Amazon EKS on AWS Fargate

Supported OS Linux Mac OS Windows

Integration version3.2.0

Overview

Note: This page describes the EKS Fargate integration. For ECS Fargate, see the documentation for Datadog’s ECS Fargate integration.

Amazon EKS on AWS Fargate is a managed Kubernetes service that automates certain aspects of deployment and maintenance for any standard Kubernetes environment. Kubernetes nodes are managed by AWS Fargate and abstracted away from the user.

Setup

These steps cover the setup of the Datadog Agent v7.17+ in a container within Amazon EKS on AWS Fargate. See the Datadog-Amazon EKS integration documentation if you are not using AWS Fargate.

AWS Fargate pods are not physical pods, which means they exclude host-based system-checks, like CPU, memory, etc. In order to collect data from your AWS Fargate pods, you must run the Agent as a sidecar of your application pod with custom RBAC, which enables these features:

  • Kubernetes metrics collection from the pod running your application containers and the Agent
  • Autodiscovery
  • Configuration of custom Agent Checks to target containers in the same pod
  • APM and DogStatsD for containers in the same pod

EC2 Node

If you don’t specify through AWS Fargate Profile that your pods should run on fargate, your pods can use classical EC2 machines. If it’s the case see the Datadog-Amazon EKS integration setup in order to collect data from them. This works by running the Agent as an EC2-type workload. The Agent setup is the same as that of the Kubernetes Agent setup, and all options are available. To deploy the Agent on EC2 nodes, use the DaemonSet setup for the Datadog Agent.

Installation

To get the best observability coverage monitoring workloads in AWS EKS Fargate, install the Datadog integrations for:

Also, set up integrations for any other AWS services you are running with EKS (for example, ELB).

Manual installation

To install, download the custom Agent image: datadog/agent with version v7.17 or above.

If the Agent is running as a sidecar, it can communicate only with containers on the same pod. Run an Agent for every pod you wish to monitor.

Configuration

To collect data from your applications running in AWS EKS Fargate over a Fargate node, follow these setup steps:

To have EKS Fargate containers in the Datadog Live Container View, enable shareProcessNamespace on your pod spec. See Process Collection.

AWS EKS Fargate RBAC

Use the following Agent RBAC when deploying the Agent as a sidecar in AWS EKS Fargate:

apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
  name: datadog-agent
rules:
  - apiGroups:
    - ""
    resources:
    - nodes
    - namespaces
    - endpoints
    verbs:
    - get
    - list
  - apiGroups:
      - ""
    resources:
      - nodes/metrics
      - nodes/spec
      - nodes/stats
      - nodes/proxy
      - nodes/pods
      - nodes/healthz
    verbs:
      - get
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
  name: datadog-agent
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: ClusterRole
  name: datadog-agent
subjects:
  - kind: ServiceAccount
    name: datadog-agent
    namespace: default
---
apiVersion: v1
kind: ServiceAccount
metadata:
  name: datadog-agent
  namespace: default

Running the Agent as a sidecar

To start collecting data from your Fargate type pod, deploy the Datadog Agent v7.17+ as a sidecar of your application. This is the minimum configuration required to collect metrics from your application running in the pod, notice the addition of DD_EKS_FARGATE=true in the manifest to deploy your Datadog Agent sidecar.

apiVersion: apps/v1
kind: Deployment
metadata:
 name: "<APPLICATION_NAME>"
 namespace: default
spec:
 selector:
   matchLabels:
     app: "<APPLICATION_NAME>"
 replicas: 1
 template:
   metadata:
     labels:
       app: "<APPLICATION_NAME>"
     name: "<POD_NAME>"
   spec:
     serviceAccountName: datadog-agent
     containers:
     - name: "<APPLICATION_NAME>"
       image: "<APPLICATION_IMAGE>"
     ## Running the Agent as a side-car
     - image: datadog/agent
       name: datadog-agent
       env:
       - name: DD_API_KEY
         value: "<YOUR_DATADOG_API_KEY>"
         ## Set DD_SITE to "datadoghq.eu" to send your
         ## Agent data to the Datadog EU site
       - name: DD_SITE
         value: "datadoghq.com"
       - name: DD_EKS_FARGATE
         value: "true"
       - name: DD_CLUSTER_NAME
         value: "<CLUSTER_NAME>"
       - name: DD_KUBERNETES_KUBELET_NODENAME
         valueFrom:
           fieldRef:
             apiVersion: v1
             fieldPath: spec.nodeName
      resources:
          requests:
            memory: "256Mi"
            cpu: "200m"
          limits:
            memory: "256Mi"
            cpu: "200m"

Note: Don’t forget to replace <YOUR_DATADOG_API_KEY> with the Datadog API key from your organization.

Note: Add your desired kube_cluster_name:<CLUSTER_NAME> to the list of DD_TAGS to ensure your metrics are tagged by your desired cluster. You can append additional tags here as space separated <KEY>:<VALUE> tags. For Agents 7.34+ and 6.34+, this is not required. Instead, set the DD_CLUSTER_NAME environment variable.

