This tool will auto discover kubernetes cluster nodes and send container and machine metrics to SignalFx. You can deploy it to run in a single pod within a Kubernetes cluster where it will collect data from the cadvisor container resource usage and performance analysis agent (that is integrated into the kubelet binary) using cluster member permissions only.

Step 1 - Create a deployment configuration based on the following example cadvisor-signalfx.yaml file. Note: Change the <API_TOKEN> based on your SignalFx account and the <CLUSTER_NAME> based on what you want to call this kubernetes cluster.

apiVersion: v1
kind: ReplicationController
metadata:
  name: "cadvisor-signalfx"
spec:
  replicas: 1
  selector:
    app: "cadvisor-signalfx"
  template:
    metadata:
      name: "cadvisor-signalfx"
      labels:
        app: "cadvisor-signalfx"
    spec:
      containers:
      - name: "cadvisor-signalfx"
        image: "quay.io/signalfx/cadvisor-integration:latest"
        env:
        - name: SFX_SCRAPPER_API_TOKEN
          value: <API TOKEN>
        - name: SFX_SCRAPPER_CLUSTER_NAME
          value: <CLUSTER NAME>
        - name: SFX_SCRAPPER_SEND_RATE
          value: 5s

Step 2 - Deploy to your cluster. e.g. kubectl create -f cadvisor-signalfx.yaml

The default configuration for this integration uses the credentials of the pod to communicate securely with the kubernetes apiserver. For more details see "Accessing the API from a Pod" section at http://kubernetes.io/docs/user-guide/accessing-the-cluster

However, if the kubernetes cluster is not setup to support accessing the api from a pod then an extra configuration is required. Add an environment variable for the IP (and port, if necessary) of the Kubernetes Master Node API to the very end of the cadvisor-signalfx.yaml file.

Example:

name:  SFX_SCRAPPER_KUBERNETES_URL
value: <http(s)://KUBERNETES_MASTER_NODE_IP_ADDR:port>

This tool is licensed under the Apache License, Version 2.0. See LICENSE for full license text.