search

mKeRix / room-assistant

Presence tracking and more for automation on the room-level
https://www.room-assistant.io
MIT License
1.27k stars 122 forks source link

SW version diverges on heterogenous clusters #317

Open embcla opened 3 years ago

embcla commented 3 years ago

Describe the bug On clusters composed of Home Assistant and raspberry pis, it is possible to set the HA instance to auto-update. If this option is set, HA will automatically update RA while the raspberry pis would not. Depending on what goes on with the versions, they may stop talking to each other (it just happened from 2.10.1 to 2.11.0) To reproduce Connect two instances in the same cluster, one must be able to auto-update. Then issue a new version that breaks compatibility: one will update while the other won't. Only way to notice is that the two instances will stop communicating, no other warning or message is seen in the logs.

Relevant logs

The logs show how more than a minute has passed since the last log line but nothing else is shown in the logs.

pi@IOT-RaspberryPi-01:~ $ sudo systemctl status room-assistant
● room-assistant.service - room-assistant service
   Loaded: loaded (/etc/systemd/system/room-assistant.service; enabled; vendor preset: enabled)
   Active: active (running) since Mon 2020-10-26 16:07:22 CET; 1min 15s ago
 Main PID: 3193 (node)
    Tasks: 11 (limit: 4915)
   CGroup: /system.slice/room-assistant.service
           └─3193 node /usr/local/bin/room-assistant -v

Oct 26 16:07:29 IOT-RaspberryPi-01 room-assistant[3193]: 26/10/2020, 16:07:29 - debug - EntitiesService: Adding new entity status-cluster-leader
Oct 26 16:07:29 IOT-RaspberryPi-01 room-assistant[3193]: 26/10/2020, 16:07:29 - debug - HomeAssistantService: Registering entity room-assistant-garage-status-cluster-leader under homeassistant/sens
Oct 26 16:07:29 IOT-RaspberryPi-01 room-assistant[3193]: 26/10/2020, 16:07:29 - debug - HomeAssistantService: Sending new state 1 for room-assistant-garage-status-cluster-size
Oct 26 16:07:29 IOT-RaspberryPi-01 room-assistant[3193]: 26/10/2020, 16:07:29 - debug - HomeAssistantService: Sending new state none for room-assistant-garage-status-cluster-leader
Oct 26 16:07:29 IOT-RaspberryPi-01 room-assistant[3193]: 26/10/2020, 16:07:29 - info - NestApplication: Nest application successfully started
Oct 26 16:07:29 IOT-RaspberryPi-01 room-assistant[3193]: 26/10/2020, 16:07:29 - debug - HomeAssistantService: Sending new attributes {"nodes":["garage"]} for room-assistant-garage-status-cluster-si
Oct 26 16:07:29 IOT-RaspberryPi-01 room-assistant[3193]: 26/10/2020, 16:07:29 - debug - HomeAssistantService: Sending new attributes {"quorumReached":true} for room-assistant-garage-status-cluster-
Oct 26 16:07:29 IOT-RaspberryPi-01 room-assistant[3193]: 26/10/2020, 16:07:29 - info - ClusterService: Added 192.168.101.4:6425 to the cluster with id server
Oct 26 16:07:29 IOT-RaspberryPi-01 room-assistant[3193]: 26/10/2020, 16:07:29 - debug - HomeAssistantService: Sending new state 2 for room-assistant-garage-status-cluster-size
Oct 26 16:07:29 IOT-RaspberryPi-01 room-assistant[3193]: 26/10/2020, 16:07:29 - debug - HomeAssistantService: Sending new attributes {"nodes":["server","garage"]} for room-assistant-garage-status-c

Relevant configuration

global:
  instanceName: server
  integrations:
    - homeAssistant
    - bluetoothClassic
bluetoothClassic:
  addresses:
    - mac1
    - mac2
    - mac3
debug_level: verbose
cluster:
  autoDiscovery: false
  peerAddresses:
    - 'ip:6425'

Expected behavior Many options are possibile

Environment

Additional context

mKeRix commented 3 years ago

I'm honestly a bit surprised that the update broke connectivity, I can't remember changing anything in the networking layer of the code. 🤔

I like the idea of at least logging when mismatching versions are detected though.

embcla commented 3 years ago

Could it be something else that I picked up at the same time then? It auto-updated and somehow the two instances weren't talking anymore