Cluster configuration & partition reassignment

Prompt A. "The replica list of a partition is [0, 1]. You must move it onto the broker set {0, 2, 3} at replication factor 2. Brokers 0 and 2 are in rack A. Broker 3 is in rack B. Describe how you pick the new replica list to (a) satisfy the rf and no-duplicate rules, (b) keep data movement minimal, and (c) respect the rack spread. Which constraint wins if they conflict?"

Prompt B. "Explain why 'rebalance by regenerating every assignment' is dangerous on a busy cluster. Explain how --execute / --verify / --throttle let you move partitions safely. Link it to the resize of a Redshift cluster at peak."

Rubric — a strong explanation covers these points (5–7):

• States the three validity rules: length == rf, no duplicate brokers, rack spread when racks are configured.
• Defines minimal movement as "keep as many current replicas as the constraints allow".
• Shows that a reorder of the same set is free but a new broker copies a whole replica.
• Explains rack awareness: replicas on distinct racks so a rack/AZ outage still leaves a replica up.
• Describes --execute, then --verify until complete, and --throttle to protect production traffic.
• Names the risk of a full reshuffle (network saturation, under-replicated partitions).
• Maps it to WLM/Redshift rebalancing or a resize at peak.