Monitoring Collection

Plugin

This plugin allows for other plugins to add data to Kibana stack monitoring documents.

OpenTelemetry Metrics

Enable Prometheus endpoint with Elastic Agent Prometheus input

1. Start local setup with fleet or a cloud cluster

2. Start Kibana

3. Set up a new agent policy and enroll a new agent in your local machine

4. Install the Prometheus Metrics package

   1. Set Hosts with localhost:5601
   2. Set Metrics Path with /(BASEPATH)/api/monitoring_collection/v1/prometheus
   3. Remove the values from Bearer Token File and SSL Certificate Authorities
   4. Set Username and Password with elastic and changeme

5. Add the following configuration to kibana.dev.yml

   ymlCopy

   # Enable the prometheus exporter
   monitoring_collection.opentelemetry.metrics:
     prometheus.enabled: true
   

Enable OpenTelemetry Metrics API exported as OpenTelemetry Protocol over GRPC

1. Start local setup with fleet or a cloud cluster

2. Start Kibana

3. Set up a new agent policy and enroll a new agent in your local machine

4. Install Elastic APM package listening on localhost:8200 without authentication

5. Add the following configuration to kibana.dev.yml

   ymlCopy

   # Enable the OTLP exporter
   monitoring_collection.opentelemetry.metrics:
     otlp.url: "http://127.0.0.1:8200"
   

You can also provide headers for OTLP endpoints that require authentication:

# Enable the OTLP exporter to an authenticated APM endpoint
monitoring_collection.opentelemetry.metrics:
  otlp:
    url: "https://DEPLOYMENT.apm.REGION.PROVIDER.elastic-cloud.com"
    headers:
      Authorization: "Bearer SECRET_TOKEN"

Alternatively, OTLP Exporter can be configured using environment variables OTEL_EXPORTER_OTLP_ENDPOINT, OTEL_EXPORTER_OTLP_METRICS_ENDPOINT and OTEL_EXPORTER_OTLP_METRICS_HEADERS. See OTLP Exporter docs for details.

It's possible to configure logging for the OTLP integration. If not informed, the default will be info

monitoring_collection.opentelemetry.metrics:
  logLevel: warn | info | debug | warn | none | verbose | all

For connection-level debug information you can set these variables:

export GRPC_NODE_TRACE="xds_client,xds_resolver,cds_balancer,eds_balancer,priority,weighted_target,round_robin,resolving_load_balancer,subchannel,keepalive,dns_resolver,fault_injection,http_filter,csds"
export GRPC_NODE_VERBOSITY=DEBUG

See the grpc-node docs for details and other settings.

Example of how to instrument the code

• First, we need to define what metrics we want to instrument with OpenTelemetry

  tsCopy

  import { Counter, Meter } from '@opentelemetry/api';
  
  export class FooApiMeters {
    requestCount: Counter;
  
    constructor(meter: Meter) {
      this.requestCount = meter.createCounter('request_count', {
        description: 'Counts total number of requests',
      });
    }
  }
  

  In this example we're using a Counter metric, but OpenTelemetry SDK provides there are other options to record metrics

• Initialize meter in the plugin setup and pass it to the relevant components that will be instrumented. In this case, we want to instrument FooApi routes.

  tsCopy

  import { IRouter } from '@kbn/core/server';
  import { FooApiMeters } from './foo_api_meters';
  import { metrics } from '@opentelemetry/api';
  
  export class FooApiPlugin implements Plugin {
    private metrics: Metrics;
    private libs: { router: IRouter, metrics: FooApiMeters};
  
    constructor() {
      this.metrics = new Metrics(metrics.getMeter('kibana.fooApi'));
    }
  
    public setup(core: CoreSetup) {
      const router = core.http.createRouter();
  
      this.libs = {
        router,
        metrics: this.metrics
      }
  
      initMetricsAPIRoute(this.libs);
    }
  }
  

  monitoring_collection plugins has to be initialized before the plugin that will be instrumented. If for some reason the instrumentation doesn't record any metrics, make sure monitoring_collection is included in the list of requiredPlugins. e.g:

  jsonCopy

  "requiredPlugins": [
    "monitoringCollection"
  ],
  

• Lastly we can use the metrics object to instrument the code

  tsCopy

  export const initMetricsAPIRoute = (libs: { router: IRouter, metrics: FooApiMeters}) => {
    router.get(
      {
        path: '/api/foo',
        validate: {},
      },
      async function (_context, _req, res) {
        metrics.requestCount.add(1);
        return res.ok({});
      }
    );