Skip to content

Latest commit

 

History

History
342 lines (267 loc) · 21.7 KB

kubernetes_deployment.md

File metadata and controls

342 lines (267 loc) · 21.7 KB
Raw

Kubernetes

We provide an OpenTelemetry Demo Helm chart to help deploy the demo to an existing Kubernetes cluster. We also provide a values.yaml template to customize the deployment for New Relic. More details on this are included below.

Helm must be installed to use the charts. Please refer to Helm's documentation to get started.

Prerequisites

  • Pre-existing Kubernetes 1.23+ Cluster
  • Helm 3.9+
  • New Relic Account

Please note that this chart is not supported for clusters running on arm64 architecture, such as kind/minikube running on Apple Silicon.

Install the Chart

Add OpenTelemetry Helm repository:

helm repo add open-telemetry https://open-telemetry.github.io/opentelemetry-helm-charts && helm repo update open-telemetry

Set your New Relic license key environment variable:

export NEW_RELIC_LICENSE_KEY='<NEW_RELIC_LICENSE_KEY>'

Set a Kubernetes secret containing your New Relic license key:

kubectl create ns opentelemetry-demo && kubectl create secret generic newrelic-license-key --from-literal=license-key="$NEW_RELIC_LICENSE_KEY" -n opentelemetry-demo

To install the chart with the release name newrelic-otel, run the following command and pass in the provided values.yaml file to customize the deployment:

helm upgrade --install newrelic-otel open-telemetry/opentelemetry-demo --version 0.32.0 --values ./helm/values.yaml -n opentelemetry-demo

Remark: If your New Relic account is in Europe, install the chart as follows instead:

helm upgrade --install newrelic-otel open-telemetry/opentelemetry-demo --version 0.32.0 --values ./helm/values.yaml --set opentelemetry-collector.config.exporters.otlp.endpoint="otlp.eu01.nr-data.net:4317" -n opentelemetry-demo

New Relic Overrides (Optional)

Optionally, you can enable a version of the recommendationService that is instrumented with New Relic APM instead of OpenTelemetry. New Relic APM instrumented services are interoperable with OpenTelemetry instrumented services as New Relic supports W3C trace context.

helm upgrade --install newrelic-otel open-telemetry/opentelemetry-demo --version 0.32.0 --values ./helm/values.yaml --values ./helm/recommendation_service_values.yaml -n opentelemetry-demo

Install Prometheus Exporters (Optional)

You can install the Prometheus Exporters for Kafka, Postgres, and Redis to expose Prometheus metrics for the Kafka, Postgres, and Redis components used by the demo application.

Add the Prometheus Helm repository:

helm repo add prometheus-community https://prometheus-community.github.io/helm-charts && helm repo update prometheus-community

Install the Prometheus Kafka exporter:

helm upgrade --install prometheus-kafka-exporter prometheus-community/prometheus-kafka-exporter --values ./helm/prometheus-kafka-exporter/values.yaml -n opentelemetry-demo

Install the Prometheus Postgres exporter:

helm upgrade --install prometheus-postgres-exporter prometheus-community/prometheus-postgres-exporter --values ./helm/prometheus-postgres-exporter/values.yaml -n opentelemetry-demo

Install the Prometheus Redis exporter:

helm upgrade --install prometheus-redis-exporter prometheus-community/prometheus-redis-exporter --values ./helm/prometheus-redis-exporter/values.yaml -n opentelemetry-demo

Install Kubernetes Integration (Optional)

You can install the New Relic Kubernetes integration, which includes the Prometheus Agent, to give you visibility into the K8s cluster used to host the demo application.

Add the New Relic Helm repository:

helm repo add newrelic https://helm-charts.newrelic.com && helm repo update newrelic

Install the New Relic Kubernetes integration (be sure to add your New Relic license key and K8s cluster name):

 helm upgrade --install newrelic-bundle newrelic/nri-bundle \
 --set global.licenseKey=<NEW_RELIC_LICENSE_KEY> \
 --set global.cluster=<K8S_CLUSTER_NAME> \
 --namespace=newrelic \
 --create-namespace \
 --set newrelic-infrastructure.privileged=true \
 --set nri-metadata-injection.enable=true \
 --set kube-state-metrics.enabled=true \
 --set newrelic-logging.enabled=false \
 --set nri-kube-events.enabled=true \
 --set newrelic-prometheus-agent.enabled=true \
 --set newrelic-prometheus-agent.lowDataMode=true \
 --set-json='newrelic-prometheus-agent.config.kubernetes.integrations_filter.app_values=["redis", "kafka", "postgres"]'

