用Helm 部署NATS

The NATS Helm charts can be used to deploy a StatefulSet of NATS servers using Helm templates which are easy to extend. Using Helm3 you can add the NATS Helm repo as follows:

helm repo add nats https://nats-io.github.io/k8s/helm/charts/
helm install my-nats nats/nats

The ArtifactHub NATS Helm package contains a complete list of configuration options. Some common scenarios are outlined below.

Configuration

Server Image

nats:
  image: nats:2.7.4-alpine
  pullPolicy: IfNotPresent

Limits

nats:
  # The number of connect attempts against discovered routes.
  connectRetries: 30

  # How many seconds should pass before sending a PING
  # to a client that has no activity.
  pingInterval:

  # Server settings.
  limits:
    maxConnections:
    maxSubscriptions:
    maxControlLine:
    maxPayload:

    writeDeadline:
    maxPending:
    maxPings:
    lameDuckDuration:

  # Number of seconds to wait for client connections to end after the pod termination is requested
  terminationGracePeriodSeconds: 60

Logging

Note: It is not recommended to enable trace or debug in production since enabling it will significantly degrade performance.

nats:
  logging:
    debug:
    trace:
    logtime:
    connectErrorReports:
    reconnectErrorReports:

TLS setup for client connections

You can find more on how to set up and troubleshoot TLS connections at: running-a-nats-service/configuration/securing_nats/tls

nats:
  tls:
    secret:
      name: nats-client-tls
    ca: "ca.crt"
    cert: "tls.crt"
    key: "tls.key"

Example of creating the nats-client-tls k8s secret with three named values matching the above setup:

kubectl create secret generic nats-client-tls --from-file=tls.crt=./broker.crt --from-file=tls.key=./broker.key --from-file=ca.crt=./ca.pem

Clustering

If clustering is enabled, then a 3-node cluster will be set up. More info at: running-a-nats-server/configuration/clustering#nats-server-clustering

cluster:
  enabled: true
  replicas: 3

  tls:
    secret:
      name: nats-server-tls
    ca: "ca.crt"
    cert: "tls.crt"
    key: "tls.key"

Example:

helm install nats nats/nats --set cluster.enabled=true

Leafnodes

Leafnode connections to extend a cluster. More info at: running-a-nats-server/configuration/leafnodes

leafnodes:
  enabled: true
  remotes:
    - url: "tls://connect.ngs.global:7422"
      # credentials:
      #   secret:
      #     name: leafnode-creds
      #     key: TA.creds
      # tls:
      #   secret:
      #     name: nats-leafnode-tls
      #   ca: "ca.crt"
      #   cert: "tls.crt"
      #   key: "tls.key"

  #######################
  #                     #
  #  TLS Configuration  #
  #                     #
  #######################
  # 
  #  # You can find more on how to setup and trouble shoot TLS connnections at:
  # 
  #  # https://docs.nats.io/running-a-nats-server/configuration/securing_nats/tls
  # 
  tls:
    secret:
      name: nats-client-tls
    ca: "ca.crt"
    cert: "tls.crt"
    key: "tls.key"

Websocket Configuration

websocket:
  enabled: true
  port: 443

  tls:
    secret:
      name: nats-tls
    cert: "fullchain.pem"
    key: "privkey.pem"

Setting up External Access

Using HostPorts

In case of both external access and advertisements being enabled, an initializer container will be used to gather the public IPs. This container will be required to have enough RBAC policy to be able to make a look up of the public IP of the node where it is running.

