Pod publishing messages with KafkaJs to Strimzi keeps crashing-CodePudding

This is my KafkaJs based publisher client. I created a container image and submitted a Pod YAML to Strimzi broker.

const { Kafka } = require('kafkajs')

async function clients() {
const kafka = new Kafka({
  clientId: 'my-app',
  brokers: ['test-kafka-bootstrap.strimzi.svc.cluster.local:9092']
})

const producer = kafka.producer()

await producer.connect()
await producer.send({
  topic: 'clients',
  messages: [
    { value: 'Hello KafkaJS user!' },
  ],
})

await producer.disconnect()
}

clients()

My Dockerfile.

FROM node:14 as build_app

WORKDIR /

WORKDIR /app
COPY app .
COPY package.json .
COPY package-lock.json .
RUN npm i
CMD ["node", "index.js"]

My pod YAML.

apiVersion: v1
kind: Pod
metadata:
  name: clients
spec:
  containers:
  - name: clients
    image: ghcr.io/org/clients:v0.0.0
    imagePullPolicy: Always

The Pod keeps crashing and kubectl logs does not show anything - its empty. A kubectl describe didnt' reveal anything either.

What am I missing?

CodePudding user response：

The logs show nothing because 1) Your code has no log output 2) Your code only send a single record, and then stops, so the container exits cleanly; it is not a long-running service. You'd have the same issues with a simple docker run command.

If you wanted a long-running service, you'd need to wrap your code in a web-service (NextJS, Express, Hapi, etc), and then add a health check probe for the k8s Service.

CodePudding user response：

you can bring up the pod in sleep mode and from there you can debug, use below yaml

apiVersion: v1
kind: Pod
metadata:
  name: clients
spec:
  containers:
  - name: clients
    image: ghcr.io/org/clients:v0.0.0
    imagePullPolicy: Always
    command: ["sleep" , "100000000"]

Now exec into the pod

kubectl exec -it <pod> -- bash

now execute

node index.js