GCP Professional Data Engineer Practice Question

A retail analytics platform stores user-specific session state in Memorystore for Redis Basic Tier (non-clustered). After a zonal outage caused a 15-minute cache outage and lost in-flight writes, the team must redesign for the following objectives:

Survive a single-zone failure within the region with automatic failover completed in under one minute.
Keep potential data loss to only the most recent few seconds of writes.
Avoid managing sharding logic in the application layer.

Which redesign best meets these requirements while minimizing operational overhead?

Migrate the existing instance to Memorystore for Redis Standard Tier in the same region; no code changes are required because it remains a non-clustered endpoint with automatic cross-zone replication and failover.
Recreate the cache as Memorystore for Redis Cluster, distributing shards across zones; update clients to use Redis Cluster-aware libraries to handle key hashing and redirection.
Switch to Memorystore for Redis Basic Tier in two different zones and add custom health-check logic in the application to route traffic to the healthy instance when a zone fails.
Continue using Basic Tier but enable Redis persistence (AOF and RDB) and store backup files in Cloud Storage, then script an automated restore procedure on outage.

Report Issue

Answer Description

Memorystore for Redis Standard Tier keeps a hot replica of every primary node in a different zone of the same region and performs asynchronous replication. If the primary's zone becomes unavailable, Memorystore automatically promotes the replica, typically completing failover in tens of seconds. Because replication is asynchronous, the worst-case data loss is limited to a few seconds of un-replicated writes. Basic Tier lacks any replica or automatic failover, so it cannot meet the RTO or RPO objectives. Memorystore for Redis Cluster would provide the same high availability but would require the application to implement Redis Cluster client logic and key hashing, which the team wants to avoid. Deploying two Basic Tier instances and handling failover in the application would add operational burden and manual intervention.

Ask Bash

Bash is our AI bot, trained to help you pass your exam. AI Generated Content may display inaccurate information, always double-check anything important.