Build fluency in the vocabulary of replicas converging to the same value once writes stop arriving.
0 / 5 completed
1 / 5
At standup, a dev mentions a replicated data store where, after writes stop arriving, every replica is guaranteed to converge to the same value, but at any given moment a read might return a slightly stale copy. What is this consistency model called?
Eventual consistency is exactly this: it guarantees that if no new writes arrive, every replica of a distributed data store will eventually converge to the same value, even though a read taken before convergence may return a stale copy. A hash collision is an unrelated hash-table concept about two keys sharing a bucket. This converge-without-guaranteeing-immediacy behavior is exactly why eventual consistency lets a system keep accepting writes even while replicas are still catching up with each other.
2 / 5
During a design review, the team picks eventual consistency for a 'like count' feature, specifically because accepting writes on any replica without waiting for cross-region agreement keeps the feature available even during a network partition. Which capability does this provide?
Eventual consistency here provides high availability during partitions, since each replica can accept writes independently and reconcile later instead of blocking until every replica agrees, which is impossible while a partition is in progress. A model requiring every replica to agree before a write is accepted must refuse writes the moment replicas can't reach each other. This accept-now-reconcile-later behavior is exactly why eventual consistency is favored for features that can tolerate brief staleness in exchange for staying available.
3 / 5
In a code review, a dev notices a 'like count' feature synchronously queries every regional replica and blocks until they all agree before returning a single count, instead of tolerating an eventually consistent read that might lag by a few seconds. What does this represent?
This is a missed eventual-consistency opportunity, since tolerating a few seconds of staleness on a feature like a like count would avoid blocking on cross-region agreement for every single read. A cache eviction policy is an unrelated concept about discarded cache entries. This synchronous-cross-region-query pattern is exactly the kind of unnecessary latency a reviewer flags once the feature is confirmed tolerant of brief staleness.
4 / 5
An incident report shows a global service went unavailable in every region during a network partition, because a feature that could tolerate a few seconds of staleness still required synchronous agreement across all replicas before accepting any write. What practice would prevent this?
Adopting eventual consistency lets each replica keep accepting writes independently during the partition, avoiding the all-region outage. Continuing to require synchronous cross-replica agreement before every write regardless of how tolerant the feature is of brief staleness is exactly what caused the outage described in this incident. This accept-then-reconcile approach is the standard fix once a feature is confirmed tolerant of brief staleness.
5 / 5
During a PR review, a teammate asks why the team chooses eventual consistency instead of strong, linearizable consistency for a high-traffic feature, given that strong consistency guarantees every read reflects the latest write. What is the reasoning?
Eventual consistency keeps the feature available and low-latency by letting replicas accept writes independently and reconcile later, at the cost of temporary staleness, while strong consistency guarantees freshness on every read but must block or reject writes when replicas can't coordinate, such as during a network partition. This is exactly the CAP-theorem trade-off that leads teams to choose eventual consistency for availability-sensitive features and strong consistency for correctness-critical ones.