Kubernetes basics review in Microservices - Scalability & System Analysis

Scalability Analysis - Kubernetes basics review

Growth Table: Kubernetes Basics

Users/Workloads	What Changes
100 users	Single Kubernetes cluster with a few nodes; simple deployments; manual scaling
10,000 users	More nodes added; use of Horizontal Pod Autoscaler; introduction of namespaces for isolation
1,000,000 users	Multiple clusters; cluster federation or multi-cluster management; advanced networking; use of ingress controllers and service meshes
100,000,000 users	Global multi-region clusters; automated cluster provisioning; heavy use of monitoring, logging, and security policies; advanced autoscaling and resource optimization

First Bottleneck

At small scale, the first bottleneck is the control plane of Kubernetes. It manages the cluster state and schedules pods. With increasing workloads, the API server and scheduler can become overwhelmed.

Also, the etcd database that stores cluster state can become a bottleneck if too many updates happen rapidly.

Scaling Solutions

• Control Plane Scaling: Use managed Kubernetes services or run highly available control plane nodes to distribute load.
• Horizontal Pod Autoscaling: Automatically scale pods based on CPU or custom metrics.
• Cluster Federation: Manage multiple clusters to distribute workloads geographically.
• Namespace and Resource Quotas: Isolate workloads and prevent resource contention.
• Use of Ingress Controllers and Service Meshes: Efficient traffic routing and observability.
• Monitoring and Logging: Use tools like Prometheus and Fluentd to track cluster health and performance.

Back-of-Envelope Cost Analysis

Assuming 10,000 concurrent users generating 100 requests per second (RPS):

• API Server handles ~1000-5000 concurrent connections; may need multiple replicas.
• Each node can run hundreds of pods; adding nodes increases capacity linearly.
• Network bandwidth depends on pod communication; 1 Gbps network can handle ~125 MB/s.
• Storage for logs and metrics grows with number of pods; consider retention policies.

Interview Tip

When discussing Kubernetes scalability, start by explaining the cluster components and their roles.

Identify the control plane as a potential bottleneck early on.

Discuss horizontal scaling of nodes and pods, and how autoscaling helps.

Mention multi-cluster strategies for very large scale.

Always relate solutions to specific bottlenecks you identify.

Self Check

Your Kubernetes API server handles 1000 QPS. Traffic grows 10x to 10,000 QPS. What do you do first?

Answer: Scale the control plane by adding more API server replicas or move to a managed Kubernetes service with a highly available control plane to handle increased load.

Key Result

Kubernetes scales by adding nodes and pods horizontally, but the control plane (API server and etcd) is the first bottleneck; scaling it and using multi-cluster setups are key for large workloads.

Practice

1. What is a pod in Kubernetes?

easy

A. A command-line tool to manage Kubernetes

B. The smallest unit that runs one or more containers together

C. A configuration file format used in Kubernetes

D. A network policy to control traffic

Solution

1. Step 1: Understand Kubernetes resource types

   Kubernetes groups containers into pods to run them together on the same host.

2. Step 2: Identify the role of a pod

   A pod is the smallest deployable unit that can contain one or more containers sharing resources.

3. Final Answer:

   The smallest unit that runs one or more containers together -> Option B

4. Quick Check:

   Pod = smallest container group [OK]

Hint: Pods group containers; smallest deployable unit [OK]

Common Mistakes:

• Confusing pods with kubectl tool
• Thinking pods are config files
• Mixing pods with network policies

2. Which command correctly lists all pods in the default namespace?

easy

A. kubectl get pods

B. kubectl list pods

C. kubectl show pods

D. kubectl describe pods all

Solution

1. Step 1: Recall kubectl commands for listing resources

   The standard command to list pods is kubectl get pods.

2. Step 2: Check other options for correctness

   kubectl list pods and kubectl show pods are invalid commands; kubectl describe pods all is incorrect syntax.

3. Final Answer:

   kubectl get pods -> Option A

4. Quick Check:

   List pods = kubectl get pods [OK]

Hint: Use 'kubectl get pods' to list pods [OK]

Common Mistakes:

• Using 'list' or 'show' instead of 'get'
• Adding unnecessary arguments like 'all'
• Confusing describe with get

3. Given this YAML snippet for a pod:

apiVersion: v1
kind: Pod
metadata:
  name: myapp-pod
spec:
  containers:
  - name: myapp-container
    image: nginx:latest
    ports:
    - containerPort: 80

What will kubectl get pods myapp-pod show after creation?

medium

A. NAME READY STATUS RESTARTS AGE
myapp-pod 1/1 Completed 0 0s

B. Error: pod not found

C. NAME READY STATUS RESTARTS AGE
myapp-pod 0/1 Pending 0 0s

D. NAME READY STATUS RESTARTS AGE
myapp-pod 1/1 Running 0 0s

Solution

1. Step 1: Understand pod creation from YAML

   The YAML defines a pod with one container running nginx, exposing port 80.

2. Step 2: Predict pod status after creation

   Immediately after creation, the pod should be running with 1 container ready, so status is Running and READY is 1/1.

3. Final Answer:

   NAME READY STATUS RESTARTS AGE myapp-pod 1/1 Running 0 0s -> Option D

4. Quick Check:

   Pod created and running = READY 1/1 Running [OK]

Hint: New pod with valid image shows READY 1/1 Running [OK]

Common Mistakes:

• Expecting Pending status without reason
• Confusing Completed with Running
• Assuming pod not found immediately after creation

4. You run kubectl apply -f pod.yaml but get an error: "error: unable to recognize \"pod.yaml\": no matches for kind \"Pod\" in version \"v2\"". What is the likely fix?

medium

A. Delete the pod.yaml and recreate it

B. Rename the file to pod.yml

C. Change apiVersion from v2 to v1 in pod.yaml

D. Run the command with sudo

Solution

1. Step 1: Analyze the error message

   The error says no matches for kind "Pod" in version "v2", meaning the apiVersion is invalid.

2. Step 2: Correct the apiVersion in YAML

   The correct apiVersion for Pod is "v1", so changing from "v2" to "v1" fixes the issue.

3. Final Answer:

   Change apiVersion from v2 to v1 in pod.yaml -> Option C

4. Quick Check:

   apiVersion must be valid (v1 for Pod) [OK]

Hint: Check apiVersion spelling and value in YAML [OK]

Common Mistakes:

• Changing file extension instead of apiVersion
• Running with sudo unnecessarily
• Deleting file without fixing content

5. You want to update a running pod's container image from nginx:1.19 to nginx:1.21 without downtime. Which Kubernetes resource and method should you use?

hard

A. Create a Deployment and update its image with kubectl set image

B. Directly edit the pod with kubectl edit pod to change the image

C. Delete the pod and create a new one with the new image

D. Use kubectl scale pod to increase replicas

Solution

1. Step 1: Understand pod immutability and updates

   Pods are immutable; you cannot update container images directly on running pods without recreating them.

2. Step 2: Use Deployment for zero downtime updates

   Deployments manage pods and allow rolling updates to change images without downtime using kubectl set image.

3. Final Answer:

   Create a Deployment and update its image with kubectl set image -> Option A

4. Quick Check:

   Use Deployment + set image for smooth updates [OK]

Hint: Use Deployment and set image for zero downtime update [OK]

Common Mistakes:

• Trying to edit pod image directly
• Deleting pod causes downtime
• Scaling pod is invalid command