For example, to set up external access for a cluster and advertise the public IP to clients:

nats:
  # Toggle whether to enable external access.
  # This binds a host port for clients, gateways and leafnodes.
  externalAccess: true

  # Toggle to disable client advertisements (connect_urls),
  # in case of running behind a load balancer (which is not recommended)
  # it might be required to disable advertisements.
  advertise: true

  # In case both external access and advertise are enabled
  # then a service account would be required to be able to
  # gather the public IP from a node.
  serviceAccount: "nats-server"

Where the service account named nats-server has the following RBAC policy for example:

---
apiVersion: v1
kind: ServiceAccount
metadata:
  name: nats-server
  namespace: default
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
  name: nats-server
rules:
- apiGroups: [""]
  resources:
  - nodes
  verbs: ["get"]
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
  name: nats-server-binding
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: ClusterRole
  name: nats-server
subjects:
- kind: ServiceAccount
  name: nats-server
  namespace: default

The container image of the initializer can be customized via:

bootconfig:
  image: natsio/nats-boot-config:latest
  pullPolicy: IfNotPresent

Using LoadBalancers

When using a load balancer for external access, it is recommended to disable advertisement so that internal IPs from the NATS Servers are not advertised to the clients connecting through the load balancer.

nats:
  image: nats:alpine

cluster:
  enabled: true
  noAdvertise: true

leafnodes:
  enabled: true
  noAdvertise: true

natsbox:
  enabled: true

You could then use an L4 enabled load balancer to connect to NATS, for example:

apiVersion: v1
kind: Service
metadata:
  name: nats-lb
spec:
  type: LoadBalancer
  selector:
    app.kubernetes.io/name: nats
  ports:
    - protocol: TCP
      port: 4222
      targetPort: 4222
      name: nats
    - protocol: TCP
      port: 7422
      targetPort: 7422
      name: leafnodes
    - protocol: TCP
      port: 7522
      targetPort: 7522
      name: gateways

Gateways

A supercluster can be formed by pointing to remote gateways. You can find more about gateways in the NATS documentation: running-a-nats-server/configuration/gateways.

gateway:
  enabled: false
  name: 'default'

  #############################
  #                           #
  #  List of remote gateways  #
  #                           #
  #############################
  # gateways:
  #   - name: other
  #     url: nats://my-gateway-url:7522

  #######################
  #                     #
  #  TLS Configuration  #
  #                     #
  #######################
  # 
  #  # You can find more on how to setup and trouble shoot TLS connnections at:
  # 
  #  # https://docs.nats.io/running-a-nats-server/configuration/securing_nats/tls
  #
  # tls:
  #   secret:
  #     name: nats-client-tls
  #   ca: "ca.crt"
  #   cert: "tls.crt"
  #   key: "tls.key"

Auth setup

Auth with a Memory Resolver

auth:
  enabled: true

  # Reference to the Operator JWT.
  operatorjwt:
    configMap:
      name: operator-jwt
      key: KO.jwt

  # Public key of the System Account
  systemAccount:

  resolver:
    ############################
    #                          #
    # Memory resolver settings #
    #                          #
    ##############################
    type: memory

    # 
    # Use a configmap reference which will be mounted
    # into the container.
    # 
    configMap:
      name: nats-accounts
      key: resolver.conf

Auth using an Account Server Resolver

auth:
  enabled: true

  # Reference to the Operator JWT.
  operatorjwt:
    configMap:
      name: operator-jwt
      key: KO.jwt

  # Public key of the System Account
  systemAccount:

  resolver:
    ##########################
    #                        #
    #  URL resolver settings #
    #                        #
    ##########################
    type: URL
    url: "http://nats-account-server:9090/jwt/v1/accounts/"

JetStream

Setting up Memory and File Storage

File Storage is always recommended, since JetStream's RAFT Meta Group will be persisted to file storage. The Storage Class used should be block storage. NFS is not recommended.

nats:
  image: nats:alpine

  jetstream:
    enabled: true

    memStorage:
      enabled: true
      size: 2Gi

    fileStorage:
      enabled: true
      size: 10Gi
      # storageClassName: gp2 # NOTE: AWS setup but customize as needed for your infra.

