Skip to content

HA integration recovery after a Finn host reboot

URL: https://mkdocs.justinsforge.com/memory/general/reference_ha_post_reboot_recovery/

When Finn (the Proxmox host) reboots, the Home Assistant VM (CT/VM 103-adjacent, forge-homeassistant VMID 100) comes back before Finn's pveproxy (port 8006) is ready. The proxmoxve integration loses the connection during that window and stays in setup_error ("Host is unreachable" / Errno 113 to 192.168.86.50:8006). It does not self-heal. Every Proxmox-derived entity (per-VM cpu/mem/disk sensors, start/stop buttons, status binary_sensors, switches) goes unavailable as a result. On 2026-05-29 this was ~207 entities (350 unavailable total dropped to 143 after the fix).

Fix (one step)

Reload the config entry, no restart needed:

POST /api/config/config_entries/entry/<proxmoxve_entry_id>/reload

proxmoxve entry_id as of 2026-05-29: 01KNAP0HMGRP0JYP5E1KYE6C1D. Or in the UI: Settings → Devices & Services → Proxmox VE → ⋮ → Reload.

To find failed entries authoritatively, use the WebSocket API config_entries/get and filter state != "loaded" (the REST /api/error_log returns 404 on this HAOS build).

Notes

  • The .50 host IP in the integration is correct (it connects fine once Finn is up); it is NOT a misconfiguration, despite Finn's primary IP being 192.168.86.67.
  • Cloud integrations (Alexa Media Player, Eight Sleep, Garmin) self-recover after the reboot; they do not need a reload.
  • androidtv (192.168.86.208) sits in setup_retry whenever the TV is powered off; self-heals when the TV turns on. Not a failure.
  • Curl/wget are blocked in the Console sandbox; hit the HA REST/WS API with Python urllib/websockets instead.

[Claude Code]