Desired replicas vs actual replicas in Kubernetes - When to Use Which

The Big Idea

What if your app could fix itself when it's too busy or too quiet, without you doing anything?

The Scenario

Imagine you have a website that suddenly gets a lot of visitors. You try to start more copies of your website manually on different servers to handle the traffic.

You have to check each server one by one to see if the website is running, and start or stop copies as needed.

The Problem

This manual way is slow and tiring. You might forget to start enough copies or accidentally start too many, wasting resources.

Also, if a copy crashes, you might not notice quickly, causing your website to slow down or stop.

The Solution

Kubernetes solves this by letting you say how many copies (replicas) you want, called the desired replicas.

It then watches the actual replicas running and automatically starts or stops copies to match your desired number.

This keeps your website running smoothly without you checking all the time.

Before vs After

✗ Before

ssh server1
start website
ssh server2
start website

✓ After

kubectl scale deployment mysite --replicas=3

What It Enables

You can focus on your app while Kubernetes keeps the right number of copies running automatically.

Real Life Example

During a big sale, your online store needs more copies to handle many buyers. Kubernetes adjusts the copies up and down without you lifting a finger.

Key Takeaways

Manually managing copies is slow and error-prone.

Desired replicas tell Kubernetes how many copies you want.

Kubernetes keeps actual replicas matching desired replicas automatically.

Practice

1. What does the desired replicas value represent in a Kubernetes Deployment?

easy

A. The number of pods currently running

B. The number of containers inside a pod

C. The number of nodes in the cluster

D. The number of pod copies Kubernetes should run

Solution

1. Step 1: Understand the role of desired replicas

   Desired replicas specify how many copies of a pod Kubernetes should maintain to run your app.

2. Step 2: Differentiate from actual replicas

   Actual replicas show how many pods are currently running, which may differ temporarily.

3. Final Answer:

   The number of pod copies Kubernetes should run -> Option D

4. Quick Check:

   Desired replicas = target pod count [OK]

Hint: Desired replicas = target pods Kubernetes tries to run [OK]

Common Mistakes:

• Confusing desired replicas with actual running pods
• Thinking desired replicas count containers, not pods
• Mixing desired replicas with cluster node count

2. Which of the following is the correct YAML snippet to set desired replicas to 3 in a Deployment?

easy

A. replicaCount: 3

B. desiredReplicas: 3

C. replicas: 3

D. pods: 3

Solution

1. Step 1: Identify the correct field name in Deployment spec

   The correct field to set desired replicas is replicas under the spec section.

2. Step 2: Verify syntax correctness

   Using replicas: 3 is valid YAML and recognized by Kubernetes.

3. Final Answer:

   replicas: 3 -> Option C

4. Quick Check:

   Use 'replicas' to set desired pod count [OK]

Hint: Use 'replicas' key in Deployment spec to set desired pods [OK]

Common Mistakes:

• Using incorrect field names like desiredReplicas or replicaCount
• Placing replicas outside the spec section
• Using plural 'pods' instead of 'replicas'

3. You run kubectl get deployment myapp and see:
DESIRED=4, CURRENT=3, UP-TO-DATE=3, AVAILABLE=3
What does this mean?

medium

A. One pod has crashed and won't restart

B. Kubernetes is still creating the 4th pod

C. All pods are running as expected

D. Deployment is scaled down to 3 pods

Solution

1. Step 1: Compare desired and current replicas

   Desired is 4 but current is 3, so Kubernetes aims for 4 pods but only 3 are running now.

2. Step 2: Interpret the status fields

   Up-to-date and available are 3, meaning 3 pods are ready and updated; the 4th pod is likely starting.

3. Final Answer:

   Kubernetes is still creating the 4th pod -> Option B

4. Quick Check:

   Desired > Current means pods are still launching [OK]

Hint: Desired > Current means pods are starting or recovering [OK]

Common Mistakes:

• Assuming pods crashed without checking events
• Thinking deployment scaled down when desired is higher
• Confusing available with desired replicas

4. You notice your Deployment's desired replicas is 5 but actual replicas stays at 3. Which is the most likely cause?

medium

A. Pods are stuck in Pending state due to resource limits

B. Deployment YAML has syntax errors

C. kubectl command was typed incorrectly

D. Actual replicas always match desired replicas instantly

Solution

1. Step 1: Understand why actual replicas lag behind desired

   If actual replicas are less than desired, pods may be stuck starting, often due to resource constraints.

2. Step 2: Eliminate other options

   Syntax errors prevent deployment creation, incorrect kubectl commands won't show this state, and actual replicas do not update instantly.

3. Final Answer:

   Pods are stuck in Pending state due to resource limits -> Option A

4. Quick Check:

   Pending pods cause actual < desired replicas [OK]

Hint: Check pod status if actual replicas < desired replicas [OK]

Common Mistakes:

• Assuming syntax errors cause replica mismatch
• Expecting instant replica count update
• Ignoring pod scheduling issues

5. You want to ensure your Deployment always has 3 pods running. You set replicas: 3 but sometimes kubectl get pods shows only 2 running pods. What should you do to fix this?

hard

A. Check pod events and logs to find why one pod is not running and fix the issue

B. Increase replicas to 4 to force Kubernetes to create more pods

C. Delete the Deployment and recreate it with replicas set to 3

D. Ignore it; actual replicas will eventually match desired without intervention

Solution

1. Step 1: Investigate why pods are missing

   Check pod events and logs to identify issues like crashes, scheduling failures, or resource limits causing pods not to run.

2. Step 2: Fix the root cause

   Resolve the identified problem (e.g., fix image, increase resources) so Kubernetes can maintain 3 running pods as desired.

3. Final Answer:

   Check pod events and logs to find why one pod is not running and fix the issue -> Option A

4. Quick Check:

   Fix pod issues to match actual to desired replicas [OK]

Hint: Always check pod status and logs if replicas don't match [OK]

Common Mistakes:

• Scaling replicas blindly without fixing pod issues
• Deleting Deployment unnecessarily
• Assuming Kubernetes fixes pod issues automatically