Helm Chart Parameters

Chart parameters are separated in 4 general sections:

  • default - Used to specify defaults applied to all demo components
  • components - Used to configure the individual components (microservices) for the demo
  • opentelemetry-collector - Used to configure the OpenTelemetry Collector

New Relic Configurations

In our values template we have disabled several observability components in favor of the New Relic tool suite:

Parameter Description Default
opentelemetry-collector.enabled Enables the OpenTelemetry Collector sub-chart true
jaeger.enabled Enables the Jaeger sub-chart false
prometheus.enabled Enables the Prometheus sub-chart false
grafana.enabled Enables the Grafana sub-chart false
opensearch.enabled Enables the Opensearch sub-chart false

OpenTelemetry Collector

Note The following parameters have a opentelemetry-collector. prefix.

  • mode: Specifies the mode in which the collector should run. In this case, it is set to run in a "daemonset" mode, which means that the collector will be running on each node in a Kubernetes cluster. We use this mode for the resourcedetection, k8sattributes processors
  • service: Configures the service that is exposed by the collector, it uses LoadBalancer type service on AWS.
  • ports: Configures the ports that the collector should listen on. In this case, it is configured to listen on several different ports for different protocols such as otlp, otlp-http, jaeger-compact, jaeger-thrift, jaeger-grpc, zipkin and metrics, with some protocols enabled and some disabled based on support and usage.
  • podAnnotations: Configures annotations that should be added to the collector's pod. In this case, it is configured to be scraped by Prometheus and specify the port for Prometheus to scrape on.
  • config :
  • service inside config is used to configure the service extensions, pipelines are used to configure the different pipelines that the collector should use to process different types of telemetry data, like traces and metrics with their respective receivers, processors and exporters.
  • ingress configuration is not present in the file since we are using the LoadBalancer type service on AWS

Note Please make sure to update the collector configuration options according to your needs before using values template file.

Component parameters

The OpenTelemetry demo contains several components (microservices). Each component is configured with a common set of parameters. All components will be defined within components.[NAME] where [NAME] is the name of the demo component.

Note The following parameters require a components.[NAME]. prefix where [NAME] is the name of the demo component.

Parameter Description Default
enabled Is this component enabled true
useDefault.env Use the default environment variables in this component true
imageOverride.repository Name of image for this component Defaults to the overall default image repository
imageOverride.tag Tag of the image for this component Defaults to the overall default image tag
imageOverride.pullPolicy Image pull policy for this component IfNotPresent
imageOverride.pullSecrets Image pull secrets for this component []
servicePort Service port used for this component nil
ports Array of ports to open for deployment and service of this component []
env Array of environment variables added to this component Each component will have its own set of environment variables
envOverrides Used to override individual environment variables without re-specifying the entire array []
resources CPU/Memory resource requests/limits Each component will have a default memory limit set
schedulingRules.nodeSelector Node labels for pod assignment {}
schedulingRules.affinity Man of node/pod affinities {}
schedulingRules.tolerations Tolerations for pod assignment []
securityContext Container security context to define user ID (UID), group ID (GID) and other security policies {}
podAnnotations Pod annotations for this component {}
ingress.enabled Enable the creation of Ingress rules false
ingress.annotations Annotations to add to the ingress rule {}
ingress.ingressClassName Ingress class to use. If not specified default Ingress class will be used. nil
ingress.hosts Array of Hosts to use for the ingress rule. []
ingress.hosts[].paths Array of paths / routes to use for the ingress rule host. []
ingress.hosts[].paths[].path Actual path route to use nil
ingress.hosts[].paths[].pathType Path type to use for the given path. Typically this is Prefix. nil
ingress.hosts[].paths[].port Port to use for the given path nil
ingress.additionalIngresses Array of additional ingress rules to add. This is handy if you need to differently annotated ingress rules []
ingress.additionalIngresses[].name Each additional ingress rule needs to have a unique name nil
command Command & arguments to pass to the container being spun up for this service []
configuration Configuration for the container being spun up; will create a ConfigMap, Volume and VolumeMount {}

The services are configured to use the OpenTelemetry exporter to send traces and metrics to a backend service. The endpoint for these services are set to use the environment variable OTEL_EXPORTER_OTLP_ENDPOINT. The endpoint is set to the value http://$(HOST_IP):4317 and http://$(HOST_IP):4318 respectively.

