[ibiltari]

ClickUp: 869fhghhe · Packaging: debian/mock-nodeconf-dropin (PR against debian/bookworm)

Problem

On mock boxes (dev.stagelab.coop) cuems-nodeconf is stubbed out (ExecStart=/bin/true). The real daemon would rewrite network_map.xml from whatever avahi hears on the LAN, and on a public VPS that would wipe the box's invented nodes. So the engine refused every adopt/unadopt from the UI with "cuems-nodeconf is not running", and the node adoption UI (#32) could not be exercised there.

Change (mock script only)

New scripts/mock_nodeconf.py, run as python3 -m cuemsengine.scripts.mock_nodeconf [--map PATH]:

• Serves /tmp/nodeconf.ipc (NNG Rep, mode 0666) like cuems-nodeconf, and answers every request.
• nodelist_modify ADD/REMOVE flips <adopted> in network_map.xml. Same rules as nodeconf: an offline node cannot be adopted, the master cannot be unadopted, an unknown uuid is an error, and "already (un)adopted" is OK.
• The rewrite is validated against the network_map.xsd the engine reads with, then renamed atomically over the file. Other fields are kept verbatim.
• No discovery at all: the nodes the box lists stay as they are.
• Removes the socket on exit, so the engine refuses at once instead of waiting its 15 s IPC timeout.

Test plan

• black, isort, flake8
• local: the 7 request cases against a copy of dev's network_map.xml; ADD+REMOVE leaves the nodes block identical
• dev.stagelab.coop through apache /ws, as the settings page sends it: ADD → OK + <adopted>True, ADD again → OK, REMOVE → OK + <adopted>False, REMOVE master → "Cannot unadopt master node"; the engine logs "network_map reloaded after a node list change"; load/GO/STOP unaffected