Kubernetes(k8s) Architecture - Distributed System
Kubernetes is a distributed system that horizontally scales containers across multiple physical hosts termed Nodes.
This produces fault-tolerant deployments which adapt to conditions such as Node resource pressure, instability, and elevated external traffic levels. If one Node suffers
an outage, Kubernetes can reschedule your containers onto neighboring healthy Nodes.
Let's stop treating Kubernetes(k8s) as black box. You can't troubleshoot what you don't understand. So, here's how Kubernetes actually works - from the inside out.

API Server - The Brain
The central access point for all control plane interactions. It receives API requests, validates them, and updates the cluster state by interacting with other
components. Every kubectl command you run hits this. It does following:
- It's the front door to your cluster
- Validates requests (authz + admission)
- Talks to etcd (the memory)
- Sends changes to all other components
If your API Server is down, K8s is blind. You can't schedule, scale, or patch anything.
etcd - The Memory
A distributed key value store used to persist all cluster data. It stores the desired and current state of Kuberates resources.Think of it like a version-controlled
database of your entire cluster's state. It is used as:
- Highly available key-value store
- Every Pod, ConfigMap, Secret and Node lives here
etcd corruption = cluster amnesia.
Controller Manager - The Conductor
Runs background loops that ensure the cluster's desired state is maintained, It manages controllers responsible for nodes, replicas, jobs, and more. It watches the orchestra,
makes sure everyone plays the right note. It does following:
- Ensures Deployments have the right number of pods
- Jobs finish, CronJobs run
- Node controllers remove dead ones
If something disappears, controller brings it back, That's reconciliation.
Scheduler - The Smart Assigner
Decides which node should run a newly created Pod. It considers resource needs, constraints, and policies to make placement decisions. Decides where your pod
runs. It does following:
- Looks at resource requests, affinity/taints/selector, node status
- Picks the best-fit node
No scheduler = no new workloads
Kubelet - The Node Agent
A node agent running on every noce in the cluster. It ensures that containers described in Pod specs are running and reports their status. Lives inside every
worker node. It does following:
- Talks to the API server
- Watches pods assigned to the node
- Starts containers via container runtime
If kubelet dies, that node is essentially offline to the cluster.
Kube Proxy - The Network Bridge
Manages networking rules on each node. It enables network communication between services and Pods across the cluster. Handles service-to-pod networking.
It does following:
- Sets up iptables/ipvs rules
- Enables service discovery via DNS
- Route traffic across pods/nodes
Broken kube-proxy = your service exists, but can't talk to pods.
Summary and Conclusions
Kubernetes isn't magic. It's just a well-coordinated distributed system with clearly defined tools.
- API Server --> Command center
- etcd --> Memory
- Controller --> Reconciler
- Scheduler --> Planner
- Kubelet --> Executor
- Proxy --> Networker

Author

Sagar Mehta is Atgen Software Solutions Founder and a recognised expert in the field of Intelligent Automation, including Robotic Process Automation, Workload Automation, DevOps, SRE and Advanced Analytics. Sagar advocates a pragmatic approach to Automation, encouraging a policy of using ‘the best tool for the job’.
Prior to co-founding Atgen Software Solutions, Sagar worked in Senior Automation roles, architecting and delivering robust, scalable solutions for many of the world’s biggest banks and working with leading Automation vendors. He developed his first automated solution in 2006 and has continued to deliver robust, scalable and sophisticated Automation ever since.
Sagar is a regular guest speaker and panellist at Automation seminars, conferences and user group events.
Contact
Have a similar problem to solve, let's work together.
Our Address
#107, Tower B, Escon Arena, Zirakpur, Punjab, India - 140603
Email Us
info@atgensoft.com
Call Us
+91-8806666141