Using with an existing PersistentVolumeClaim

For example, given the following PersistentVolumeClaim:

---
kind: PersistentVolumeClaim
apiVersion: v1
metadata:
  name: nats-js-disk
  annotations:
    volume.beta.kubernetes.io/storage-class: "default"
spec:
  accessModes:
    - ReadWriteOnce
  resources:
    requests:
      storage: 3Gi

You can start JetStream so that one pod is bound to it:

nats:
  image: nats:alpine

  jetstream:
    enabled: true

    fileStorage:
      enabled: true
      storageDirectory: /data/
      existingClaim: nats-js-disk
      claimStorageSize: 3Gi

Clustering example

nats:
  image: nats:alpine

  jetstream:
    enabled: true

    memStorage:
      enabled: true
      size: "2Gi"

    fileStorage:
      enabled: true
      size: "1Gi"
      storageDirectory: /data/
      storageClassName: default

cluster:
  enabled: true
  # Cluster name is required, by default will be release name.
  # name: "nats"
  replicas: 3

Misc

NATS Box

A lightweight container with NATS and NATS Streaming utilities deployed along the cluster to confirm the setup. You can find the image at: https://github.com/nats-io/nats-box

natsbox:
  enabled: true
  image: nats:alpine
  pullPolicy: IfNotPresent

  # credentials:
  #   secret:
  #     name: nats-sys-creds
  #     key: sys.creds

Configuration Reload sidecar

The NATS config reloader image to use:

reloader:
  enabled: true
  image: natsio/nats-server-config-reloader:latest
  pullPolicy: IfNotPresent

Prometheus Exporter sidecar

You can toggle whether to start the sidecar to be used to feed metrics to Prometheus:

exporter:
  enabled: true
  image: natsio/prometheus-nats-exporter:latest
  pullPolicy: IfNotPresent

Prometheus operator ServiceMonitor support

You can enable Prometheus operator ServiceMonitor:

exporter:
  # You have to enable exporter first
  enabled: true
  serviceMonitor:
    enabled: true
    ## Specify the namespace where Prometheus Operator is running
    # namespace: monitoring
    # ...

Pod Customizations

Security Context

 # Toggle whether to use setup a Pod Security Context
 # ## ref: https://kubernetes.io/docs/tasks/configure-pod-container/security-context/
securityContext:
  fsGroup: 1000
  runAsUser: 1000
  runAsNonRoot: true

Affinity

https://kubernetes.io/docs/concepts/configuration/assign-pod-node/#affinity-and-anti-affinity

matchExpressions must be configured according to your setup

affinity:
  nodeAffinity:
    requiredDuringSchedulingIgnoredDuringExecution:
      nodeSelectorTerms:
        - matchExpressions:
            - key: node.kubernetes.io/purpose
              operator: In
              values:
                - nats
  podAntiAffinity:
    requiredDuringSchedulingIgnoredDuringExecution:
      - labelSelector:
          matchExpressions:
            - key: app
              operator: In
              values:
                - nats
                - stan
        topologyKey: "kubernetes.io/hostname"

Service topology

Service topology is disabled by default but can be enabled by setting topologyKeys. For example:

topologyKeys:
  - "kubernetes.io/hostname"
  - "topology.kubernetes.io/zone"
  - "topology.kubernetes.io/region"

CPU/Memory Resource Requests/Limits

Sets the pods CPU/memory requests/limits

nats:
  resources:
    requests:
      cpu: 2
      memory: 4Gi
    limits:
      cpu: 4
      memory: 6Gi

No resources are set by default.

Annotations

https://kubernetes.io/docs/concepts/overview/working-with-objects/annotations

podAnnotations:
  key1 : "value1",
  key2 : "value2"

Name Overrides

Can change the name of the resources as needed with:

nameOverride: "my-nats"

Image Pull Secrets

imagePullSecrets:
- name: myRegistry

Adds this to the StatefulSet:

spec:
  imagePullSecrets:
    - name: myRegistry