Hi Dave,
Machine A is running a model that writes files periodically to a file system NFS mounted from machine B. Machine B backs up this file system periodically to machine C via rsync. If this backup runs while machine A is writing a file, the model on A crashes.
OpenBSD FFS1 and FFS2 do not implement snapshots. If Machine B runs OpenBSD, you (or cron) should stop the model on Machine A from any disk writes on B before the backup starts, and not resume it until the backup is complete. That way, the data on B stays consistent until it is rsync'ed to C.
If it is a higher priority to keep the model running and writing to disk at all times, than it is to have OpenBSD on Machine B, perhaps use FreeBSD with ZFS and its snapshots capability on Machine B. You would still want to ensure rsync backs up from the latest snapshot and not the live system to prevent the model on Machine A from crashing for the same reason.
However, a whole second operating system is a lot to learn, automate, and maintain (especially across version upgrades) than simply stopping the model from writing to disk during the backup.
If rsync takes a long time and you don't want the model stopped for that long, consider switching from rsync to borgbackup. I have not compared the two. But I hear it's faster than rsync in many cases. With borg (or any versioned backup tool), Machine C would need plenty of free storage space, since borg keeps old versions of your data, not just the latest like rsync.
