Core: Unifi presence detection ignores detection_time

[aaronwolen]

Home Assistant release with the issue: 0.102.2

Last working Home Assistant release (if known):

Operating environment (Hass.io/Docker/Windows/etc.): Both HASS and the Unifi controller (v5.12.35) are running in docker on Ubuntu server v19.04. Here are the relevant entries from docker-compose.yaml:

  unifi:
    container_name: unifi
    image: jacobalberty/unifi:latest
    restart: always
    depends_on:
      - mongo
    networks:
      - unifi
    ports:
      - '10001:10001/udp'
      - '3478:3478/udp'
      - '8080:8080/tcp'
      - '8443:8443/tcp'
      - '27117:27117/tcp'
    volumes:
      - ./unifi/data:/unifi/data
      - ./unifi/log:/unifi/log
      - ./unifi/cert:/unifi/cert
      - ./unifi/init:/unifi/init.d
      - ./unifi/var/run:/var/run/unifi
      - /etc/localtime:/etc/localtime:ro
    environment:
      DB_URI: mongodb://mongo/unifi
      STATDB_URI: mongodb://mongo/unifi_stat
      DB_NAME: unifi
      TZ: America/Chicago
      UNIFI_UID: 999
      UNIFI_GUID: 999
      RUNAS_UID0: 'false'

  hass:
    container_name: hass
    image: homeassistant/home-assistant:stable
    volumes:
      - ./hass:/config
      - /etc/localtime:/etc/localtime:ro
    restart: always
    network_mode: host
    environment:
      TZ: 'America/Chicago'
    devices: 
      - "/dev/ttyACM0"

Integration: https://www.home-assistant.io/integrations/unifi/

Description of problem: Devices tracked by the Unifi integration are marked as not_home well before the amount of time specified by detection_time has elapsed.

Problem-relevant configuration.yaml entries and (fill out even if it seems unimportant): The Unifi component was configured via the Integrations menu, and detection_time was set to 400 seconds using the integration's config options menu. I confirmed the setting was recorded in core.config_entries:

I also attempted to increase detection_time in configuration.yaml see below) to see if that helped but this had no effect.

unifi:
  controllers:
    - host: localhost
      site: !secret UNIFI_SITE_ID
      detection_time: 600

Traceback (if applicable):

Additional information:

Having compared the state changes recorded in HASS' database to the connect/disconnect events reported by Unifi, it looks like tracked devices are marked as not_home immediately, without taking detection_time into account at all. Here's an example for one of the affected devices:

Unifi Events

Database

(Note: UTC times were converted to my local time zone to facilitate comparison)

[probot-home-assistant[bot]]

Hey there @kane610, mind taking a look at this issue as its been labeled with a integration (unifi) you are listed as a codeowner for? Thanks!

[Kane610]

Thanks for reporting, I will look into this in a couple of weeks

[mjmccans]

I have been struggling with the same issue and for me it ended up being related to the Unifi Controller bug referenced in the is_connected method in the device_tracker.py file where a disconnected wireless device still shows it connected as a LAN device. This was also leading to some devices toggling between Home and Not_Home many times per hour. In that circumstance the current code will immediately mark the device as not connected and will hot honour the detection_time set by the user.

While a bit sloppy, I was able to fix my issue with some changes to the is_connected method to store a temporary timestamp if self.is_wired != self.client.is_wired and use that instead of last_seen. I have attached my updated code for the method, which has been working well for me for about a day now without any issues. This may not be the best way to solve the issue, but it does the trick for me and handles the Unifi Controller bug while still honouring the detection_time option set by the user.

UnifiBug.txt

[Kane610]

@mjmccans great input!

[rwarner]

I'm also seeing the exact same situation on my end. Same variables set in core.config_entries appropriately and everything. I'm also easily able to see when the device switches APs in the house that this has been happening and triggering the not_home on my device tracker / Person.

I'll have to try your change @mjmccans -- would hate to disable all my automations in the meantime. Would there be an alternative workaround?

[rwarner]

Just an update -- I integrated @mjmccans fix into my local Home Assistant python Unifi script and I haven't been experiencing these disconnects in the past two days.

[Kane610]

@rwarner that's great, I just pushed a fix based on @mjmccans fix. It will probably be a part of the 103 release

[mamartins]

Was this fix already deployed?

I have been experiencing the same issue, devices that are disconnected that still show a 'ap_mac' instead of "None". Sometimes it works, others don't. I found this because my automations based on specific AP connections are messed up.

[Kane610]

Yes this was released with HASS 0.103

[mamartins]

them this is another bug. I'm I the only one with this broken behaviour?

[Kane610]

Possibly Noone else has reported it