How to handle versions for millions of objects

We are setting up Pimcore 6.x for a customer. At go-live we’re expecting somewhere between 25-30 million objects spread over some 20 classes. We’re experiencing problems with the version files filling up the inodes of the disk.

Does anyone have any experience working with versions and these numbers of objects? Any tips, apart from the obvious options to turn off the feature or throw a larger disk on the server, would be welcome.

How much work would it be to store the versions in a database instead? Mysql or perhaps a redis or something similar.

Need to look at the same problem, maybe S3 setup could solve it?

https://pimcore.com/docs/6.x/Development_Documentation/Installation_and_Upgrade/System_Setup_and_Hosting/Amazon_AWS_Setup/Amazon_AWS_S3_Setup.html

Hi,
we would love to see a PR for configuring DB as alternative version storage.
Architecture wise most things should be prepared. You can override the version model, as stated here: https://github.com/pimcore/pimcore/issues/3342

BR
Christian

No time to build anything fancy. At least not for this project.

We’ve solved it using a S3 bucket. We mounted the bucket on our server using https://github.com/s3fs-fuse/s3fs-fuse. This way nothing needs to be done in Pimcore. Pimcore writes files to the version folder as usual. But on a lower level the version folder is actually a S3 bucket.