MLA-C01 Deployment and Orchestration of ML Workflows MEDIUM
PRODUCTION SCENARIO
A team is updating the model behind a production SageMaker real-time endpoint. They want the update to run in exactly two traffic steps, a small portion first and the remainder after a baking period, returning to the previous fleet automatically if a CloudWatch alarm fires.

Which deployment guardrails configuration meets these requirements?

Answering here is anonymous. Nothing is saved unless you sign in.

Show answer and explanation

Answer: Canary traffic shifting with CloudWatch alarms and a baking period

Canary shifting moves one small portion of traffic to the green fleet, bakes it under CloudWatch alarms, then shifts the remainder, which is exactly two steps. Linear shifting also bakes and rolls back but spreads the move across ten steps rather than two. All-at-once moves every request in one step, and a rolling deployment provisions capacity in equal batches of the size you set rather than one small portion followed by the rest.
Free

Keep practicing MLA-C01

18 original MLA-C01 practice questions, each with an explanation and a source link. No account needed.

Start free practice set → Timed, explained, free