Demo Manual Reconciliation

In this lesson we perform a manual reconciliation of an Argo CD application to illustrate why making ad-hoc changes directly in the Kubernetes cluster results in an OutOfSync state. You will:

Confirm cluster namespaces.
Synchronize an Argo CD application from Git.
Inspect the resulting live resources.
Make an intentional manual change in-cluster to cause drift.
Observe Argo CD detecting the OutOfSync state and restore the desired state by synchronizing from Git.

1. Confirm existing namespaces

First, verify which namespaces are present in the cluster:

# kubectl get namespaces
NAME                 STATUS   AGE
argocd               Active   16m
default              Active   3h1m
kube-node-lease      Active   3h1m
kube-public          Active   3h1m
kube-system          Active   3h1m

2. Synchronize the Argo CD application

In the Argo CD web UI, click “Synchronize” for the application. Choose to auto-create the namespace if prompted and synchronize both the Deployment and Service manifests. Argo CD will apply those manifests and report the created ReplicaSet and Service details (such as the NodePort).

The image shows an interface for ArgoCD, displaying the status and details of an application called "highway-animation," which is healthy and synced to the latest commit. A visual workflow represents the deployment process from service to pod.

You can inspect logs, events, and a summary of the applied YAML from the Argo CD UI.

3. Inspect the created Pod manifest (live cluster)

Here is the Pod manifest that was created by the Deployment in the cluster:

apiVersion: v1
kind: Pod
metadata:
  creationTimestamp: '2025-07-29T08:40:03Z'
  generateName: highway-animation-c5ccdf6b-
  labels:
    app: highway-animation
    pod-template-hash: c5ccdf6b
  name: highway-animation-c5ccdf6b-5pl4d
  namespace: highway-animation
  ownerReferences:
    - apiVersion: apps/v1
      blockOwnerDeletion: true
      controller: true
      kind: ReplicaSet
      name: highway-animation-c5ccdf6b
      uid: 175a22f1-a852-4b16-b992-8bf08f8793a52
  resourceVersion: '15018'
  uid: 57c6c5a5-3497-4104-87c2-26cb35703be2
spec:
  containers:
    - env:
        - name: POD_COUNT
          value: '1'
      image: siddharth67/highway-animation:blue
      imagePullPolicy: IfNotPresent
      name: highway-animation
      ports:
        - containerPort: 3000
          protocol: TCP
      resources: {}
      terminationMessagePath: /dev/termination-log
      terminationMessagePolicy: File

4. Confirm the new namespace and resources

After synchronization, the highway-animation namespace should exist:

# kubectl get namespaces
NAME                  STATUS   AGE
argocd                Active   17m
default               Active   3h2m
highway-animation     Active   35s
kube-node-lease       Active   3h2m
kube-public           Active   3h2m
kube-system           Active   3h2m

List all resources in the highway-animation namespace:

# kubectl -n highway-animation get all
NAME                                      READY   STATUS    RESTARTS   AGE
pod/highway-animation-c5ccdf6b-5pl4d     1/1     Running   0          47s

NAME                                  TYPE       CLUSTER-IP       EXTERNAL-IP   PORT(S)            AGE
service/highway-animation-service     NodePort   10.100.200.170   <none>        3000:32000/TCP     47s

NAME                                  READY   UP-TO-DATE   AVAILABLE   AGE
deployment.apps/highway-animation     1/1     1            1           47s

NAME                                      DESIRED   CURRENT   READY   AGE
replicaSet.apps/highway-animation-c5ccdf6b   1         1         1       47s

The app is reachable via the NodePort (e.g., http://localhost:32000) and will render the highway animation. With the Deployment set to 1 replica, you will see one vehicle.

5. Create drift: manually scale the Deployment in-cluster

In a non-GitOps workflow, an operator might directly edit the Deployment in the cluster to increase replicas:

# kubectl -n highway-animation edit deployment highway-animation

Make these edits in the live Deployment manifest:

Change replicas: 1 to replicas: 5
Update the container environment variable POD_COUNT value from '1' to '5'

After saving, Kubernetes will create 4 additional pods and the app will show five vehicles. Important: This manual change modifies the live cluster state but does not update the Git repository. That causes the live state to diverge from the desired state declared in Git.

