In this Dev Diary, we’re taking a closer look at how we reduced the costs and complexity of our development and QA environments by rethinking how they run. What started as a perfectly functional Kubernetes setup eventually became too expensive and too complicated for what we needed, so we found a different path.

Our website consists of several services:

1. An Admin CMS built with React
   
2. A backend API built with Nest.js
   
3. A Next.js website handling server-side rendering
   
4. A PostgreSQL database
   

All of these components were deployed as Docker containers inside a Kubernetes cluster. While this setup offered flexibility and strong performance, it came with a clear drawback: cost.

The problem

The total cost reached around 100 euros per environment per month. This was acceptable for production but too expensive for development and QA environments. A secondary issue was the ongoing maintenance of the Kubernetes clusters — they require regular upgrades, and tasks like debugging or exporting the database were more complicated than necessary.

The solution

We decided to migrate from Kubernetes to managed services. Based on previous positive experience, we chose Google Cloud — specifically Cloud Run and Cloud SQL with a PostgreSQL database.

Cloud Run offers request-based billing, meaning you don’t pay when a container is inactive. This fits perfectly for development and QA environments, which are idle most of the time. Each service also allows fine-grained control over instance size — CPU, RAM, and other resources — so you can tailor configurations exactly to your needs. And in the case of higher load, Cloud Run automatically scales the number of containers.

As a bonus, access to logs and the database is significantly easier compared to the Kubernetes setup.

Verdict

The best part? The total cost of a single environment, including the database, is now around 5 euros per month. Even under higher load, it rarely exceeds 10 euros. This brings substantial savings without any compromise in functionality.

Production environments will naturally have higher costs due to high availability and performance requirements, but even then, the expenses remain lower than with the original solution.