My Daemonset pod is only created on the master node

yemmey

I created a Daemonset from the ds.yaml file, but it created only 1 pod in the master node instead of 1 pod in each node. below is my ds.yaml file.

apiVersion: apps/v1
kind: DaemonSet
metadata:
name: ds-one
spec:
selector:
matchLabels:
system: DaemonSetOne
template:
metadata:
labels:
system: DaemonSetOne
spec:
containers:
- name: nginx
image: nginx:1.15.1
ports:
- containerPort: 80

alihasanahmedk

give label to your each node.
for example:
kubectl label node master system=daemonsetOne
kubectl label node worker system=daemonsetOne

Then add nodeSelector in your yaml file.

apiVersion: apps/v1
kind: DaemonSet
metadata:
  name: ds-one
spec:
  selector:
    matchLabels:
      system: DaemonSetOne
  template:
    metadata:
      labels:
        system: DaemonSetOne
      spec:
        containers:
        - name: nginx
          image: nginx:1.15.1
          ports:
          - containerPort: 80
nodeSelector:
  system: daemonsetOne

chrispokorni

Hi @yemmey,

The DaemonSet operator is expected to schedule one Pod on each node, without the need of a nodeSelector. The use of the nodeSelector property helps when a DaemonSet is expected to schedule Pods on a subset of the cluster's nodes, which have been pre-labeled accordingly.

The Pods of a DaemonSet may not be scheduled on a node when the node is tainted, not ready, when it is unschedulable, or when dependencies cannot be met.

What is the output of the kubectl get nodes -o wide command?

Regards,
-Chris

yemmey

chrispokorni

Hi @yemmey,

This output seems ok.
What output is produced by:

kubectl get nodes --show-labels

kubectl get pod -A

and

kubectl describe node | grep -i taint

Regards,
-Chris

yemmey

Apologies for the late response, here is the feedback below.

kubectl get nodes --show-labels:

kubectl get pod -A

kubectl describe node | grep -i taint

chrispokorni

Hi @yemmey,

The output displaying taints is not as expected, according to the installation and configuration steps from lab exercise 3.3. A tainted control-plane node will prevent our pods from being scheduled on the control-plane node unless presenting a matching toleration. This is the reason why the DaemonSet operator only shows one running pod, instead of two.
Please revisit lab exercise 3.3 to remove the taint.

Regards,
-Chris

yemmey

Got it! Thanks so much