| DBaaS | DIY | |
|---|---|---|
| Time to solution (initial design, setup, expertise) | ✔ | ❌ |
| Dependence on specialists (retention, HR, availability) | ✔ | ❌ |
| Costs to init (design, setup) | ✔ | ❌ |
| Operational complexity (monitoring) | ✔ | ❌ |
| Operational costs (backup, upgrade) | ✔ | ❌ |
| Costs of performance (more CPU cycles for same budget) | ❌ | ✔ |
| Costs of data volume (scaling with volume) | ❌ | ✔ |
| Control & Special features (DB or environment specific) | ❌ | ✔ |
| Security in general (maintenance, patch level, monitoring) | ✔ | ❌ |
| Specific security requirements (legal requirements, confidelity) | ❌ | ✔ |
In my case (at this point in the course) using Cloud SQL would require
- additional effort of implementation
- additional costs of running the cluster (though still well within the initial credits)
- increase complexity of the CD pipeline
- sacrifice the strict isolation between namespaces created by CI
- or lose the flexibility of spinning them up
"When you delete an instance, you can't reuse the name of the deleted instance until one week from the deletion date."
So I've concluded that I stick with the current solution of StatefulSet.