Daemonsets

• Replicasets and Deployments allow multiple copies of the application to be available across the worker nodes.
• DaemonSets are like Replicasets - they allow you to deploy multiple instances of pods.
  • It runs one copy of the pod on every node in the cluster.
• When a Node is removed, the pod is also automatically removed.
• The DaemonSet ensures that one copy of the pod is present across all nodes in the cluster.
• Good use cases are a Monitoring Solution / Logs Viewer
  • Deploys the monitoring agent in the form of a pod on all nodes in the cluster.
• Another component that is requires on every node in the cluster is kube-proxy
  • DaemonSets can also be used for this purpose.
• Networking is another good idea and an agent can be deployed on every node in the cluster.
• Creating a DaemonSet is similar to the replicaset creation process.
• Example Daemon Set:

  apiVersion: apps/v1
  kind: DaemonSet
  metadata:
   name: monitoring-daemon
  spec:
  selector:
    matchLabels:
      app: monitoring-agent
  template:
    metadata:
      labels:
        app: monitoring-agent
    spec:
      containers:
      - name: monitoring-agent
        image: monitoring-agent
  

• Exactly the same as a Replicaset definition, except kind is set as DaemonSet.
• Then create it with kubectl create -f <yaml>.
• View the command using kubectl get daemonsets
• To see more detail about a particular daemonset command, use kubectl describe daemonsets <daemonset_name>
• We can replace the node_name property on a pod to bypass the scheduler and place a pod on a node directly.
  • This was in version 1.12
• From v1.12 onwards, NodeAffinity and the default scheduler are used.