Having issues where Rhasspy seems to get stuck in a bad state after some time has passed and no longer will wake. When I check my Rhasspy logs, the last thing I generally see when I get stuck in a bad state is:
[DEBUG:2020-08-23 03:30:28,906] rhasspywake_raven_hermes: Receiving audio
Currently using Raven, but have experienced the same issues with other wake word systems.
Some info on my setup below, mainly hoping for some suggestions on where I should be looking for warning signs in my logs or suggestions on how to debug this.
Base server - Runs a mosquitto MQTT instance, and a Rhasspy instance in charge of STT, TTS, Intent Recognition, Dialog Handling, and Intent Handling.
Satellite - Runs a rhasspy instance handling wake word detection (Raven), with arecord for audio (via a Matrix Creator hat for my Pi 3B+) input, and Intent Handling. Audio input/output is using UDP on port 12202.
Upon starting up both servers, my first few attempts at waking my satellite work as expected. Wake word detection seems to be relatively good (80-90%), and once woken, my intents are recognized and executed as expected.
What Doesn’t Work
After some period of time, or perhaps some number of wake attempts, things just seem to go to haywire. I can attempt to wake 5-10 times consecutively, not seem to get a beep_hi, then 5 minutes later get consecutive beeps that seem as though they’re in response to my earlier attempts at waking (with no significant noise in the room between those times).
Also if left running for an extended period of time, my MQTT server (running via docker) seems to get overloaded (causing my entire server to slow to a near halt). Restarting the docker instance seemed to immediately free things up. I’ve only had this happen once so far, so I’m not sure if it’s part of this same issue, but figured it was worth mentioning.