ffind-dev/pve-ups v3.5.0

on GitHub

Added

• Proxmox Backup Server as a shutdown target. Each host now has a type (Proxmox VE or
  Proxmox Backup Server) that decides how it is talked to. PBS uses its own
  PBSAPIToken=<id>:<secret> header scheme and needs the Sys.PowerManagement privilege
  on /system/status, which is why entering one as a PVE host used to fail as an invalid
  token. The node name is a free label for PBS entries: PBS ignores the node in the API
  path, so the shutdown always addresses /nodes/localhost/status.
• Self-test results per host. The scheduled credential check now records its outcome
  for every target (credentials_ok, power_mgmt_ok, last_test_at, last_test_error).
  They appear in /api/status and in the dashboard's host table, so an expired token is
  visible long before an outage needs it.
• /api/health reports the shutdown targets: hosts_total, hosts_ok,
  hosts_selftest_ok and hosts_selftest_at. The status field and the HTTP code are
  deliberately unchanged — they still track the engine, not the credentials.
• Live preview of the shutdown sequence under the host list in the wizard. It shows the
  order the engine would actually use, including which hosts share a stage and therefore go
  at the same time — previously that was only readable in a tooltip, and the staged
  behaviour was not visible at all. The Order field now also says 0 = first in its
  label, and it is hidden while This host is ticked: that flag is the first sort key, so
  a marked host is last whatever number it carries. The value is kept, so unticking
  restores it.

Changed

• Hosts are shut down in stages instead of strictly one after another. Targets sharing
  a shutdown order are now commanded at the same time, and the appliance's own host still
  forms the final stage. Together with a hard per-target deadline this means one machine
  that stops responding can no longer delay the other hosts, the poll loop or the battery
  countdown.
• The feed diagram shows a host's name only; the product name is no longer prefixed, which
  kept overflowing the node. The type stays visible as a chip in the dashboard's host table
  and in the host card's heading, shortened there to Proxmox BS so a Backup Server row is
  no wider than a Proxmox VE one.
• This host is hidden for Proxmox Backup Server entries in LXC deployments, where the
  combination is impossible — an LXC never runs on a Backup Server. It stays available in
  Docker deployments, where the container may genuinely sit on the PBS and the mark decides
  whether it shuts itself down last.