Nouvelles annonces sur les technologies sans serveur et réseau ainsi que sur le RUM (Real-User Monitoring) dévoilées à la conférence Dash ! Nouvelles annonces dévoilées à la conférence Dash !

Running Endpoints Checks with Autodiscovery

Cette page n'est pas encore disponible en français, sa traduction est en cours.
Si vous avez des questions ou des retours sur notre projet de traduction actuel, n'hésitez pas à nous contacter.

How it Works

The Cluster Check feature provides the ability to autodiscover and perform checks on load-balanced cluster services (eg. Kubernetes services). The Endpoints Checks feature extends this mechanism to monitor any endpoint behind cluster services.

The Cluster Agent holds the configurations and exposes them to node-based Agents so they can consume and convert them into Endpoints Checks.

Endpoints Checks are scheduled by Agents that run on the same node as the pod(s) that back the endpoint(s) of the monitored service.

The Agents connect to the Cluster Agent every 10 seconds and retrieve the check configurations to run. Metrics coming from Endpoints Checks will be submitted with service, pod, and host tags.

This feature is currently supported on Kubernetes for versions 6.12.0+ of the Agent, and versions 1.3.0+ of the Cluster Agent.

Example: three NGINX pods exposed by the nginx service

# kubectl get svc nginx -o wide
nginx   ClusterIP   <none>        80/TCP    1h    app=nginx
# kubectl get pods --selector app=nginx
NAME                     READY   STATUS    RESTARTS   AGE
nginx-758655469f-59q9z   1/1     Running   0          20h
nginx-758655469f-k8zrc   1/1     Running   0          20h
nginx-758655469f-lk9p6   1/1     Running   0          20h
# kubectl get ep nginx -o yaml
- addresses:
  - ip:
    nodeName: gke-cluster-default-pool-4658d5d4-k2sn
      kind: Pod
      name: nginx-758655469f-lk9p6
  - ip:
    nodeName: gke-cluster-default-pool-4658d5d4-p39c
      kind: Pod
      name: nginx-758655469f-59q9z
  - ip:
    nodeName: gke-cluster-default-pool-4658d5d4-p39c
      kind: Pod
      name: nginx-758655469f-k8zrc

By design, Endpoints Checks are scheduled by Agents that run on the same node as the pods that back the endpoints of the nginx service, so only Agents running on the nodes gke-cluster-default-pool-4658d5d4-k2sn and gke-cluster-default-pool-4658d5d4-p39c will schedule the checks on the nginx pods.

This works like that to leverage autodiscovery, and attach pod and container tags to the metrics coming from these pods.

How to set it up

Cluster Agent setup

This feature requires a running Cluster Agent with the Cluster Checks feature enabled.

Agent setup

Enable the endpointschecks configuration providers on the Datadog Node Agent. This can be done in two ways:

  • By setting the DD_EXTRA_CONFIG_PROVIDERS environment variable:
  • Or adding it to the datadog.yaml configuration file:
  - name: endpointschecks
    polling: true

Restart the Agent to apply the configuration change.

Note: To enable both Cluster Checks and Endpoints Checks, clusterchecks and endpointschecks configuration providers should be both enabled on the Datadog Node Agent.

Setting up Check Configurations via Kubernetes Service Annotations

Similar to annotating Kubernetes Pods, Services can be annotated with the following syntax: '[<CHECK_NAME>]' '[<INIT_CONFIG>]' '[<INSTANCE_CONFIG>]' '[<CHECK_NAME>]' '[<INIT_CONFIG>]' '[<INSTANCE_CONFIG>]'

The %%host%% template variable is supported and is replaced by the service’s and endpoints’ IPs. The kube_namespace and kube_service tags are automatically added to the instances.

Note: For now, it is required to set both* and* annotations to enable the Endpoints Checks on the service’s endpoints.

Example: HTTP check on an NGINX-backed service with NGINX check on the service’s endpoints

The following service definition exposes the pods from the my-nginx deployment. It then runs an HTTP check to measure the latency of the load-balanced service and an NGINX check on the pod(s) that back the endpoint(s) of the service to collect NGINX metrics and service checks on the pod level:

apiVersion: v1
kind: Service
  name: my-nginx
    run: my-nginx
  annotations: '["http_check"]' '[{}]' |
          "name": "My Nginx Service",
          "url": "http://%%host%%",
          "timeout": 1
      ] '["nginx"]' '[{}]' |
          "name": "My Nginx Service Endpoints",
          "nginx_status_url": "http://%%host%%:%%port%%/nginx_status"
  - port: 80
    protocol: TCP
    run: my-nginx


Troubleshooting Endpoints Checks is similar to troubleshooting Cluster Checks—the only difference is on the node-based Agents, where scheduled Endpoints Checks appear alongside the Cluster Check.

Note: Endpoints Checks are scheduled by Agents that run on the same node as the pod(s) that back the endpoint(s) of the service.

Autodiscovery in the node-based Agent

The Agent configcheck command should show the instance, with the endpoints-checks source:

# kubectl exec <NODE_AGENT_POD_NAME> agent configcheck
=== nginx check ===
Source: endpoints-checks
Instance ID: nginx:My Nginx Service Endpoints:96eff84ce7d742b9
name: My Nginx Service Endpoints
- kube_deployment:nginx
- kube_namespace:default
- kube_service:nginx
- pod_phase:running
Init Config:
Auto-discovery IDs:
* kube_endpoint://default/nginx
* kubernetes_pod://e8667db4-5f8e-11e9-ae71-42010af0016b
* kube_service://6c964fcf-5d35-11e9-ae71-42010af0016b

Agent status

The Agent status command should show the check instance running and reporting successfully.

# kubectl exec <NODE_AGENT_POD_NAME> agent status
    nginx (3.1.0)
      Instance ID: nginx:My Nginx Service Endpoints:96eff84ce7d742b9 [OK]
      Total Runs: 2
      Metric Samples: Last Run: 7, Total: 14
      Events: Last Run: 0, Total: 0
      Service Checks: Last Run: 1, Total: 2
      Average Execution Time : 86ms

Further Reading