CompTIA Cloud+ (CV0-004)TroubleshootingHard
A cloud architect is troubleshooting a highly available database cluster deployed across multiple availability zones. Users are reporting occasional read timeouts, especially during peak load. Monitoring shows that the primary database instance's CPU utilization is consistently high (90%+), while replica instances are underutilized. The application is configured to send all write operations to the primary and distribute read operations across replicas. What is the MOST likely performance bottleneck?
- AInsufficient I/O operations per second (IOPS) on the primary database storage.
- BNetwork latency between the application and the database cluster.
- CThe read replica instances are not properly configured to handle read traffic.
- DThe application's load balancing for read operations is not effectively distributing queries.
Show answer & explanationAnswer & explanation
Correct answer: D. The application's load balancing for read operations is not effectively distributing queries.
If the primary is overloaded with CPU while replicas are underutilized, and the application is supposed to distribute reads, it indicates that the application's read distribution mechanism is failing to direct queries to the available replicas, effectively bottlenecking reads at the primary.
Why the other options are wrong
- A. High CPU on the primary suggests processing bottleneck, not necessarily storage IOPS, especially if replicas are idle.
- B. Network latency would affect all operations, but the symptom points to primary CPU overload with underutilized replicas.
- C. If replicas weren't configured, they wouldn't be 'underutilized'; they'd likely be failing or not connected.
Database Read/Write Splitting
A common database scaling strategy where write operations are directed to a primary instance, and read operations are distributed among one or more replica instances.
- Improves read scalability and performance.
- Reduces load on the primary database.
- Requires careful application configuration to route queries correctly.
Memory trick: Distribute Reads, Don't Let One Database Do Everything.