40840 GET /Traeffic
2022/12/30 11:43:14 10.42.0.22:40840 GET /Traeffic
2022/12/30 11:43:14 10.42.0.22:40840 GET /Traeffic
2022/12/30 11:43:14 10.42.0.22:40840 GET /Traeffic
And voilà, you can see how it was properly distributed between our 3 replicas.
I hope you had as much fun as I did playing with Kubernetes.