Kubernetes

In this section we will cover how to deploy a NEO•ONE Node to a Kubernetes cluster.

If you are unfamiliar with Kubernetes visit their getting started page, in particular we will be implementing a StatefulSet locally using Kubernetes through Docker.

Requirements
Getting Started
Pause/Shutdown
Configuration
Health Probes
Exposing Pods of a StatefulSet Locally

Requirements

Docker
- Minimum: at least 2GB Memory, and 50GB Storage allocated
- Recommended: 4GB Memory, 60GB+ Storage allocated (you will need ~60GB of storage per pod)
kubectl
- You can enable Kubernetes through the docker GUI; Docker >> Preferences… >> Kubernetes >> Enable Kubernetes

Getting Started

The following deployment spec will create a StatefulSet of n nodes defined by spec.replicas. Each requests 60GB of storage and 4GB of memory. If you do not have a default storage class set (docker will automatically create one for local deployments) you will need to create one, see storage classes for more information.

A requirement of StateFul sets is a headless service, so we’ll start by creating neo-one-service:

#node-svc.yml
apiVersion: v1
kind: Service
metadata:
  name: neo-one-node
  labels:
    app: neo-one-node
spec:
  ports:
    - port: 8080
      name: node
  clusterIP: None
  selector:
    app: neo-one-node

followed by the StatefulSet itself which we’ll save as node-spec.yml:

# node-spec.yml
apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: node
spec:
  serviceName: "neo-one-node"
  replicas: 1
  selector:
    matchLabels:
      app: neo-one-node
  template:
    metadata:
      labels:
        app: neo-one-node
    spec:
      containers:
      - name: neo-one-node
        image: neoonesuite/node
        imagePullPolicy: IfNotPresent
        ports:
        - containerPort: 8080
          name: node
        volumeMounts:
        - name: node-data
          mountPath: /root/.local/share/neo-one
        args: [
          "--node.rpcURLs=http://seed6.ngd.network:10332",
          "--node.rpcURLs=https://seed1.red4sec.com:10332"
        ]
        resources:
          requests:
            memory: "4Gi"
            cpu: "1"
  volumeClaimTemplates:
  - metadata:
      name: node-data
    spec:
      accessModes: [ "ReadWriteOnce" ]
      resources:
        requests:
          storage: 60Gi

Running this deployment with

kubectl apply -f node-svc.yml
kubectl apply -f node-spec.yml

will start a single pod which:

makes a persistent volume claim for 60GB of storage.
starts the node with this volume mounted to the default node-storage path
if node-data isn’t present, restore from the public google-cloud backup
sync the node using two seeds from http://monitor.cityofzion.io/

There are two main benefits to deploying the nodes this way. If a pod needs to restart for any reason it will always attempt to bind to the same persistent volume and will not start until it is scheduled on the same machine as that volume. It also makes it incredibly simple to scale the number of nodes you would like to run.

Pause/Shutdown

The simplest way to pause any pods from scheduling/running is by setting spec.replicas: 0 in your node-spec.yml and running

kubectl apply -f node-spec.yml

To purge the persisted volume space and shutdown the headless service you can simply delete both with:

kubectl delete svc neo-one-service
kubectl delete statefulset node

Configuration

Configuration Reference

While the above examples shows how to configure our pods with environment variables we can also mount a ConfigMap to our pods. As an example we will use a basic configuration:

#/path/to/config.json
{
  "telemetry": {
    "logging": {
      "level": "debug"
    }
  }
}

Creating a ConfigMap is as easy as running:

#create the configMap
kubectl create configmap example-config-map --from-file=config=/path/to/config.json
#inspect the configMap
kubectl describe configmap example-config-map

Then, we can apply the ConfigMap by modifying the node-spec above to the following:

# node-spec.yml
apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: node
spec:
  serviceName: 'neo-one-node'
  podManagementPolicy: 'Parallel'
  replicas: 1
  selector:
    matchLabels:
      app: neo-one-node
  template:
    metadata:
      labels:
        app: neo-one-node
    spec:
      containers:
        - name: neo-one-node
          image: neoonesuite/node
          ports:
            - containerPort: 8080
              name: node
          volumeMounts:
            - name: node-data
              mountPath: /root/.local/share/neo-one
            - name: config-volume
              mountPath: /etc/neo-one/config
              subPath: config
          resources:
            requests:
              memory: '4Gi'
              cpu: '1'
      volumes:
        - name: config-volume
          configMap:
            name: example-config-map
            optional: true

  volumeClaimTemplates:
    - metadata:
        name: node-data
      spec:
        accessModes: ['ReadWriteOnce']
        resources:
          requests:
            storage: 60Gi

Note

Because of how rc searches for configuration files the key of the configMap MUST be config no extension.

Health Probes

Sometimes a node will go down and need to be restarted, more commonly a node gets 'stuck' while syncing the blockchain data from other nodes. Because of this it can be extremely useful to configure health probes alongside your pods.

