CKA Troubleshooting Walkthrough

Deployment Scaled to 2 Replicas but Only 1 Exists

Lab Scenario

The deployment pink-depl-cka14-trb was expected to run 2 replicas.

However:

kubectl get deploy pink-depl-cka14-trb

NAME                  READY   UP-TO-DATE   AVAILABLE
pink-depl-cka14-trb   0/2     1            0

The deployment wanted 2 replicas, but only 1 replica was being managed.

Initial Clues

Deployment YAML:

spec:
  replicas: 2

ReplicaSet:

kubectl describe rs pink-depl-cka14-trb-df6c5bb49

Replicas: 1 current / 1 desired
Annotations:
deployment.kubernetes.io/desired-replicas: 1
deployment.kubernetes.io/max-replicas: 2

Red Flag:

Deployment wants 2 replicas.
ReplicaSet only wants 1 replica.
Those values should normally match.

Important Discovery

Checking deployment status revealed:

metadata:
  generation: 3

status:
  observedGeneration: 1

Generation was 3, but observedGeneration was still 1.

This means the Deployment Controller had not processed the latest changes.

The deployment object was being updated, but the controller responsible for reconciling Deployments wasn't working correctly.

Verify Controller Manager Health

kubectl get pods -n kube-system | grep controller

Output:

kube-controller-manager-cluster4-controlplane   0/1   CrashLoopBackOff

This immediately pointed to the control plane rather than the deployment itself.

Root Cause

Inspecting the static pod manifest:

sudo cat /etc/kubernetes/manifests/kube-controller-manager.yaml

Found a typo:

command:
- kube-controller-manage

Correct value:

command:
- kube-controller-manager

A single missing "r" caused the kube-controller-manager to crash continuously.

Without the controller manager running:

Deployments cannot reconcile desired state.
ReplicaSets are not scaled correctly.
Generation updates are not observed.
Controllers stop performing their normal duties.

Fix

sudo vi /etc/kubernetes/manifests/kube-controller-manager.yaml

Replace:

kube-controller-manage

With:

kube-controller-manager

Save the file and wait for the kubelet to recreate the static pod.

Verify:

kubectl get pods -n kube-system | grep controller

Expected:

kube-controller-manager-cluster4-controlplane   1/1   Running

Final Verification

kubectl get deploy pink-depl-cka14-trb

Expected:

NAME                  READY   UP-TO-DATE   AVAILABLE
pink-depl-cka14-trb   2/2     2            2

The deployment controller observes the latest generation, scales the ReplicaSet, and creates the missing pod.

CKA Exam Takeaway

If you ever see:

spec.replicas: 2

status.replicas: 1

generation: 3
observedGeneration: 1

Think beyond the deployment itself and inspect the control plane.

A mismatch between generation and observedGeneration is a strong indicator that the controller responsible for processing that resource is unhealthy.