Blue-green deployment for models in MLOps - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding Blue-Green Deployment Basics

What is the main advantage of using a blue-green deployment strategy for machine learning models?

A It deploys models only during off-peak hours to reduce server load.

B It allows testing a new model version in production without downtime by switching traffic between two identical environments.

C It automatically retrains the model using live data without human intervention.

D It merges two different models into one to improve accuracy.

💡 Hint

Think about how to update a model without affecting users.

Blue-green deployment uses two identical environments (blue and green). One serves live traffic while the other is updated. Switching traffic avoids downtime and allows quick rollback.

💻 Command Output

intermediate

Interpreting Deployment Switch Command Output

You run a command to switch traffic from the blue environment to the green environment in your model deployment setup. What output indicates a successful switch?

MLOps

kubectl rollout status deployment/model-green
kubectl patch service model-service -p '{"spec":{"selector":{"env":"green"}}}'

A

deployment "model-green" successfully rolled out
service "model-service" patched

B

error: deployment "model-green" not found
service "model-service" patched

C

deployment "model-blue" successfully rolled out
service "model-service" patched

D

deployment "model-green" rolled out with warnings
service "model-service" patch failed

💡 Hint

Look for confirmation messages indicating success for both rollout and patch.

The command output shows the green deployment rolled out successfully and the service selector updated to route traffic to green.

❓ Configuration

advanced

Configuring Service Selector for Blue-Green Deployment

Which Kubernetes service selector configuration correctly routes traffic to the green environment in a blue-green deployment?

MLOps

apiVersion: v1
kind: Service
metadata:
  name: model-service
spec:
  selector:

A env: green

B version: blue

C env: blue

D version: green

💡 Hint

The selector must match the label of the green deployment pods.

The service selector uses the label 'env: green' to route traffic to pods labeled as green environment.

❓ Troubleshoot

advanced

Troubleshooting Traffic Not Switching in Blue-Green Deployment

After switching the service selector to the green environment, users still receive predictions from the blue model. What is the most likely cause?

A The model code in green pods is identical to blue pods.

B The green deployment pods crashed and are not running.

C The service selector was not updated correctly and still points to blue pods.

D The load balancer caches old responses and ignores the service selector.

💡 Hint

Check if the service selector matches the intended environment labels.

If the service selector still points to blue pods, traffic will not switch even if green pods are ready.

🔀 Workflow

expert

Blue-Green Deployment Workflow for Model Update

Arrange the steps in the correct order for performing a blue-green deployment of a new machine learning model version.

A 3,1,2,4

B 2,1,3,4

C 1,3,2,4

D 1,2,3,4

💡 Hint

Think about deploying, testing, switching traffic, then monitoring.

The correct workflow is to deploy the new model, test it, switch traffic, then monitor to ensure stability.