To enable health checks on the NEO•ONE Node you must first enable the options in your configuration. Assuming you have set up configuration mounting described above the following example config.json will enable health checks:

{
  "node": {
    "rpcURLs": [
      "http://seed1.ngd.network:10332",
      "http://seed2.ngd.network:10332",
      "http://seed3.ngd.network:10332",
      "http://seed4.ngd.network:10332",
      "http://seed5.ngd.network:10332",
      "http://seed6.ngd.network:10332",
      "http://seed7.ngd.network:10332",
      "http://seed8.ngd.network:10332",
      "http://seed9.ngd.network:10332",
      "http://seed10.ngd.network:10332",
      "https://seed1.cityofzion.io:443",
      "https://seed2.cityofzion.io:443",
      "https://seed3.cityofzion.io:443",
      "https://seed4.cityofzion.io:443",
      "https://seed5.cityofzion.io:443"
    ]
  },
  "network": {
    "listenTCP": {
      "port": 8081
    }
  },
  "rpc": {
    "http": {
      "port": 8080
    },
    "liveHealthCheck": {
      "rpcURLs": [
        "http://seed1.ngd.network:10332",
        "http://seed2.ngd.network:10332",
        "http://seed3.ngd.network:10332",
        "http://seed4.ngd.network:10332",
        "http://seed5.ngd.network:10332",
        "http://seed6.ngd.network:10332",
        "http://seed7.ngd.network:10332",
        "http://seed8.ngd.network:10332",
        "http://seed9.ngd.network:10332",
        "http://seed10.ngd.network:10332",
        "https://seed1.cityofzion.io:443",
        "https://seed2.cityofzion.io:443",
        "https://seed3.cityofzion.io:443",
        "https://seed4.cityofzion.io:443",
        "https://seed5.cityofzion.io:443"
      ]
    },
    "readyHealthCheck": {
      "rpcURLs": [
        "http://seed1.ngd.network:10332",
        "http://seed2.ngd.network:10332",
        "http://seed3.ngd.network:10332",
        "http://seed4.ngd.network:10332",
        "http://seed5.ngd.network:10332",
        "http://seed6.ngd.network:10332",
        "http://seed7.ngd.network:10332",
        "http://seed8.ngd.network:10332",
        "http://seed9.ngd.network:10332",
        "http://seed10.ngd.network:10332",
        "https://seed1.cityofzion.io:443",
        "https://seed2.cityofzion.io:443",
        "https://seed3.cityofzion.io:443",
        "https://seed4.cityofzion.io:443",
        "https://seed5.cityofzion.io:443"
      ]
    }
  }
}

Be sure to apply this new config to your ConfigMap example-config-map by running:

kubectl delete configmap example-config-map
kubectl create configmap example-config-map --from-file=config=/path/to/config.json

After our config has been mapped we can add the liveness/readiness configurations to our node-spec.yml:

apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: node
spec:
  selector:
    matchLabels:
      app: neo-one-node
  serviceName: 'neo-one-node'
  podManagementPolicy: 'Parallel'
  replicas: 1
  template:
    metadata:
      labels:
        app: neo-one-node
    spec:
      containers:
        - name: neo-one-node
          image: neoonesuite/node
          ports:
            - containerPort: 8080
              name: node
          volumeMounts:
            - name: node-data
              mountPath: /root/.local/share/neo-one
            - name: config-volume
              mountPath: /etc/neo-one/config
              subPath: config
          resources:
            requests:
              memory: '4Gi'
              cpu: '1'
          livenessProbe:
            httpGet:
              path: /live_health_check
              port: 8080
            initialDelaySeconds: 30
            periodSeconds: 60
          readinessProbe:
            httpGet:
              path: /ready_health_check
              port: 8080
            initialDelaySeconds: 30
            periodSeconds: 15
      volumes:
        - name: config-volume
          configMap:
            name: example-config-map
            optional: true

  volumeClaimTemplates:
    - metadata:
        name: node-data
      spec:
        accessModes: ['ReadWriteOnce']
        resources:
          requests:
            storage: 60Gi

After starting this StatefulSet with

kubectl apply -f node-spec.yml

you should be able to inspect the state of the health checks (once they start running) with

kubectl describe pod node-0

see Kubernetes Documentation for more information on health probes.

Exposing Pods of a StatefulSet Locally

Since we start our StatefulSet under a headless service we do not have direct access to the underlying endpoints (see headless services). The solution to this issue is by creating another service which explicitly targets the pods created by the stateful set. For our single pod example it is as simple as starting a service:

#node-0-svc.yml
apiVersion: v1
kind: Service
metadata:
  name: node-0-svc
spec:
  type: LoadBalancer
  externalTrafficPolicy: Local
  selector:
    statefulset.kubernetes.io/pod-name: node-0
  ports:
    - protocol: TCP
      port: 8080
      targetPort: 8080

Voilà! You should now be able to access the pod locally! You can test this by running

curl localhost:8080/live_health_check

Edit this page

Local Docker Development

Docker Compose

DOCS

Installation Main Concepts Advanced Guides API Reference Contributing

CHANNELS

GitHub Stack Overflow Discord Chat Twitter YouTube

Tutorial Blog