This message was deleted.

By best it depends on your infrastructure. Druid is able to export its metrics in different ways: •

http

-> it can periodically send this JSON payload into an endpoint you want •

kafka

-> it sends the metrics payload into a Kafka topic •

prometheus

-> exposes metrics into a local endpoint so a Prometheus scrapper can fetch them • there are others, like

statsd

, `influxdb`… They are all useful depending on what your infrastructure supports. If your k8s cluster has Prometheus running, it’s prob the best for you, so choose

prometheus

and you prob just need to add some annotations in your deployment like:

spec:
  template:
    metadata:
      annotations:
        <http://prometheus.io/scrape|prometheus.io/scrape>: "true"
        <http://prometheus.io/path|prometheus.io/path>: /metrics
        <http://prometheus.io/port|prometheus.io/port>: "9102"

You prob have other deployments into your repo with a template you can use

Thank you for your response! It appears that all methods, except Prometheus, rely on pushing metrics, whereas Prometheus requires pulling them from a local endpoint. Given this information, should I enable annotation in all Druid components? I came across an open-source project called Prometheus Druid Exporter (https://github.com/prometheus-community/helm-charts/tree/main/charts/prometheus-druid-exporter). Can we leverage this project to retrieve metrics from Druid components? As a newcomer to this subject, I greatly appreciate your assistance.

Don't the other services in your k8s use Prometheus as well? If so they almost sure are scrapped by Prometheus agent, probably via some k8s annotations similar to the one I showed before. And yes, ideally you should activate it in all Druid services as all of them got interesting metrics to monitor the cluster

We are using exporters within our Kubernetes cluster to obtain metrics for all our applications(https://github.com/prometheus-community/helm-charts/tree/main/charts/kube-prometheus-stack/templates/exporters). Currently, we are focusing on exploring additional metrics for druid. Our goal is to optimize the configuration and identify any bottlenecks. I have another question: Does the exported metrics provide information about failed queries?