Prepare and test your skills

Prepare and test your skills

Free sample question with answers and detailed explanation below. The Professional Cloud Database Engineer practice bank has an additional 769 practice questions.Try 10 free realistic questions: instant score, no credit card, and no account required.
You manage a production Cloud SQL for MySQL instance configured with 32 GB of RAM and 8 vCPUs. Users report intermittent query latency spikes, and you inspect the instance's telemetry in Cloud Monitoring over the past seven days. You observe the following patterns:
usage metric: Consistently at 93% and trending toward 95%.total_usage metric: Consistently at 97%, leaving minimal difference between total_usage and usage.How should you diagnose this bottleneck and resize the instance?
A Cloud SQL read replica is an asynchronous copy of the primary instance that handles read-only SELECT traffic. Deploying replicas is a standard architectural pattern for horizontally scaling read-heavy workloads and reducing query load on the primary database.
usage is already at 93%.The primary instance showed healthy memory headroom (for example, usage below 75%) but suffered from high CPU utilization (over 90%) driven predominantly by concurrent read queries that could be safely routed to a replica.
total_usage includes the operating system cache, which Cloud SQL automatically repurposes to prevent out-of-memory errors.In Cloud SQL, the total_usage metric captures both the memory actively held by the database engine container and the memory temporarily allocated to the operating system (OS) page cache. Because OS cache memory can be reclaimed by the system when needed, a high total_usage value alone does not always indicate an emergency.
total_usage and usage to see how much memory is actually reclaimable. Here, usage is 93% and total_usage is 97%, meaning only 4% of total memory is OS cache.usage approaching the 95% threshold recommended for instances larger than 16 GB, the instance is dangerously close to an out-of-memory (OOM) failure.usage is actively reserved by database processes and buffers.usage exceeds 90–95% leaves no buffer for traffic bursts.total_usage were at 97% while usage remained low (for example, 55%), demonstrating that over 40% of RAM was harmlessly occupied by repurposable OS cache rather than database processes.
In Cloud SQL, persistent disk IOPS and throughput limits scale alongside disk size and the number of vCPUs provisioned on the machine type. Increasing vCPUs provides greater compute concurrency and higher storage bandwidth ceilings for CPU-bound or I/O-throttled workloads.
usage at 93% and rising, leaving the instance exposed to an out-of-memory (OOM) crash.usage causes cache misses, swapping pressure, and OOM terminations.Cloud Monitoring showed CPU utilization consistently at or above 95% alongside long-running queries or high active thread counts while memory usage remained comfortably below 80%.
Evaluating Cloud SQL memory health requires interpreting the usage and total_usage metrics together in Cloud Monitoring. The usage metric measures memory consumed directly by database processes and containers, while total_usage reflects both database process memory and operating system (OS) cache.
usage should remain below 95% to prevent out-of-memory (OOM) errors. At 93% and rising, the database processes have almost exhausted available RAM.total_usage (97%) and usage (93%) indicates very little OS cache remains available to repurpose.Scaling memory directly resolves both the imminent OOM risk and the cache-miss latency spikes. Unlike adding vCPUs or deploying replicas on equally undersized hardware, increasing instance memory addresses the root cause—an undersized memory tier for the workload's working set.
Answer them yourself and get your score the moment you finish. No account needed.
Full access includes
Keep the momentum going with these hand-picked practice scenarios
Last updated Validated