Running the Cluster Agent or the Cluster Checks Runner

Datadog recommends you run the Cluster Agent to access features such as events collection, Kubernetes resources view, and cluster checks.

When using EKS Fargate, there are two possible scenarios depending on whether or not the EKS cluster is running mixed workloads (Fargate/non-Fargate).

If the EKS cluster runs Fargate and non-Fargate workloads, and you want to monitor the non-Fargate workload through Node Agent DaemonSet, add the Cluster Agent/Cluster Checks Runner to this deployment. For more information, see the Cluster Agent Setup.

The Cluster Agent token must be reachable from the Fargate tasks you want to monitor. If you are using the Helm Chart or Datadog Operator, this is not reachable by default because a secret in the target namespace is created.

You have two options for this to work properly:

  • Use an hardcoded token value (clusterAgent.token in Helm, credentials.token in the Datadog Operator); convenient, but less secure.
  • Use a manually-created secret (clusterAgent.tokenExistingSecret in Helm, not available in the Datadog Operator) and replicate it in all namespaces where Fargate tasks need to be monitored; secure, but requires extra operations.

If the EKS cluster runs only Fargate workloads, you need a standalone Cluster Agent deployment. And, as described above, choose one of the two options for making the token reachable.

Use the following Helm values.yaml:

datadog:
  apiKey: <YOUR_DATADOG_API_KEY>
  clusterName: <CLUSTER_NAME>
agents:
  enabled: false
clusterAgent:
  enabled: true
  replicas: 2

In both cases, you need to change the Datadog Agent sidecar manifest in order to allow communication with the Cluster Agent:

       env:
        - name: DD_CLUSTER_AGENT_ENABLED
          value: "true"
        - name: DD_CLUSTER_AGENT_AUTH_TOKEN
          value: <hardcoded token value> # Use valueFrom: if you're using a secret
        - name: DD_CLUSTER_AGENT_URL
          value: https://<CLUSTER_AGENT_SERVICE_NAME>.<CLUSTER_AGENT_SERVICE_NAMESPACE>.svc.cluster.local:5005
        - name: DD_ORCHESTRATOR_EXPLORER_ENABLED # Required to get Kubernetes resources view
          value: "true"
        - name: DD_CLUSTER_NAME
          value: <CLUSTER_NAME>

Cluster performance

For insights into your EKS cluster performance, enable a Cluster Check Runner to collect metrics from the kube-state-metrics service.

Metrics collection

Integration metrics

Use Autodiscovery labels with your application container to start collecting its metrics for the supported Agent integrations.

apiVersion: apps/v1
kind: Deployment
metadata:
 name: "<APPLICATION_NAME>"
 namespace: default
spec:
 replicas: 1
 selector:
   matchLabels:
     app: "<APPLICATION_NAME>"
 template:
   metadata:
     labels:
       app: "<APPLICATION_NAME>"
     name: "<POD_NAME>"
     annotations:
      ad.datadoghq.com/<CONTAINER_NAME>.check_names: '[<CHECK_NAME>]'
      ad.datadoghq.com/<CONTAINER_IDENTIFIER>.init_configs: '[<INIT_CONFIG>]'
      ad.datadoghq.com/<CONTAINER_IDENTIFIER>.instances: '[<INSTANCE_CONFIG>]'
   spec:
     serviceAccountName: datadog-agent
     containers:
     - name: "<APPLICATION_NAME>"
       image: "<APPLICATION_IMAGE>"
     ## Running the Agent as a side-car
     - image: datadog/agent
       name: datadog-agent
       env:
       - name: DD_API_KEY
         value: "<YOUR_DATADOG_API_KEY>"
         ## Set DD_SITE to "datadoghq.eu" to send your
         ## Agent data to the Datadog EU site
       - name: DD_SITE
         value: "datadoghq.com"
       - name: DD_EKS_FARGATE
         value: "true"
       - name: DD_KUBERNETES_KUBELET_NODENAME
         valueFrom:
           fieldRef:
             apiVersion: v1
             fieldPath: spec.nodeName
      resources:
          requests:
            memory: "256Mi"
            cpu: "200m"
          limits:
            memory: "256Mi"
            cpu: "200m"

Notes:

  • Don’t forget to replace <YOUR_DATADOG_API_KEY> with the Datadog API key from your organization.
  • Container metrics are not available in Fargate because the cgroups volume from the host can’t be mounted into the Agent. The Live Containers view reports 0 for CPU and Memory.

