Kubernetes

How to answer in an interview

Kubernetes is an open-source container orchestration platform that automates the deployment, scaling, and management of containerized applications across a cluster of machines. It handles tasks like scheduling containers, self-healing failed instances, load balancing, and rolling updates, abstracting away much of the operational complexity of running containers at scale.

Explain like I am 5

Kubernetes is like an automatic manager for a bunch of boxes (containers), deciding where they go, restarting them if they break, and making sure there are always enough copies running.

How to answer in an interview

A pod is the smallest deployable unit in Kubernetes, representing one or more tightly coupled containers that share the same network namespace and storage volumes. Containers within the same pod can communicate via localhost and are always scheduled together on the same node.

Explain like I am 5

A pod is like a small shared apartment where one or a few containers live together, sharing the same address and utilities, and always moving in and out of the building at the same time.

How to answer in an interview

A container is a single running instance of an application packaged with its dependencies. A pod is a Kubernetes abstraction that wraps one or more containers, providing them a shared network and storage context, and it's the actual unit Kubernetes schedules and manages, rather than individual containers directly.

Explain like I am 5

A container is one single box. A pod is the shared room that holds one or more boxes together, and Kubernetes only manages rooms, not individual boxes directly.

How to answer in an interview

A namespace is a way to logically partition a Kubernetes cluster into multiple virtual clusters, allowing resource isolation between different teams, projects, or environments within the same physical cluster, along with the ability to set resource quotas per namespace.

Explain like I am 5

A namespace is like dividing one big office building into separate labeled sections for different teams, so each team's stuff stays organized and separate, even though they're all in the same building.

How to answer in an interview

kubectl is the command-line tool used to interact with a Kubernetes cluster, allowing users to deploy applications, inspect and manage cluster resources, view logs, and troubleshoot issues by communicating with the cluster's API server.

Explain like I am 5

kubectl is the remote control you use to talk to and manage your Kubernetes cluster from your computer's terminal.

How to answer in an interview

The control plane manages the overall cluster state and includes components like the API server, which handles all requests; etcd, a key-value store holding cluster state; the scheduler, which assigns pods to nodes; and the controller manager, which ensures the actual state matches the desired state. Worker nodes run the actual application workloads via the kubelet agent and container runtime, communicating with the control plane.

Explain like I am 5

The control plane is like the head office making decisions and keeping records, while the worker nodes are like the branches actually doing the work, following instructions sent from the head office.

How to answer in an interview

A Deployment is a Kubernetes object that manages a set of identical pods, ensuring a specified number of replicas are running at all times, and it manages rolling updates and rollbacks to safely update application versions without downtime. It automatically creates and manages an underlying ReplicaSet.

Explain like I am 5

A Deployment is like giving Kubernetes a standing instruction: 'always keep exactly this many copies of my app running, and update them smoothly whenever I release a new version.'

How to answer in an interview

A Service provides a stable network endpoint and DNS name for a set of pods, load balancing traffic across them even as individual pods are created, destroyed, or rescheduled with changing IP addresses. It uses label selectors to determine which pods it routes traffic to.

Explain like I am 5

A Service is like a single phone number that always reaches the right department, even though the specific person answering the phone might change from call to call.

How to answer in an interview

A ConfigMap stores non-sensitive configuration data, like environment variables or config files, separately from application code, so configuration can change without rebuilding container images. A Secret is similar but designed for sensitive data like passwords or API keys, stored in a base64-encoded format with additional access controls.

Explain like I am 5

A ConfigMap is like a settings sheet you can update without rebuilding the whole app. A Secret is the same idea but specifically for sensitive stuff like passwords, handled a bit more carefully.

How to answer in an interview

A ReplicaSet ensures a specified number of identical pod replicas are running at any given time, automatically creating new pods to replace any that fail or are deleted. In practice, ReplicaSets are usually managed indirectly through Deployments rather than created directly.

Explain like I am 5

A ReplicaSet is like a rule that says 'always keep exactly five copies of this running,' automatically replacing any copy that disappears.

How to answer in an interview

A rolling update gradually replaces old pod versions with new ones, a few at a time, ensuring the application remains available throughout the update with zero downtime. If issues are detected, Kubernetes allows an easy rollback to the previous stable version using the deployment's revision history.

Explain like I am 5

A rolling update is like swapping out old team members for new ones a few at a time, so the team never stops working entirely. A rollback is quickly bringing back the old team if the new ones aren't working out.

How to answer in an interview

A liveness probe checks whether a container is still running correctly, and if it fails, Kubernetes restarts the container. A readiness probe checks whether a container is ready to accept traffic, and if it fails, Kubernetes temporarily removes the pod from a service's load balancing pool without restarting it.

Explain like I am 5

A liveness probe asks 'are you still alive?' and restarts you if you're not. A readiness probe asks 'are you ready for customers yet?' and just pauses sending you customers if you're not, without restarting you.

How to answer in an interview

An Ingress manages external HTTP and HTTPS access to services within a cluster, providing features like path-based and host-based routing, SSL termination, and load balancing, all managed through a single entry point rather than exposing each service individually. It requires an Ingress controller running in the cluster to actually implement the routing rules.

Explain like I am 5

An Ingress is like a smart receptionist at the front door of the cluster, reading the visitor's request and directing them to exactly the right internal department based on the web address they asked for.

How to answer in an interview

Self-healing refers to Kubernetes' ability to automatically detect and recover from failures, such as restarting containers that crash, rescheduling pods from failed nodes onto healthy ones, and replacing pods that fail health checks, continuously reconciling the actual cluster state with the declared desired state.

Explain like I am 5

Self-healing means Kubernetes constantly checks that everything matches what you asked for, and automatically fixes things itself, like replacing a broken pod, without needing a human to notice and step in.

How to answer in an interview

ClusterIP, the default type, exposes a service only internally within the cluster. NodePort exposes the service on a static port on each node's IP, making it reachable from outside the cluster. LoadBalancer provisions an external load balancer, typically through a cloud provider, giving the service a single external IP that distributes traffic into the cluster.

Explain like I am 5

ClusterIP is like an internal office extension only staff can call. NodePort opens up a specific door on every building for outsiders too. LoadBalancer gives you one clean public phone number that automatically routes to the right place.

How to answer in an interview

Horizontal Pod Autoscaling (HPA) automatically adjusts the number of pod replicas in a Deployment based on observed metrics, such as CPU or memory utilization, or custom metrics, scaling out during high demand and scaling back in during low demand to optimize resource usage and cost.

Explain like I am 5

Horizontal pod autoscaling is like automatically adding more cashiers when the store gets busy, and sending some home again once things quiet down.

How to answer in an interview

A StatefulSet manages pods that require stable, unique network identities and persistent storage that survives rescheduling, such as databases, and it deploys and scales pods in a defined, ordered sequence. A Deployment, in contrast, treats pods as interchangeable and doesn't guarantee ordering or stable identity, making it suitable for stateless applications.

Explain like I am 5

A Deployment treats all its pods as identical and interchangeable, like disposable cups. A StatefulSet treats each pod as a unique, named individual with its own dedicated storage, like assigned lockers that stay tied to the same person.

How to answer in an interview

A DaemonSet ensures that a copy of a specific pod runs on every node, or a selected subset of nodes, in the cluster, commonly used for cluster-wide services like log collectors, monitoring agents, or network plugins that need to run on each node.

Explain like I am 5

A DaemonSet is like assigning exactly one guard to stand at every single entrance of a building, automatically, no matter how many entrances get added later.

How to answer in an interview

A PersistentVolume (PV) represents a piece of actual storage provisioned in the cluster, independent of any pod's lifecycle. A PersistentVolumeClaim (PVC) is a request for storage made by a user or application, which Kubernetes binds to a matching available PV, abstracting away the underlying storage details from the application.

Explain like I am 5

A PersistentVolume is like an actual storage unit available for rent. A PersistentVolumeClaim is like a request form saying 'I need a storage unit of this size,' and Kubernetes matches it up with an available one automatically.

How to answer in an interview

etcd is a distributed, consistent key-value store that serves as Kubernetes' primary data store, holding the entire cluster's configuration and state, including information about nodes, pods, and other resources. All control plane components read from and write to etcd, making its availability and consistency critical to the cluster's reliability.

Explain like I am 5

etcd is like the master notebook where Kubernetes writes down everything about the cluster's current setup, and everyone in the control plane checks that same notebook to know what's really going on.