Are snapshots point-in-time of the entire disk?

I’m using Longhorn encrypted volumes with Rancher to run a Postgres database. We are questioning weather we need to pause disk writes in the database before creating a snapshot of the disk. Postgres is good at recovering (if the process is killed for example), but only if the whole disk is a snapshot of a single point in time.

Does the Longhorn Snapshot feature take an image of the whole disk at one point in time? Similar to Microsoft’s volume shadow copy?

When you perform a snapshot, Longhorn takes an image of the entire kubernetes volume at one point in time: Longhorn | Documentation