I have two Java microservices (caller.jar which calls called.jar)
We can set the caller service http port through an env-var CALLERPORT and the address of called service through an env-var CALLEDADDRESS. So caller uses two env-var.
We must set also called service env-var CALLEDPORT in order to set the specific http port on which called service is listening for http requests.
I don't know exactly how to simply expose these variables from a Dockerfile, in order to set them using Kubernetes.
Here is how I made the two Dockerfiles:
Dockerfile of caller
FROM openjdk:8-jdk-alpine
# ENV CALLERPORT (it's own port)
# ENV CALLEDADDRESS (the other service address)
ADD caller.jar /
CMD ["java", "-jar", "caller.jar"]
Dockerfile of called
FROM openjdk:8-jdk-alpine
# ENV CALLEDPORT (it's own port)
ADD called.jar /
CMD ["java", "-jar", "called.jar"]
With these I've made two Docker images:
Then I've made the two deployments.yaml in order to let K8s deploy (on minikube) the two microservices using replicas and loadbalancers.
deployment-caller.yaml
apiVersion: v1
kind: Service
metadata:
name: caller-loadbalancer
spec:
type: LoadBalancer
ports:
- port: 8080
targetPort: 8080
selector:
app: caller
---
apiVersion: extensions/v1beta1
kind: Deployment
metadata:
name: caller
labels:
app: caller
spec:
replicas: 2
minReadySeconds: 15
strategy:
type: RollingUpdate
rollingUpdate:
maxUnavailable: 1
maxSurge: 1
selector:
matchLabels:
app: caller
tier: caller
strategy:
type: Recreate
template:
metadata:
labels:
app: caller
tier: caller
spec:
containers:
- image: myaccount/caller
name: caller
env:
- name: CALLERPORT
value: "8080"
- name: CALLEDADDRESS
value: called-loadbalancer # WHAT TO PUT HERE?!
ports:
- containerPort: 8080
name: caller
And deployment-called.yaml
apiVersion: v1
kind: Service
metadata:
name: called-loadbalancer
spec:
type: LoadBalancer
ports:
- port: 8081
targetPort: 8081
selector:
app: called
---
apiVersion: extensions/v1beta1
kind: Deployment
metadata:
name: called
labels:
app: called
spec:
replicas: 2
minReadySeconds: 15
strategy:
type: RollingUpdate
rollingUpdate:
maxUnavailable: 1
maxSurge: 1
selector:
matchLabels:
app: called
tier: called
strategy:
type: Recreate
template:
metadata:
labels:
app: called
tier: called
spec:
containers:
- image: myaccount/called
name: called
env:
- name: CALLEDPORT
value: "8081"
ports:
- containerPort: 8081
name: called
IMPORTANT: The single services work well if called singularly (such as calling an healthcheck endpoint) but, when calling the endpoint which involves the communication between the two services, then there is this error:
java.net.UnknownHostException: called
The pods are correctly running and active, but i guess the problem is the part of deployment.yaml in which I must define how to find the pointed service, so here:
spec:
containers:
- image: myaccount/caller
name: caller
env:
- name: CALLERPORT
value: "8080"
- name: CALLEDADDRESS
value: called-loadbalancer # WHAT TO PUT HERE?!
ports:
- containerPort: 8080
name: caller
Neither
called
nor
called-loadbalancer
nor
http://caller
kubectl get pods,svc -o wide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
pod/called-855cc4d89b-4gf97 1/1 Running 0 3m23s 172.17.0.4 minikube <none> <none>
pod/called-855cc4d89b-6268l 1/1 Running 0 3m23s 172.17.0.5 minikube <none> <none>
pod/caller-696956867b-9n7zc 1/1 Running 0 106s 172.17.0.6 minikube <none> <none>
pod/caller-696956867b-djwsn 1/1 Running 0 106s 172.17.0.7 minikube <none> <none>
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE SELECTOR
service/called-loadbalancer LoadBalancer 10.99.14.91 <pending> 8081:30161/TCP 171m app=called
service/caller-loadbalancer LoadBalancer 10.107.9.108 <pending> 8080:30078/TCP 65m app=caller
service/kubernetes ClusterIP 10.96.0.1 <none> 443/TCP 177m <none>
works if put in that line of the deployment.yaml. So what to put in this line?
To access services inside Kubernetes you should use this DNS: