I have some operations questions about Asami. I have a prod app using it, with storage on a shared disk. I assume it is not safe/possible to bring up a new version of the app accessing the same dB files as the old one, right? So I need to accept some downtime on each deployment. Q2: the cloud takes daily snapshot of the disk. If I restore it, will the DB be fine? Or is it possible that the snapshot would be taken at an unfortunate moment when the files are in some inconsistent state and thus unusable after a restore? 🙏🙏🙏

A1: I think it is. It only allows a single writer at a time, and reading is immutable. The main question would be if I remembered to handle the blank node counter correctly. If not, it’s still safe so long as only one process writes. Q2: this situation was explicitly designed for. It’s possible to snapshot during a write. In that case the final transaction will be corrupt. But a subsequent process opening it should see that and truncate that transaction. All indexes are fully forced to disk before the transaction record is written. (For this reason breaking up large transaction into multiple small transactions actually slows Asami down)

So you should be OK on both of these… you’re supposed to be anyway