top of page
  • Writer's pictureSadik Bakiu

Three Cost Management Pitfalls on GCP Cloud SQL

A set of clouds loosing money

When the organization is growing it is very important to keep the costs as low as possible. A common oversight in managing cloud resources is the assumption that costs are paused when instances are not in operation. In this post, we will be talking about three ways one could be incurring unnecessary costs on GCP's Cloud SQL even though you have shut down the instance.

Cloud SQL is a managed service on GCP that offers managed relational database management systems like Postgres, MySQL, or SQL Server.

This is especially relevant for cases when one might shut down their instance and have the impression that you might not be paying for the resource that you are not using.

CUDs do not apply to Cloud Storage costs

Committed Use Discounts (CUDs) from GCP reward customers with reduced rates in return for a commitment to a minimum spend on certain services. For Cloud SQL, these discounts apply to vCPU and memory usage but notably exclude allocated storage costs. Therefore, any SSD storage allocated for a Cloud SQL database remains at full price, unaffected by CUDs.

Storage Costs Persist Even for Inactive Instances

Even if a Cloud SQL instance is shut down, users are still liable for the full cost associated with the allocated SSD storage. This pricing model stems from the fact that the allocated space on SSDs is reserved exclusively for the user's data, regardless of the instance's operational status. Consequently, the storage space cannot be repurposed or allocated to other users, leading to ongoing charges. This can significantly impact budgeting and cost management.

In case the instance is seldom used, it is important to consider using other services like BigQuery.

One pays for Public IPs not in use

IPv4 addresses have become a scarce commodity due to the limitations of the original internet architecture, which did not anticipate the exponential growth in demand. In contrast, the IPv6 protocol offers a solution to this constraint by providing a vastly larger address space. However, the widespread adoption of IPv6 is still in progress, requiring time before it is fully integrated across the internet.

In case a Public IPv4 has been assigned to a Cloud SQL instance and the instance is not active, one will be charged $0.01 per hour the instance is not used. It is worth noting that assigning public IP addresses to your Cloud SQL instances is discouraged as well.


This blog post details a few cost management pitfalls for Cloud SQL, emphasizing that SSD storage costs persist even when instances are inactive, that CUDs don't apply to storage, and that there are additional costs related to the use of IPv4 addresses. It is important to routinely review billing reports to identify opportunities for cost optimization.


bottom of page