Cloud SQL Read Replica
Does ZopNight manage Cloud SQL Read Replica?
A Cloud SQL read replica bills at full instance rates, using the same vCPU, memory, and storage meters as its primary, and does not stop when the primary stops. ZopNight links every replica to its primary in topology and flags under-utilized replicas, since one forgotten replica can double a database's spend.
Rules that fire on Cloud SQL Read Replica
No active rule family targets Cloud SQL Read Replica today. Rules that used to are retired, and retired rules publish no pages and fire no findings. Scheduling and permissions coverage are unaffected.
A Cloud SQL read replica is a continuously synchronized copy of a primary instance used to offload read traffic, billed at full instance rates. Replicas provisioned for load that never materialized double database spend for no benefit.
A copy that costs as much as the original
Nothing about replica pricing is discounted: the replica meters its own vCPUs, memory, and provisioned storage exactly as a standalone instance would, hour after hour, while continuously applying the primary’s changes. The read traffic it was created to absorb never enters the equation. A primary with one same-sized replica is, on the invoice, two databases, and replicas can themselves be sized differently from their primary, so the copy is sometimes the more expensive of the pair.
Why stopping the primary is not enough
Cloud SQL scheduling operates on the primary’s activation policy, and a replica does not follow its primary down. Stop the primary overnight and the replica keeps running, keeps replicating nothing, and keeps billing. That is why scheduling is handled via the primary while replicas need their own decision: keep, resize, or remove. Teams that schedule primaries without auditing replicas routinely recover half of what they expected.
Topology linkage in ZopNight’s inventory
ZopDev discovers replicas via Cloud Asset Inventory, links them to their primaries in topology, and flags under-utilized replicas through its recommendation rules. The linkage matters because a replica’s name rarely announces what it serves; connecting each one to its primary and its read load is what separates a load-bearing replica from an artifact of an old scaling plan.
Replica decisions that age badly
Replicas provisioned ahead of projected read growth that never arrived. Reporting replicas kept after the reporting workload moved to BigQuery. Cross-region replicas created for a disaster-recovery drill and retained without a failover plan that would ever use them. Each pattern shares a trait: the original justification expired, the meter did not.
Spotting replicas in the Cloud SQL console
Google Cloud console → SQL lists replicas alongside primaries. The Instances table shows each replica indented under (or labeled with) its primary, with its own machine type and storage figures to compare against actual read traffic.