This configuration is necessary because the service is deployed as a Kubernetes daemonset. A daemonset ensures that all, or some, nodes run a copy the collector pod. This means the IP address of the endpoint may be different depending on the node that the pod is running on. By using the variable HOST_IP the exporter will automatically substitute the correct IP address of the endpoint when the pod starts up, regardless of which node it's running on.

So, this configuration ensures that the exporter sends traces and metrics to the correct endpoint, even if the pod is running on different nodes in the kubernetes cluster.

Default parameters (applied to all demo components)

Property Description Default
default.env Environment variables added to all components Array of several OpenTelemetry environment variables
default.envOverrides Used to override individual environment variables without re-specifying the entire array. []
default.image.repository Demo components image name otel/demo
default.image.tag Demo components image tag (leave blank to use app version) nil
default.image.pullPolicy Demo components image pull policy IfNotPresent
default.image.pullSecrets Demo components image pull secrets []
default.schedulingRules.nodeSelector Node labels for pod assignment {}
default.schedulingRules.affinity Man of node/pod affinities {}
default.schedulingRules.tolerations Tolerations for pod assignment []
default.securityContext Demo components container security context {}
serviceAccount The name of the ServiceAccount to use for demo components ""

Use the Demo

The demo application will need the services exposed outside of the Kubernetes cluster in order to use them. You can expose the services to your local system using the kubectl port-forward command or by configuring service types (ie: LoadBalancer) with optionally deployed ingress resources.

Expose services using kubectl port-forward

To expose the frontendproxy service use the following command (replace newrelic-otel with your Helm chart release name accordingly):

kubectl port-forward svc/newrelic-otel-frontendproxy 8080:8080

In order for spans from the browser to be properly collected, you will also need to expose the OpenTelemetry Collector's OTLP/HTTP port (replace newrelic-otel with your Helm chart release name accordingly):

kubectl port-forward svc/newrelic-otel-otelcol 4318:4318

Note kubectl port-forward will proxy the port until the process terminates. You may need to create separate terminal sessions for each use of kubectl port-forward, and use CTRL-C to terminate the process when done.

The following services are available at these paths once the proxy is exposed:

Expose services using service type configurations

Note Kubernetes clusters may not have the proper infrastructure components to enable LoadBalancer service types or ingress resources. Verify your cluster has the proper support before using these configuration options.

Each demo service (ie: frontendproxy) offers a way to have its Kubernetes service type configured. By default these will be ClusterIP but you can change each one using the serviceType property for each service.

To configure the frontendproxy service to use a LoadBalancer service type you would specify the following in your values file:

components:
  frontendProxy:
    serviceType: LoadBalancer

Note It is recommended to use a values file when installing the Helm chart in order to specify additional configuration options.

The Helm chart does not provide facilities to create ingress resources. If required these would need to be created manually after installing the Helm chart. Some Kubernetes providers require specific service types in order to be used by ingress resources (ie: EKS ALB ingress, requires a NodePort service type).

In order for spans from the browser to be properly collected, you will also need to expose the OpenTelemetry Collector's OTLP/HTTP port to be accessible to user web browsers. The location where the OpenTelemetry Collector is exposed must also be passed into the frontend service using the PUBLIC_OTEL_EXPORTER_OTLP_TRACES_ENDPOINT environment variable. You can do this using the following in your values file:

components:
  frontend:
    env:
      - name: PUBLIC_OTEL_EXPORTER_OTLP_TRACES_ENDPOINT
        value: "http://otel-demo-collector.mydomain.com:4318/v1/traces"

With the frontendproxy and Collector exposed, you can access the demo UI at the base path for the frontendproxy. Other demo components can be accessed at the following sub-paths:

  • Webstore: / (base)
  • Load Generator UI: /loadgen/ (must include trailing slash)