Problem / Motivation
Today VMs boot from a single disk image. This makes upgrades and “factory reset” workflows heavy (copy full images), and makes it hard to keep a golden base immutable.
Proposed behavior
- Allow a node to boot from a read-only base image plus a writable overlay (qcow2 overlay recommended).
- Overlay location (example):
/var/lib/aos-unit/overlays/<node>/overlay.qcow2
- Base image stays in
/var/lib/aos-unit/images/ (or existing image dir).
Implementation sketch
- If overlay missing:
qemu-img create -f qcow2 -b <base> -F <base_format> <overlay>
- QEMU uses overlay as the main
-drive, base referenced as backing file.
- Optional config flag per node:
overlay: true|false (default false initially).
Acceptance criteria
- Overlay can be enabled per-node.
- Base image never modified.
- Overlay lifecycle documented (create, reuse, delete/reset).
- Works for qcow2 base; define behavior for raw base (either supported with
-F raw or explicitly rejected with a clear error).
Problem / Motivation
Today VMs boot from a single disk image. This makes upgrades and “factory reset” workflows heavy (copy full images), and makes it hard to keep a golden base immutable.
Proposed behavior
/var/lib/aos-unit/overlays/<node>/overlay.qcow2/var/lib/aos-unit/images/(or existing image dir).Implementation sketch
qemu-img create -f qcow2 -b <base> -F <base_format> <overlay>-drive, base referenced as backing file.overlay: true|false(default false initially).Acceptance criteria
-F rawor explicitly rejected with a clear error).