Anyone use OverlayFS for their filesystem for a server? I'm curious if it's a viable solution beyond what appears to be "merely a hack to save write cycles on flaky SD cards or flash drives. In particular, curious for a solution that works for standard server distros (not just Raspberry Pi OS's raspi-config).
Would a package manager hook to commit changes on update to the system (i.e. write to lower layer) be all one needs to use the server without much attention or care that OverlayFS is in effect or are there other aspects to consider? I assume a systemd timer can be set to automatically commit the entire rootfs every hour or so and then the user elect to commit important changes manually themselves with a command.
Curious if it gets more involved than this.
For my purposes, I intend to use only 1 attached storage for NAS but create a partition for the system (the drive is connected via USB-SATA so it should be more reliable hence I prefer the system on it over an SD card or flash drive) . I don't have interact with the server much and it's running some basic services for my clients that aren't performance critical. I'm thinking OverlayFS can alleviate system writes and having the system's state be updated periodically as opposed to real time is tolerable.
Not necessarily related but I'm also thinking about using Btrfs for atomic snapshots. I think it would be a decent skill to learn something useful (applicable to "real" workstations/servers) by using the Pi Linux server as a target. My hesitation with OverlayFS is that it might be too niche of a solution as I don't see much examples that fully integrate it with typical system usage (i.e. package manager hooks, systemd services, and scripts) to make using OverlayFS a transparent experience.
Would a package manager hook to commit changes on update to the system (i.e. write to lower layer) be all one needs to use the server without much attention or care that OverlayFS is in effect or are there other aspects to consider? I assume a systemd timer can be set to automatically commit the entire rootfs every hour or so and then the user elect to commit important changes manually themselves with a command.
Curious if it gets more involved than this.
For my purposes, I intend to use only 1 attached storage for NAS but create a partition for the system (the drive is connected via USB-SATA so it should be more reliable hence I prefer the system on it over an SD card or flash drive) . I don't have interact with the server much and it's running some basic services for my clients that aren't performance critical. I'm thinking OverlayFS can alleviate system writes and having the system's state be updated periodically as opposed to real time is tolerable.
Not necessarily related but I'm also thinking about using Btrfs for atomic snapshots. I think it would be a decent skill to learn something useful (applicable to "real" workstations/servers) by using the Pi Linux server as a target. My hesitation with OverlayFS is that it might be too niche of a solution as I don't see much examples that fully integrate it with typical system usage (i.e. package manager hooks, systemd services, and scripts) to make using OverlayFS a transparent experience.
Statistics: Posted by zff — Sat Jun 28, 2025 9:02 am