LearnGlossary

What is a DaemonSet in Kubernetes?

Updated 25 October 2026 · Published by Zoorik

A DaemonSet in Kubernetes runs one copy of a pod on each node. A new node gets its copy. When a node leaves, Kubernetes removes its copy.

Every node runs one copy, and a new node gets its own Illustration of one DaemonSet and 4 nodes. A box at the top reads DaemonSet, log collector. A line from it splits into 3 arrows, labelled runs one copy on each node, one into a log collector pod box inside each of Node 1, Node 2 and Node 3. A second arrow from the DaemonSet, labelled adds a pod, points into a new log collector pod box inside the fourth node, labelled Node 4 joins. Each node holds 1 copy, so 4 nodes hold 4 copies. Every node runs one copy, and a new node gets its own DaemonSetlog collectorruns one copy on each nodeadds a podlog collector podNode 1log collector podNode 2log collector podNode 3log collector podNode 4 joins
Illustration: a DaemonSet keeps one copy of its pod on each node, and adds one when a node joins.

Some nodes get no copy:

  • A node that doesn’t match its node selector or node affinity.
  • A node with a NoSchedule or NoExecute taint the pod doesn’t tolerate. A taint keeps pods away.

The use of a DaemonSet is to run a helper on each node. The docs name three:

  • cluster storage
  • log collection
  • node monitoring

A Kubernetes DaemonSet example

The docs’ example runs the fluentd-elasticsearch image. Each copy requests 100m of CPU, or 0.1 CPU, and 200Mi of memory, or 200 mebibytes.

DaemonSet vs Deployment

Both keep pods running.

  • A Deployment runs as many copies as you set, on whatever nodes fit.
  • A DaemonSet suits a helper that must run on all nodes, or on some.

Why it matters for your bill

The scheduler counts each pod’s request against its node, so a DaemonSet’s request takes room on every node you pay for. A container with only a limit usually gets that limit as its request.

Say the docs’ example runs on 10 nodes. Its requests total 1 CPU and 2,000Mi, or 0.6 CPU and 1,200Mi on 6 nodes.

  • Coral rebalances CPU and memory requests on AKS and EKS to match real use, so the same pods fit on fewer nodes.
  • It drains spare nodes so your autoscaler can remove them, and it never moves DaemonSet pods.
  • The free assessment works out your own figure.

Read next: Kubernetes node size best practices: large or small nodes

Sources: Kubernetes, “DaemonSet” (opens in a new tab); Kubernetes, “Resource Management for Pods and Containers” (opens in a new tab); Kubernetes, “Taints and Tolerations” (opens in a new tab); Kubernetes, “Kubelet Configuration (v1beta1)” (opens in a new tab).

Find your own number

One line in your cloud console. One PDF report with your estimated saving at list price.

Estimate your saving

Talk to the builders

Thirty minutes with the people who build Coral.

Book a demo (opens in a new tab)

Or write to founders@zoorik.com