presto - moving worker from docker to kubernetes

It looks like you've combined parts of a Deployment and a Service; they're two different objects. You can break this up:

---apiVersion: apps/v1 kind: Deploymentmetadata:  name: presto-workerspec:  selector:    matchLabels:      app: presto-worker  replicas: 2   template:    metadata:      labels:        app: presto-worker    spec:      containers:      - name: presto-image        image: docker.io/mystuff/presto-image:latest        ports:        - containerPort: 8080---apiVersion: v1 kind: Servicemetadata:  name: presto-workerspec:  type: NodePort  selector:    matchLabels:      app: presto-worker  ports:  - name: http    port: 8080

The Service selector points at specific pods; it needs to match the deployment spec's pod template labels. The Deployment selector names the pods that the Deployment manages. In both cases they point at pods but they're for different purposes.

Here's a working Helm chart (i.e., template package of k8s resources) for presto: https://github.com/helm/charts/tree/master/stable/presto.

Here's basic design of a presto cluster in k8s from above Chart:

1. single coordinator. discoverable (by workers) over http://:8080 - this is where you need a k8s service exposed for coordinator so that all workers can perform discovery against this coordinator (actually it's the discover service exposed by coordinator).
2. a number of stateless workers via a k8s deployment. You do not need to expose any ports nor k8s service since no external process will reach to worker process at all.
3. uses k8s configMap to inject customized presto configuration. the most important config for a presto cluster is discovery-server.enabled=true in <presto-home>/etc/config.properties of your presto coordinator or your coordinator is not discoverable at all. (a.k.a., can not have external worker process over network).

According to this question, you need to make sure presto coordinator is accessible by worker process via DNS name like http://my-presto-coordinator:8080.

This is what I got from chart stable/presto by running helm template . (render all template in stdout). you will need to replace RELEASE-NAME with lowercase strings to use it:

---# Source: charts/presto/templates/deployment-worker.yamlapiVersion: apps/v1beta2kind: Deploymentmetadata:  name: RELEASE-NAME-presto-worker  labels:    app: presto    chart: presto-0.1    release: RELEASE-NAME    heritage: Tiller    component: workerspec:  replicas: 2  selector:    matchLabels:      app: presto      release: RELEASE-NAME      component: worker  template:    metadata:      labels:        app: presto        release: RELEASE-NAME        component: worker    spec:      volumes:        - name: config-volume          configMap:            name: RELEASE-NAME-presto-worker      containers:        - name: presto-worker          image: "bivas/presto:0.196"          imagePullPolicy: IfNotPresent          command: ["/bin/bash"]          args:            - /etc/presto/docker-presto.sh          volumeMounts:            - mountPath: /etc/presto              name: config-volume          livenessProbe:            exec:              command:                - /bin/bash                - /etc/presto/health_check.sh            initialDelaySeconds: 10            periodSeconds: 25          readinessProbe:            exec:              command:                - /bin/bash                - /etc/presto/health_check.sh            initialDelaySeconds: 5            periodSeconds: 10          resources:            {}