Build fluency in the vocabulary of replacing a fleet's instances in small batches without taking the service offline.
0 / 5 completed
1 / 5
A teammate explains that a fleet of twenty application instances is updated a few at a time, taking a small batch out of the load balancer, replacing it with the new version, and adding it back before moving to the next batch, so the service never goes fully offline. What deployment strategy is being described?
A rolling deployment is exactly this: a small batch of instances is taken out of the load balancer, updated to the new version, and returned to service before the next batch is updated, cycling through the whole fleet without ever taking the service fully offline. A DNS zone transfer is an unrelated concept about replicating name server records. This batch-by-batch-replace approach is exactly why rolling deployments are the default update strategy for most stateless service fleets.
2 / 5
During a design review, the team adopts a rolling deployment for a twenty-instance API fleet, specifically so only a small batch is ever offline at once and the service keeps serving traffic throughout the entire rollout. Which capability does this provide?
A rolling deployment here provides continuous availability during an update, since only a small batch of instances is ever out of the load balancer at any given moment while the rest keep serving traffic. Taking the entire fleet offline at once to update every instance simultaneously would cause a full outage for the duration of the update. This keep-most-instances-serving behavior is exactly why rolling deployments are favored for services that cannot tolerate a maintenance window.
3 / 5
In a code review, a dev notices a deployment script stops every instance in the fleet simultaneously to install the new version, causing a full outage window, instead of cycling through small batches while the rest keep serving traffic. What does this represent?
This is a missed rolling-deployment opportunity, since cycling through small batches would keep most instances serving traffic instead of taking the whole fleet down at once. A cache eviction policy is an unrelated concept about discarded cache entries. This stop-everything-at-once pattern is exactly the kind of avoidable outage a reviewer flags once continuous availability during updates is the goal.
4 / 5
An incident report shows a scheduled deployment caused a five-minute full outage because every instance in the fleet was stopped at the same time to install the update, rather than being replaced in small batches. What practice would prevent this?
Switching to a rolling deployment that updates only a small batch of instances at a time while the rest keep serving traffic eliminates the full outage window entirely. Continuing to stop every instance in the fleet simultaneously regardless of how much downtime that causes for users is exactly what caused the outage described in this incident. This small-batch-cycling approach is the standard fix once simultaneous full-fleet restarts are confirmed to be causing outages.
5 / 5
During a PR review, a teammate asks why the team reaches for a rolling deployment instead of a blue-green deployment that swaps an entirely separate, pre-warmed environment into place instantly, given that blue-green avoids any mixed-version period. What is the reasoning?
A rolling deployment trades a brief mixed-version period across batches for not needing to provision and maintain a whole second full-scale environment, while blue-green avoids mixed versions but costs double the infrastructure to keep a pre-warmed standby ready. This is exactly why rolling deployments are favored when infrastructure cost matters more than avoiding a brief mixed-version window, while blue-green remains preferable when an instant, fully tested cutover matters more than cost.