I have a project that runs on Kubernentes and uses a microservices architecture.

The application will need to create and serve PDF files in two of its APIs.

I wonder what is the best practice to store such data in an on-premises Kubernetes cluster. S3 would have been a perfect fit but that is not possible since i'm not running on AWS.

I guess K8S persistent volume is an option, but I worry about handling data corruptuon and how to handlescaling. The number of PDFs can reach the hundreds of millions.

Would you consider a NoSQL solution as a viable alternative? Any specific implementation that would provide benefits compared to others?

Thanks.