SharkIQ Integration loses connection frequently; integration must be reloaded daily.

The problem

My Shark vacuums lose connection with home assistant frequently, approximately once per day (but I have not timed it). They show ‘unavailable’ state. Reloading the integration fixes it immediately and the correct state is shown. From logs it appears to be an authentication issue.

Environment

• Home Assistant Core release with the issue: 2020-12-1
• Last working Home Assistant Core release (if known): never
• Operating environment (OS/Container/Supervised/Core): Supervised
• Integration causing this issue: sharkiq
• Link to integration documentation on our website: https://www.home-assistant.io/integrations/sharkiq/

Problem-relevant configuration.yaml

none, set up with UI

Traceback/Error logs

There are over 1200 instances of the word ‘shark’ in my logs, but this seems to be the one causing the problem. If you’d like all of the shark errors I can dive deeper.

2020-12-20 13:16:31 ERROR (MainThread) [homeassistant.components.sharkiq] Error fetching sharkiq data: Ayla Networks API authentication expired.  Re-authenticate and retry.

Additional information

Since this is seemingly an issue with Ayla Networks and not HA, perhaps adding a service to reload the integration without visiting the integrations page would be a good workaround. Then I could simply automate that if my vacuums become unavailable the integration is reloaded.

[dmatora]

Fixed it with automation that reloads integration once robot becomes unavailable

shell_command:
  shark_restart: "/usr/bin/curl -X 'POST' -H 'authorization: Bearer <token>' http://localhost:8123/api/config/config_entries/entry/<entry>/reload"

alias: SharkIQ
description: ''
trigger:
  - platform: state
    entity_id: vacuum.baby_shark
    to: unavailable
    for: '00:05:00'
condition: []
action:
  - service: shell_command.shark_restart
    data: {}
mode: single

<token> is long lived token I’ve generated in profile section) <entry> is integration id I’ve captured using chrome inspector while reloading integration from gui

[ajmarks]

Hi guys, sorry again for not fixing this yet, but I wanted to at least give everybody an update. The issue is in the authentication flow. The login credentials are good for 24 hours, and then they need to be refreshed. That refresh is what’s not working, but it is why the workaround of restarting the integration fixes it. I’ve been slammed with a new job the past few months, but I feel really bad that this is still broken, and I’m going to fix this as soon as I can.