A team is designing a fault-tolerant Google Cloud architecture for a customer-facing application that uses Cloud SQL for its system of record and Memorystore for Redis as a write-through cache. They want to know which failure scope each managed service can absorb without an operator-led recovery, so they can document the resilience properties of the overall stack honestly. Which statement most accurately describes the fault tolerance built into these services?
- ACloud SQL regional HA and Memorystore for Redis standard tier both automatically fail over within their region and also automatically fail over to a different region if the home region becomes unavailable, provided cross-region replicas are configured for either service.
- BCloud SQL regional HA transparently survives the loss of any zone in the region but Memorystore for Redis standard tier requires an operator to trigger replica promotion, because Memorystore exposes no automatic zonal failover for cache workloads.
- CCloud SQL regional HA and Memorystore for Redis standard tier both transparently survive the loss of a single zone within their region without operator intervention, but neither service provides automatic failover to a different region if the entire region becomes unavailable. Correct
- DCloud SQL regional HA tolerates only the loss of the primary zone while keeping a defined standby zone available, whereas Memorystore for Redis standard tier tolerates the loss of any zone in the region because clients are routed through a regional load balancer to any healthy shard.
Why A is wrong: Configuring cross-region replicas is good practice for DR and so this option sounds reasonable, but neither service automatically promotes a cross-region replica during a regional outage; that step is operator-initiated, not a built-in automatic failover.
Why B is wrong: Some candidates assume managed caches require manual failover because they prioritise speed over availability, which makes this distractor plausible, but standard tier Memorystore does provide automatic replica promotion on zonal failure within the region.
Why C is correct: Regional HA in Cloud SQL synchronously mirrors data to a second zone and automatically promotes the standby on zonal failure; standard tier Memorystore for Redis has a replica in another zone with automatic promotion. Cross-region disaster recovery, however, requires explicit setup and an operator-led promotion.
Why D is wrong: Mapping HA to fixed zones and giving Redis a multi-zone routing story makes the model seem complementary, but Cloud SQL HA is not pinned to a specific surviving zone and Memorystore does not front shards with a regional load balancer that routes across zones independently of primary placement.