Overview

This integration gets container metrics from kubelet

• Visualize and monitor kubelet stats
• Be notified about kubelet failovers and events.

Setup

Installation

The Kubelet check is included in the Datadog Agent package, so you don’t need to install anything else on your servers.

Configuration

Edit the kubelet.d/conf.yaml file, in the conf.d/ folder at the root of your Agent’s configuration directory. See the sample kubelet.d/conf.yaml for all available configuration options.

Validation

Run the Agent’s status subcommand and look for kubelet under the Checks section.

Compatibility

The kubelet check can run in two modes:

• The default prometheus mode is compatible with Kubernetes version 1.7.6 or superior
• The cAdvisor mode (enabled by setting the cadvisor_port option) should be compatible with versions 1.3 and up. Consistent tagging and filtering requires at least version 6.2 of the Agent.

OpenShift <3.7 support

The cAdvisor 4194 port is disabled by default on OpenShift. To enable it, you need to add the following lines to your node-config file:

kubeletArguments:
  cadvisor-port: ["4194"]

If you cannot open the port, disable both sources of container metric collection, by setting:

• cadvisor_port to 0
• metrics_endpoint to ""

The check can still collect:

• kubelet health service checks
• pod running/stopped metrics
• pod limits and requests
• node capacity metrics

Data Collected

Service Checks

kubernetes.kubelet.check.ping
Returns CRITICAL if the Kubelet doesn’t respond to Ping. OK, otherwise
Statuses: ok, critical

kubernetes.kubelet.check.docker
Returns CRITICAL if the Docker service doesn’t run on the Kubelet. OK, otherwise
Statuses: ok, critical

kubernetes.kubelet.check.syncloop
Returns CRITICAL if the syncloop health check is down. OK, otherwise
Statuses: ok, critical

kubernetes.kubelet.check
Returns CRITICAL if the overall Kubelet health check is down. OK, otherwise
Statuses: ok, critical

Excluded containers

To restrict the data collected to a subset of the containers deployed, set the DD_CONTAINER_EXCLUDE environment variable. Metrics are not included from the containers specified in that environment variable.

For network metrics reported at the pod level, containers cannot be excluded based on name or image name since other containers can be part of the same pod. So, if DD_CONTAINER_EXCLUDE applies to a namespace, the pod-level metrics are not reported if the pod is in that namespace. However, if DD_CONTAINER_EXCLUDE refers to a container name or image name, the pod-level metrics are reported even if the exclusion rules apply to some containers in the pod.

Troubleshooting

Need help? Contact Datadog support.