You run Cloud SQL for a large CRM and must avoid running out of storage, keep CPU utilization near 75% per core, and keep replication lag under 60s as load grows. Which actions meet these requirements?

The correct option directly addresses each requirement. Enabling automatic storage increase prevents running out of storage. Creating a Stackdriver alert for CPU exceeding 75% and then changing the instance type (scaling up or down as needed) allows for maintaining CPU utilization near 75% per core. Sharding the database is an effective strategy to reduce replication lag in a growing system by distributing the write load across multiple instances. Incorrect options: Changing to a fixed 32-core machine type without monitoring doesn't guarantee 75% CPU utilization; it might be over-provisioned or under-provisioned. Deploying Memcache reduces read load but doesn't directly address replication lag, which is often caused by write-heavy operations. Manually increasing storage after an alert is reactive and risks downtime if not done promptly. Deploying Memcached helps CPU but doesn't address replication lag. Changing instance type to reduce replication lag might help if the master is bottlenecked, but sharding is a more scalable solution for high write loads causing lag. Similar to the previous, manual storage increase is less ideal than automatic. Memcached helps CPU. Changing instance type might help replication lag, but sharding is a more robust solution for sustained growth.