Learn the vocabulary of every region accepting writes simultaneously instead of one region alone.
0 / 5 completed
1 / 5
At standup, a dev mentions a deployment where every region accepts both reads and writes simultaneously, rather than one region handling writes while the others sit idle as standbys. What is this architecture called?
Multi-region active-active has every region accept both reads and writes simultaneously, rather than one region handling writes while the others sit idle as standbys waiting for a failover. Active-passive failover is the contrasting pattern where only one region is actually live for writes at any given time. This simultaneous-write architecture is what lets every region serve local traffic with low latency, at the cost of needing to handle writes landing in more than one place at once.
2 / 5
During a design review, the team acknowledges that the same record could be modified in two different regions at nearly the same moment, and wants an explicit strategy for reconciling that instead of silently picking one write to keep. Which capability supports this?
An explicit conflict-resolution strategy, such as last-write-wins backed by a vector clock, or a CRDT-based merge, gives the system a deliberate, well-understood way to reconcile two writes to the same record that landed in different regions nearly simultaneously. Letting whichever write happens to arrive first silently overwrite the other risks losing a change nobody intended to discard. This explicit strategy is what makes conflicting concurrent writes a handled case instead of an unpredictable race.
3 / 5
In a code review, a dev notices a metric tracking how far behind one region's replicated data is from another region's most recent writes, and an alert configured to fire if that gap grows too large. What does this represent?
Cross-region replication lag being monitored as an inconsistency window tracks how far behind one region's replicated data is from another region's most recent writes, since a multi-region active-active system can't guarantee every region sees every write instantaneously. A read replica describes a deployment pattern, not a lag metric itself. This monitoring is essential for understanding how large a window of potential staleness or conflict the system is actually operating with at any given time.
4 / 5
An incident report shows the same customer record was updated in two different regions within the same second, and one of those updates was silently overwritten and lost because no conflict-resolution strategy had been implemented. What practice would prevent this?
Implementing an explicit conflict-resolution strategy ensures a concurrent cross-region write to the same record is reconciled deliberately, whether by timestamp ordering, a vector clock, or a CRDT merge, instead of one write silently disappearing. Continuing to let whichever write reaches storage first silently overwrite the other is exactly what caused the lost update in this incident. This deliberate reconciliation is a mandatory design decision for any active-active system, not an optional afterthought.
5 / 5
During a PR review, a teammate asks why the team accepts the complexity of conflict resolution in a multi-region active-active design instead of using a simpler active-passive setup where only one region ever accepts writes. What is the reasoning?
Active-passive keeps every write simple by routing it to one region, but that region becomes both a single point of failure and a source of added latency for any user physically far from it. Active-active serves writes locally in every region, keeping latency low everywhere, at the cost of needing an explicit strategy to resolve the occasional conflict when the same record is written in two regions at once. The tradeoff is a direct one between operational simplicity and both availability and latency, and different applications will land on different sides of it depending on how tolerable an occasional conflict actually is.