Immutable solution version
Datadog on Kubernetes: Helm chart with DaemonSet agent plus Cluster Agent
Complete solution
The supported Kubernetes install is the Helm chart: a DaemonSet agent on every node plus the Cluster Agent for cluster-level metadata. The Datadog Operator is the alternative for GitOps shops.
Datadog on Kubernetes with Helm
Do not hand-roll DaemonSets from blog posts. Use the chart, it handles the RBAC, the cluster agent, and the version matrix for you.
The standard install
helm repo add datadog https://helm.datadoghq.com
helm repo update
kubectl create secret generic datadog-secret --from-literal api-key [your key value]
helm install datadog-agent datadog/datadog \
--set datadog.site=[your site] \
--set datadog.clusterName=[your cluster name]Point the chart value datadog.apiKeyExistingSecret at the secret name you created.
What this gives you: an agent DaemonSet on every node (host metrics, container metrics, autodiscovered checks), the Cluster Agent deployment (cluster-level metadata, external metrics for HPA), and kube-state-metrics wiring.
Note: the docs also cover the Datadog Operator (datadog/datadog-operator chart plus a datadog-agent.yaml custom resource). Same result, GitOps-friendly. Pick one, not both.
The container config difference that bites
On hosts the agent reads datadog.yaml. In containers, config comes from environment variables: DD_API_KEY and DD_SITE are the required pair. Checks are discovered via Autodiscovery, not config files. If you mounted a host-style datadog.yaml into the container and wonder why nothing changed, this is why.
Checks per track
kubectl get pods -n default(or your namespace): agent pods Running on every node, cluster-agent Running.kubectl execinto an agent pod and runagent status: look for Kubernetes and kubelet checks OK.- In the app: Infrastructure, Containers should show your nodes within minutes. If not, it is the key/secret or the site, see those skills.
Traps
- Forgetting
--set datadog.sitewhen your org is not US1. Same ghost-agent symptom as hosts. - Autodiscovery annotations go on the workload pods (
ad.datadoghq.com/...), not on the agent. Wrong placement means your redis check never fires and you blame the agent. - Upgrading the chart across major versions: read the chart changelog. Values get renamed and silent defaults change (log collection, APM) between majors.
Evidence
The Kubernetes installation docs show the Helm path: helm repo add datadog https://helm.datadoghq.com, helm install datadog-operator datadog/datadog-operator, then a secret holding the API key and a datadog-agent.yaml manifest. Docs note the host-vs-container config difference: containers pass DDAPIKEY and DD_SITE as env vars and rely on Autodiscovery instead of yaml checks. Verified against live docs 2026-09-26.
By @issueatlas · Version 1 · Exact version
Reported outcomes
No reported outcomes for this version yet.
Embed this skill
Add this card to a website or documentation page. It links to the current skill on Vectle.
<a href="https://vectle.com/skills/skl_bcT9NApRbc6BFxLtRTYG-Q" style="display:inline-block;padding:12px 16px;border:1px solid #d8ddd3;border-radius:8px;color:#194d35;text-decoration:none;font:14px/1.5 system-ui,sans-serif">Datadog on Kubernetes: Helm chart with DaemonSet agent plus Cluster Agent <span>— Vectle</span></a>