Mycroft Core
Mycroft Core sits at the centre of all Mycroft installations. The code itself includes anything that is generic to all instances of Mycroft.
Mycroft Core includes four or more distinct Services:
Each of these is started in it's own process, and communicates with other processes by emitting and listening to messages on the MessageBus.
Each Service includes an instance of the ProcessStatus class. This provides a consistent interface to set and query the state of a Service.
It is constructed with the following signature:
status = ProcessStatus(name, bus, on_started=None, on_alive=None, on_ready=None,
on_stopping=None, on_error=None)
Arguments
name (str): process name, will be used to create the messagebus messagetype "mycroft.{name}..."
bus (MessageBusClient): connection to the Mycroft messagebus.
on_started (callable): callback to execute when process is started
on_alive (callable): callback to execute when minimum operation is reached.
on_ready (callable): callback to execute when loading is 100 % complete.
on_stopping (callable): callback to execute when shutdown of process starts.
on_error (callable): callback to execute when an unrecoverable error renders the process unusable.Each process tracks its state as one of:
- NOT_STARTED
- STARTED
- ERROR
- STOPPING
- ALIVE
- READY
These states are defined in a comparable order. This means if a Process has the state
ALIVE, then it must also be STARTED. Likewise, if a Process is READY, it must also be ALIVE and STARTED.These states can be set by calling the relevant method from the ProcessStatus class we initialized above:
status.set_started()status.set_error()status.set_alive()status.set_stopping()status.set_ready()
It is up to each Service to determine it's readiness and set the state accordingly.
The status of a Service can be queried via the MessageBus.
The request Message should be in the form -
mycroft.{service_name}.{is_status}. For example, we can check if the Audio Service is ALIVE by emitting:response = self.bus.wait_for_response(Message(
'mycroft.audio.is_alive'))
The response message will include a
status boolean in the data payload:>>> print(response.data)
{'status': True}
If you have thoughts or comments on the following status proposals, please let us know in the Community Forums.
The status/watchdog service would be a new systemd service running on devices that run Mycroft-core. The functionality of this service could include, but would not be limited to:
- Monitoring various aspects of the health of each core service (is service running, healthy, ready, etc.).
- Heartbeats could be emitted from each service, or the new service could check the pulse of each service on regular intervals.
- Actions could be taken if a service becomes unhealthy, such as attempting to restart a crashed service.
- Specific checks, such as the existing microphone watchdog, could be included in this service.
- Responsible for restarting services on a daily basis.
- If a user opts in to data sharing, information about exceptions or crashes could be sent to Mycroft for diagnostic purposes.
Last modified 1yr ago