-
fix(controller): survive installer reruns — keep wizard topology, heal compose drift
Release / frontend (push) Canceled after 0sRelease / linux (amd64) (push) Canceled after 0sRelease / linux (arm64) (push) Canceled after 0sRelease / windows (push) Canceled after 0sRelease / macos (push) Canceled after 0sRelease / release (push) Canceled after 0sBuild and Push Docker Image / build-and-push (push) Canceled after 0sreleased this
2026-06-12 08:05:49 +03:00 The wizard's managed-postgres choice silently reverted to sqlite on
valheim: install.sh's write_compose unconditionally regenerated the
sqlite BASE docker-compose.yml on every rerun/update/repair, dropping
the controller-provisioned db service, while the controller's
.applied-hash shortcut concluded there was nothing to re-apply.- install.sh: when desired-state.json exists, docker-compose.yml is
controller-owned — keep it and only refresh the installer-owned
override (host caps); - controller: when the on-disk compose no longer matches an
ALREADY-APPLIED desired state, log and re-apply (restores a clobbered
topology; never springs a half-finished DB switch — drift is only
consulted for the state whose hash was recorded as applied); - compose: disable the image's HTTP healthcheck on the controller
service (it serves no HTTP — every controller showed a permanent
misleading '(unhealthy)').
Downloads
- install.sh: when desired-state.json exists, docker-compose.yml is