I have updated to beta4 with settings kept from beta1. Still having issues. As you can see most of my devices shows up as unknown and connect top column is empty and mesh page does not show the "paired" flint 3 (even though when i log into flint 3 with ip address it shows everything is okay and mesh is working fine).
I have asked claude to SSH into my flint 4 and investigate based on my previous findings and it fixed the issue! Here is what it said:
On the main router the prplmesh controller (beerocks_controller) was never starting. Only gl-mesh was running, and /tmp/beerocks contained nothing but uds_broker.
The cause is that the 6 GHz radio had its channel saved as 0 instead of auto:
wireless.MT7990_1_1.channel='auto' # 2.4G
wireless.MT7990_1_2.channel='auto' # 5G
wireless.MT7990_1_3.channel='0' # 6G
hostapd and mtkdat treat 0 the same as auto, so 6 GHz works normally (ACS picked ch 53, 320 MHz). gl-wld takes 0 literally, though. It loops forever on:
gl-wld: (wld.lua:633) radio channel not ready: rax0 channel: 0
gl-mesh: [WLD-SYNC] query initialized state completed, trigger on_active Initialized false
gl-mesh: [AMX] subscription activation failed 2 path X_PRPLWARE-COM_WiFiController.Network.Device ... failed to subscribe AMX events: -1
WiFi.Initialized stays false, so gl-mesh never starts the controller or agent. The node list and the "Connect To" column both come from the controller's data model, so they show up empty.
The Flint 3 was never actually unpaired. It was still in /etc/config/gl-mesh and kept retrying, but the main router rejected it every 3 s:
gl-mesh: (vendor.lua:541) not found device 94:83:c4:b0:9a:bd
Workaround:
uci set wireless.MT7990_1_3.channel=auto
uci commit wireless
Within a few seconds gl-wld reported "wld initialized" and the controller and agent started. The Flint 3 rejoined by itself (join state report ... is_new_join: 0), with no re-pairing needed.
Suggested fix: gl-wld should treat channel=0 as auto, the same way hostapd.lua and mtkdat.lua already do. I upgraded from beta1 with settings kept, so the 0 may have come from the old config or from config migration.
The "Unknown" client names are a separate issue, and I haven't looked into those yet.