DogStatsD

Set up the container port 8125 over your Agent container to forward DogStatsD metrics from your application container to Datadog.

apiVersion: apps/v1
kind: Deployment
metadata:
 name: "<APPLICATION_NAME>"
 namespace: default
spec:
 replicas: 1
 selector:
   matchLabels:
     app: "<APPLICATION_NAME>"
 template:
   metadata:
     labels:
       app: "<APPLICATION_NAME>"
     name: "<POD_NAME>"
   spec:
     serviceAccountName: datadog-agent
     containers:
     - name: "<APPLICATION_NAME>"
       image: "<APPLICATION_IMAGE>"
     ## Running the Agent as a side-car
     - image: datadog/agent
       name: datadog-agent
       ## Enabling port 8125 for DogStatsD metric collection
       ports:
        - containerPort: 8125
          name: dogstatsdport
          protocol: UDP
       env:
       - name: DD_API_KEY
         value: "<YOUR_DATADOG_API_KEY>"
         ## Set DD_SITE to "datadoghq.eu" to send your
         ## Agent data to the Datadog EU site
       - name: DD_SITE
         value: "datadoghq.com"
       - name: DD_EKS_FARGATE
         value: "true"
       - name: DD_KUBERNETES_KUBELET_NODENAME
         valueFrom:
           fieldRef:
             apiVersion: v1
             fieldPath: spec.nodeName
      resources:
          requests:
            memory: "256Mi"
            cpu: "200m"
          limits:
            memory: "256Mi"
            cpu: "200m"

Note: Don’t forget to replace <YOUR_DATADOG_API_KEY> with the Datadog API key from your organization.

Live containers

Datadog Agent v6.19+ supports live containers in the EKS Fargate integration. Live containers appear on the Containers page.

Live processes

Datadog Agent v6.19+ supports live processes in the EKS Fargate integration. Live processes appear on the Processes page. To enable live processes, enable shareProcessNamespace in the pod spec.

Kubernetes resources view

To collect Kubernetes resource views, you need a Cluster Agent setup.

Log collection

Collecting logs from EKS on Fargate with Fluent Bit.

Monitor EKS Fargate logs by using Fluent Bit to route EKS logs to CloudWatch Logs and the Datadog Forwarder to route logs to Datadog.

  1. To configure Fluent Bit to send logs to CloudWatch, create a Kubernetes ConfigMap that specifies CloudWatch Logs as its output. The ConfigMap specifies the log group, region, prefix string, and whether to automatically create the log group.

     kind: ConfigMap
 apiVersion: v1
 metadata:
   name: aws-logging
   namespace: aws-observability
 data:
   output.conf: |
     [OUTPUT]
         Name cloudwatch_logs
         Match   *
         region us-east-1
         log_group_name awslogs-https
         log_stream_prefix awslogs-firelens-example
         auto_create_group true

  2. Use the Datadog Forwarder to collect logs from Cloudwatch and send them to Datadog.

Traces collection

Set up the container port 8126 over your Agent container to collect traces from your application container. Read more about how to set up tracing.

apiVersion: apps/v1
kind: Deployment
metadata:
 name: "<APPLICATION_NAME>"
 namespace: default
spec:
 replicas: 1
 selector:
   matchLabels:
     app: "<APPLICATION_NAME>"
 template:
   metadata:
     labels:
       app: "<APPLICATION_NAME>"
     name: "<POD_NAME>"
   spec:
     serviceAccountName: datadog-agent
     ## Putting the agent in the same namespace as the application for origin detection with cgroup v2
     shareProcessNamespace: true
     containers:
     - name: "<APPLICATION_NAME>"
       image: "<APPLICATION_IMAGE>"
     ## Running the Agent as a side-car
     - image: datadog/agent
       name: datadog-agent
       ## Enabling port 8126 for Trace collection
       ports:
        - containerPort: 8126
          name: traceport
          protocol: TCP
       env:
       - name: DD_API_KEY
         value: "<YOUR_DATADOG_API_KEY>"
         ## Set DD_SITE to "datadoghq.eu" to send your
         ## Agent data to the Datadog EU site
       - name: DD_SITE
         value: "datadoghq.com"
       - name: DD_EKS_FARGATE
         value: "true"
       - name: DD_APM_ENABLED
         value: "true"
       - name: DD_KUBERNETES_KUBELET_NODENAME
         valueFrom:
           fieldRef:
             apiVersion: v1
             fieldPath: spec.nodeName
      resources:
          requests:
            memory: "256Mi"
            cpu: "200m"
          limits:
            memory: "256Mi"
            cpu: "200m"

Note: Don’t forget to replace <YOUR_DATADOG_API_KEY> with the Datadog API key from your organization.

Events collection

To collect events from your AWS EKS Fargate API server, run a Datadog Cluster Agent within your EKS cluster and Enable Event collection for your Cluster Agent.

Optionally, deploy cluster check runners in addition to setting up the Datadog Cluster Agent to enable cluster checks.

Note: You can also collect events if you run the Datadog Cluster Agent in a pod in Fargate.

Process collection

For Agent 6.19+/7.19+, Process Collection is available. Enable shareProcessNamespace on your pod spec to collect all processes running on your Fargate pod. For example:

apiVersion: v1
kind: Pod
metadata:
  name: <NAME>
spec:
  shareProcessNamespace: true
...

Note: CPU and memory metrics are not available.

Data Collected

Metrics

The eks_fargate check submits a heartbeat metric eks.fargate.pods.running that is tagged by pod_name and virtual_node so you can keep track of how many pods are running.

Service Checks

eks_fargate does not include any service checks.

Events

eks_fargate does not include any events.

Troubleshooting

Need help? Contact Datadog support.

Further Reading

Additional helpful documentation, links, and articles: