All Products
Search

This Product

    Container Service for Kubernetes:Create a DaemonSet

    Document Center

    Container Service for Kubernetes:Create a DaemonSet

    Last Updated:Mar 26, 2026

    Node-level services such as log collectors and monitoring agents must run on every node in your cluster. A DaemonSet ensures exactly one pod runs on each node: when a node joins the cluster, the DaemonSet creates a pod on it automatically; when a node is removed, the pod is cleaned up. This topic explains how DaemonSet scheduling works and how to create one using the console or kubectl.

    Common use cases:

    • Log collection daemons (for example, Fluentd)

    • Node monitoring agents (for example, Prometheus Node Exporter)

    For workloads that require replica counts or advanced scheduling beyond one-pod-per-node placement, use a Deployment instead. For the full DaemonSet specification, see the Kubernetes documentation.

    How scheduling works

    By default, a DaemonSet places one pod on every node. Three mechanisms can restrict which nodes receive pods.

    Taints and tolerations

    DaemonSet pods respect node taints. A pod does not run on a node if it cannot tolerate the node's taints. The following tolerations are added to DaemonSet pods automatically:

    Toleration keyEffectBehavior
    node.kubernetes.io/unschedulableNoSchedulePods are scheduled on unschedulable nodes.
    node.kubernetes.io/not-readyNoExecutePods run on nodes that are not ready. Running pods are not evicted for 300 seconds.
    node.kubernetes.io/unreachableNoExecutePods run on nodes that are unreachable. Running pods are not evicted for 300 seconds.

    nodeSelector

    If a DaemonSet includes a nodeSelector, pods run only on nodes that match the label. For example, nodeSelector: { disktype: ssd } limits the DaemonSet to nodes labeled disktype=ssd.

    Affinity and anti-affinity

    Node affinity, pod affinity, and pod anti-affinity rules also apply to DaemonSet pods.

    Prerequisites

    Before you begin, ensure that you have:

    Create a DaemonSet using the console

    1. Log on to the Container Service Management Console. In the left navigation pane, click Clusters.

    2. On the Clusters page, click the name of your cluster. In the left navigation pane, choose Workloads > DaemonSets.

    3. On the DaemonSets page, click Create from Image.

    4. Configure the DaemonSet. The form is identical to a Deployment with two differences: For all other configuration options, see Create a Deployment.

      • Basic Information: No Replicas setting. The number of pods is determined by the number of nodes.

      • Advanced: No Scaling setting.

    Create a DaemonSet using kubectl

    1. Save the following YAML to a file named daemonset.yaml.

      apiVersion: apps/v1
kind: DaemonSet
metadata:
  name: nginx-test
  namespace: default  # Change the namespace as needed.
  labels:
    app: nginx
spec:
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
      - name: nginx
        image: anolis-registry.cn-zhangjiakou.cr.aliyuncs.com/openanolis/nginx:1.14.1-8.6
        resources:
          limits:
            cpu: '1'
            memory: 2Gi
          requests:
            cpu: 500m
            memory: 512Mi

    2. Apply the manifest.

      kubectl apply -f daemonset.yaml

      Expected output:

      daemonset.apps/nginx-test created

    3. Verify that a pod is running on each node.

      kubectl get pods --all-namespaces -o wide | grep nginx-test

      The output lists one pod per node, each assigned to a different node IP address.

      default     nginx-test-8mqvh     1/1     Running     0          3m38s   192.168.*.**    cn-shanghai.192.168.**.250   <none>           <none>
default     nginx-test-ltlx6     1/1     Running     0          3m38s   192.168.*.**    cn-shanghai.192.168.**.98    <none>           <none>
default     nginx-test-n6zrv     1/1     Running     0          3m38s   192.168.*.**    cn-shanghai.192.168.**.17    <none>           <none>

    Example: view kube-proxy pods

    The Kubernetes component kube-proxy is deployed as a DaemonSet, with one pod per node. Run the following command to inspect it:

    kubectl get pods --all-namespaces -o wide | grep kube-proxy

    Expected output:

    kube-system     kube-proxy-worker-hfzkh     1/1     Running     0          2d21h   192.168.*.92    cn-shanghai.192.168.*.92   <none>           <none>
kube-system     kube-proxy-worker-pxnnf     1/1     Running     0          2d21h   192.168.*.11    cn-shanghai.192.168.*.11   <none>           <none>
kube-system     kube-proxy-worker-r2t26     1/1     Running     0          2d21h   192.168.*.7     cn-shanghai.192.168.*.7    <none>           <none>

    Because kube-proxy uses hostNetwork: true, each pod's IP address matches its node's IP address.

    What's next