Learn the vocabulary of data types that let replicas converge deterministically without coordination, even across partitions.
0 / 5 completed
1 / 5
A teammate explains that a distributed data type is designed so that any two replicas, having received the same set of updates in any order, always converge to the identical final state, with no coordination or conflict-resolution round trip required between replicas. What class of data structure is being described?
A conflict-free replicated data type, or CRDT, is exactly this: it is mathematically designed, often using commutative, associative, and idempotent merge operations, so that any two replicas that have received the same set of updates, in any order, converge to the identical final state with no coordination round trip and no manual conflict resolution needed. A DNS zone transfer is an unrelated concept about replicating name server records. This converge-without-coordination guarantee is exactly why CRDTs are favored for offline-first and multi-region collaborative applications.
2 / 5
During a design review, the team adopts a CRDT-based counter for a multi-region 'like' count that must keep accepting increments in every region even during a network partition between regions. Which capability does this provide?
A CRDT-based counter here provides partition-tolerant, coordination-free convergence, since each region's increments merge deterministically into the same final total once connectivity is restored, with no cross-region round trip required to accept an increment. Requiring a distributed lock across all regions before any increment is accepted would make the counter unavailable in any region cut off during a partition. This accept-locally-merge-later behavior is exactly why CRDTs are favored for multi-region counters that must stay available during partitions.
3 / 5
In a code review, a dev notices a multi-region 'like' counter requires a cross-region distributed lock before incrementing, making the counter unavailable in any region that loses connectivity, instead of using a CRDT counter that merges increments deterministically after the fact. What does this represent?
This is a missed CRDT opportunity, since a CRDT counter would let each region accept increments locally and merge them deterministically instead of requiring an unavailable cross-region lock. A cache eviction policy is an unrelated concept about discarded cache entries. This cross-region-lock-required pattern is exactly the kind of availability gap a reviewer flags once partition tolerance is required.
4 / 5
An incident report shows the 'like' counter stopped accepting increments in an entire region during a network partition, because incrementing required first acquiring a distributed lock held across all regions, and that region could not reach the lock service. What practice would prevent this?
Replacing the locked counter with a CRDT counter lets each region accept increments locally and merge them deterministically once connectivity is restored, so no region is blocked during a partition. Continuing to require a cross-region distributed lock before every increment regardless of how many regions become unreachable during a partition is exactly what caused the outage described in this incident. This accept-locally-merge-later approach is the standard fix once cross-region locking is confirmed to break availability during partitions.
5 / 5
During a PR review, a teammate asks why the team reaches for a CRDT instead of a strongly consistent, lock-coordinated counter, given that a coordinated counter never needs a merge step at all. What is the reasoning?
A CRDT trades a small amount of eventual-consistency lag and merge-function design effort for availability during partitions, while a lock-coordinated counter guarantees immediate global consistency but becomes unavailable in any region that cannot reach the coordination service. This is exactly why CRDTs are favored for multi-region, partition-tolerant counters and collaborative data, while lock-coordinated counters remain preferable when immediate global consistency matters more than availability during a partition.