Behaviour of the integration when a connection is lost

Dehumanizer77 commented 3 months ago

Hi, I'm not sure what has changed, if it's caused by changes in HA or in the integration, but previously when a connection to the controller was lost (e.g. the usb cable was disconnected), all alarm sections immediately changed to "Unavailable". Now this does not happen anymore, I can only detect it using binary_sensor.jablotron_active, but that lags a log (like 30 seconds or so) and all other sensors look like everything is fine (alarm sections keep their previous state etc.). Is there any way how to bring back the previous behaviour?

Dehumanizer77 commented 3 months ago

To further explain, the problem is that if the connection is lost just for a few seconds (like accidentally disconnecting the usb cable and connecting it back immediately), I am unable to detect this and after reconnection the integration does not work anymore, everything looks normal but it does not communicate, until the integration is restarted. Previously I had an automation which would restart the integration when this happened (i.e. the alarm section became unavailable), but now I have no way of knowing that something happened and that the integration needs to be restarted, if the disconnection is shorther than the 30 or so seconds when the jablotron_active sensor catches it.

Dehumanizer77 commented 3 months ago

I have checked logs and I might have an idea what is a problem (or how can it be fixed/changed) I see this: OSError: [Errno 30] Read-only file system: '/dev/jablotron' 2024-05-08 11:07:02.905 ERROR (ThreadPoolExecutor-110_1) [custom_components.jablotron100] Write error: [Errno 30] Read-only file system: '/dev/jablotron' Traceback (most recent call last): File "/config/custom_components/jablotron100/jablotron.py", line 1069, in _keepalive self._send_packet(self.create_packet_command(COMMAND_HEARTBEAT)) File "/config/custom_components/jablotron100/jablotron.py", line 1097, in _send_packet self._send_packet_by_stream(packet) File "/config/custom_components/jablotron100/jablotron.py", line 1100, in _send_packet_by_stream stream = self._open_write_stream() ^^^^^^^^^^^^^^^^^^^^^^^^^ File "/config/custom_components/jablotron100/jablotron.py", line 1114, in _open_write_stream return open(self._serial_port, "wb", buffering=0) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ which means the integration is trying to write to the device while it's not available (resulting in the "read-only filesystem" error, since it can't create anything in /dev of course) Maybe it would be to sufficient to catch this exception and set the whole alarm to unavailable if this happens.

Dehumanizer77 commented 3 months ago

Actually, the binary_sensor.jablotron_active is my own sensor, which I was trying to mitigate this issue with, so please disregard that. But everything else is still valid.