Pod-to-Pod communication in Kubernetes - Mini Project: Build & Apply

Pod-to-Pod communication

📖 Scenario: You are working with Kubernetes to deploy two simple pods that need to talk to each other inside the same cluster. This is like two friends in the same room passing notes directly without using the internet.

🎯 Goal: You will create two pods with specific names, configure a service to allow them to communicate, and verify that one pod can reach the other by using the service name.

📋 What You'll Learn

Create two pods named pod-a and pod-b running the busybox image

Create a service named pod-b-service that targets pod-b

Use kubectl exec to ping pod-b-service from pod-a to test communication

💡 Why This Matters

🌍 Real World

Pod-to-pod communication is essential for microservices running inside Kubernetes clusters to work together and share data.

💼 Career

Understanding how to configure pods and services for communication is a fundamental skill for Kubernetes administrators and DevOps engineers.

1

Create two pods named pod-a and pod-b

Create two pods named pod-a and pod-b using the busybox image. Each pod should run a simple command to sleep for 3600 seconds to keep them alive.

Kubernetes

# Create pod-a and pod-b with busybox image and sleep command
# Your code here

Hint

Use two separate pod definitions separated by ---. Each pod uses the busybox image and runs sleep 3600.

2

Create a service named pod-b-service targeting pod-b

Create a ClusterIP service named pod-b-service that selects the pod with the label app: pod-b. Add the label app: pod-b to the pod-b pod metadata.

Kubernetes

apiVersion: v1
kind: Pod
metadata:
  name: pod-a
spec:
  containers:
  - name: busybox
    image: busybox
    command: ["sleep", "3600"]
---
apiVersion: v1
kind: Pod
metadata:
  name: pod-b
  # Add label app: pod-b here
spec:
  containers:
  - name: busybox
    image: busybox
    command: ["sleep", "3600"]

# Create a ClusterIP service named pod-b-service selecting app: pod-b
# Your code here

Hint

Remember to add the label app: pod-b to the pod-b metadata and create a service with selector matching that label.

3

Test pod-a can ping pod-b-service

Use kubectl exec to run the command ping -c 3 pod-b-service inside the pod-a pod to test connectivity.

Kubernetes

kubectl exec pod-a -- ping -c 3 pod-b-service
# Your code here

Hint

Use kubectl exec pod-a -- ping -c 3 pod-b-service to check if pod-a can reach pod-b via the service.

4

Display the ping command output

Run the command kubectl exec pod-a -- ping -c 3 pod-b-service and display the output to confirm pod-to-pod communication.

Kubernetes

# Run the ping command and show output
# Your code here

Hint

The ping output should show successful packets sent and received with 0% packet loss.

Practice

1.

How do pods inside a Kubernetes cluster typically communicate with each other?

easy

A. Using external DNS names only

B. Using pod IP addresses or service names

C. Through direct access to the host machine's network

D. By sharing the same storage volume

Solution

1. Step 1: Understand pod networking basics

   Pods have their own IP addresses inside the cluster and can communicate directly using these IPs.

2. Step 2: Recognize the role of services

   Services provide stable DNS names to reach pods even if pod IPs change.

3. Final Answer:

   Using pod IP addresses or service names -> Option B

4. Quick Check:

   Pod communication = IP or service name [OK]

Hint: Pods use IPs or services to talk inside cluster [OK]

Common Mistakes:

• Thinking pods use external DNS only
• Assuming pods share host network by default
• Believing storage volumes enable communication

2.

Which of the following is the correct way to define a Kubernetes Service to expose pods for pod-to-pod communication?

apiVersion: v1
kind: Service
metadata:
  name: my-service
spec:
  selector:
    app: myapp
  ports:
  - protocol: TCP
    port: 80
    targetPort: 8080

easy

A. The service exposes port 8080 externally but does not select pods

B. The selector matches pods with label app: myapp but ports are reversed

C. The selector matches pods with label app: myapp and forwards port 80 to 8080

D. The service definition is missing the type field, so it won't work

Solution

1. Step 1: Check the selector field

   The selector app: myapp correctly matches pods with that label.

2. Step 2: Verify ports mapping

   Port 80 on the service forwards traffic to port 8080 on the pods, which is correct.

3. Final Answer:

   The selector matches pods with label app: myapp and forwards port 80 to 8080 -> Option C

4. Quick Check:

   Selector and ports correct = The selector matches pods with label app: myapp and forwards port 80 to 8080 [OK]

Hint: Selector matches pod labels; ports map service to pod [OK]

Common Mistakes:

• Confusing port and targetPort roles
• Thinking 'type' is mandatory for basic service
• Ignoring selector means no pods matched

3.

Given two pods in the same namespace, pod-a and pod-b, where pod-b runs a web server on port 80, what is the expected output of this command run inside pod-a?

curl http://pod-b:80

Assuming pod-b is reachable by its pod IP but not by name.

medium

A. Connection refused or host not found error

B. Successful HTTP response from pod-b's web server

C. Timeout error because pods cannot communicate

D. DNS resolution error for pod-b

Solution

1. Step 1: Understand pod DNS naming

   Pods do not have DNS names by default; only services have stable DNS names.

2. Step 2: Analyze the command

   Trying to curl pod-b by name will fail DNS resolution because pod names are not DNS entries.

3. Final Answer:

   DNS resolution error for pod-b -> Option D

4. Quick Check:

   Pod names not DNS = DNS resolution error [OK]

Hint: Pod names are not DNS names; use service names [OK]

Common Mistakes:

• Assuming pod names resolve in DNS
• Expecting direct pod name communication
• Ignoring need for service abstraction

4.

You have a service named backend exposing pods on port 8080. A pod tries to connect using curl http://backend:80 but gets a connection refused error. What is the likely cause?

medium

A. The service port is 8080, but curl uses port 80

B. The service selector does not match any pods

C. The pod network is not enabled

D. The service type is NodePort instead of ClusterIP

Solution

1. Step 1: Check service port configuration

   The service exposes port 8080, but the curl command targets port 80, causing connection refusal.

2. Step 2: Understand port mapping importance

   Clients must use the service port, not the pod's internal port, to connect successfully.

3. Final Answer:

   The service port is 8080, but curl uses port 80 -> Option A

4. Quick Check:

   Port mismatch causes connection refused [OK]

Hint: Match curl port to service port exactly [OK]

Common Mistakes:

• Ignoring port mismatch between service and client
• Assuming network is disabled without checking
• Confusing service types with connectivity issues

5.

You want to ensure stable pod-to-pod communication in a Kubernetes cluster where pods restart frequently and get new IPs. Which approach best solves this problem?

hard

A. Use a Kubernetes Service to provide a stable DNS name and load balancing

B. Use pod IP addresses directly in application configuration

C. Configure pods to share the same static IP address

D. Use host networking mode for all pods

Solution

1. Step 1: Identify the problem with pod IPs

   Pods get new IPs on restart, so direct IP communication is unstable.

2. Step 2: Use Kubernetes Service for stability

   Services provide a stable DNS name and load balancing to reach pods regardless of IP changes.

3. Step 3: Evaluate other options

   Static IPs for pods are not practical; host networking breaks isolation; direct IP use is unreliable.

4. Final Answer:

   Use a Kubernetes Service to provide a stable DNS name and load balancing -> Option A

5. Quick Check:

   Service = stable pod communication [OK]

Hint: Use services for stable pod communication despite IP changes [OK]

Common Mistakes:

• Hardcoding pod IPs in configs
• Trying to assign static IPs to pods
• Using host networking unnecessarily