Document the host network prerequisites for updates #50
Loading…
Reference in a new issue
No description provided.
Delete branch "docs/update-host-prereqs"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
The updater will drive
wlan0over D-Bus to raise an access point for nodes to pull firmware from, so the host has to have NetworkManager owning the radio.systemd-networkdis already enabled on this machine forcan*, which makes that worth checking rather than assuming.Also says updates are meant to be finalised over Ethernet, and that running one over WiFi costs the Home Assistant connection from elsewhere on the network until it finishes.
First commit is the existing untracked
rpi-setup.mddraft as-is, so the second commit reads as the change. Documentation only — nothing here configures the host, and there is no code to test.Closes CCS-UHA-9.
Two points inline, neither blocking.
@ -0,0 +28,4 @@RestartSec=100ms```Firmware updates need NetworkManager to own `wlan0`: the updater drives it overNine lines of prose where the rest of the file is one short line per command block. The Ethernet and WiFi paragraphs need no commands so some prose is unavoidable, but this reads a register looser than the file. Roughly half the words would carry the same facts, e.g. "Firmware updates need NetworkManager to own
wlan0; the updater drives it over D-Bus to raise the access point nodes pull firmware from. Bookworm does that by default, butsystemd-networkdis enabled above, so confirm:"@ -0,0 +35,4 @@```nmcli general statusnmcli device statusThe check has no pass criterion, so you can run it and still not know. Add what a good answer looks like —
nmcli device statusshould showwlan0asconnected/disconnectedrather thanunmanaged— and ideally a word on what to do if it isn't.Round 2: the two earlier points are taken. Two smaller ones from the new wording.
@ -0,0 +28,4 @@RestartSec=100ms```Firmware updates drive `wlan0` over D-Bus to raise the access point nodes pullTwo nicks from the tightening: "the access point nodes pull from" garden-paths on "access point nodes" (the earlier "the access point the nodes pull firmware from" didn't), and "Bookworm does by default" is missing its object — "does so by default".
@ -0,0 +37,4 @@nmcli device status````wlan0` must read `connected` or `disconnected`, not `unmanaged`. Unmanagedunavailableis the third managed state and it's a likely one on a fresh Pi — NetworkManager owns the radio but rfkill has it soft-blocked because no wireless country is set. The criterion as written sends you hunting for a/etc/systemd/network/unit that isn't there. Onlyunmanagedmeans something else claims it, so: "wlan0must not readunmanaged."All three taken, nothing further from me.