Host-based persistence was very common in the early Docker days when people were running containers on predefined nodes without schedulers like Docker Swarm, Kubernetes, or Mesos. Back then, we would choose a node where we'll run a container and put it there. Upgrades were performed on the same server. In other words, we packaged applications as containers and, for the most part, treated them as any other traditional service. If a node fails... tough luck! It's a disaster with or without containers.
Since serves were prederfined, we could persist the state on the host and rely on backups when that host dies. Depending on the backup frequency, we could lose a minute, an hour, a day, or even a whole week worth of data. Life is hard.
The only positive thing about this approach is that persistence is easy. We would mount a host volume inside a container. Files are persisted...