6. Example: live (cluster) Deployment after manual edit

Live (cluster) Deployment snippet showing the manual change:

apiVersion: apps/v1
kind: Deployment
metadata:
  generation: 2
  name: highway-animation
  namespace: highway-animation
spec:
  progressDeadlineSeconds: 600
  replicas: 5
  revisionHistoryLimit: 10
  selector:
    matchLabels:
      app: highway-animation
  strategy:
    rollingUpdate:
      maxSurge: 25%
      maxUnavailable: 25%
  template:
    spec:
      containers:
        - name: highway-animation
          image: siddharth67/highway-animation:blue
          env:
            - name: POD_COUNT
              value: '5'

7. Desired (Git) Deployment in the repository

The desired manifest stored in Git still declares 1 replica:

apiVersion: apps/v1
kind: Deployment
metadata:
  name: highway-animation
  namespace: highway-animation
spec:
  progressDeadlineSeconds: 600
  replicas: 1
  revisionHistoryLimit: 10
  selector:
    matchLabels:
      app: highway-animation
  strategy:
    rollingUpdate:
      maxSurge: 25%
      maxUnavailable: 25%
  template:
    spec:
      containers:
        - name: highway-animation
          image: siddharth67/highway-animation:blue
          env:
            - name: POD_COUNT
              value: '1'

Because the live cluster has 5 replicas while Git declares 1, Argo CD will detect a difference and mark the application OutOfSync. In the Argo CD UI you can click “Diff” to view the highlighted mismatches between the live and desired manifests.

8. Reconcile back to Git (manual sync)

To restore the cluster to the Git-declared desired state, click “Synchronize” in Argo CD. Argo CD will reconcile the resources and update the live cluster to match the manifests in the repository. In this scenario it will update the Deployment (not the Service) to revert the replica count back to 1.

The image shows an Argo CD dashboard displaying the status and configuration of a Kubernetes application named "highway-animation," indicating a healthy sync status with a visual representation of its components and pods.

After synchronization, Argo CD will revert the cluster to the Git desired state and the application will again show a single pod/vehicle.

Quick reference: kubectl & Argo CD actions

Action	Command or UI step	Purpose
List namespaces	`kubectl get namespaces`	Verify cluster namespaces
List resources in namespace	`kubectl -n highway-animation get all`	Inspect deployed resources
Edit a live Deployment	`kubectl -n highway-animation edit deployment highway-animation`	Make ad-hoc changes (not recommended)
Sync app in Argo CD	Argo CD UI → Select app → “Synchronize”	Apply Git manifests to cluster
View diff in Argo CD	Argo CD UI → Select app → “Diff”	Compare live vs. desired manifests

Best practice

Never make long-lived configuration changes directly in the cluster. Always update the Git repository with the desired state and let your GitOps operator (Argo CD) reconcile the cluster. This ensures a single source of truth, prevents configuration drift, and makes rollbacks and audits straightforward.

Links and references

Argo CD documentation: https://argo-cd.readthedocs.io/
Kubernetes: Deployments — https://kubernetes.io/docs/concepts/workloads/controllers/deployment/
GitOps principles: https://www.weave.works/technologies/gitops/

Introduction

GitOps Terminology

GitOps Principles

GitOps Patterns

Tooling

Related Practices

Demo Manual Reconciliation

1. Confirm existing namespaces

2. Synchronize the Argo CD application

3. Inspect the created Pod manifest (live cluster)

4. Confirm the new namespace and resources

5. Create drift: manually scale the Deployment in-cluster

6. Example: live (cluster) Deployment after manual edit

7. Desired (Git) Deployment in the repository

8. Reconcile back to Git (manual sync)

Quick reference: kubectl & Argo CD actions

Best practice

Links and references

Watch Video

​1. Confirm existing namespaces

​2. Synchronize the Argo CD application

​3. Inspect the created Pod manifest (live cluster)

​4. Confirm the new namespace and resources

​5. Create drift: manually scale the Deployment in-cluster

​6. Example: live (cluster) Deployment after manual edit

​7. Desired (Git) Deployment in the repository

​8. Reconcile back to Git (manual sync)

​Quick reference: kubectl & Argo CD actions

​Best practice

​Links and references

Watch Video

1. Confirm existing namespaces

2. Synchronize the Argo CD application

3. Inspect the created Pod manifest (live cluster)

4. Confirm the new namespace and resources

5. Create drift: manually scale the Deployment in-cluster

6. Example: live (cluster) Deployment after manual edit

7. Desired (Git) Deployment in the repository

8. Reconcile back to Git (manual sync)

Quick reference: kubectl & Argo CD actions

Best practice

Links and references