Retain active-state (closes #201)

[avrovulcanxh607]

Monitored unit active-states are now retained. To try & prevent invalid data getting "stuck" in the broker, we publish "unknown" as the status to every monitored unit on shutdown.

[fphammerle]

Thank you!

Could you add a line to the # Unreleased section of the CHANGELOG.md?

Did you set the executable bit on setup.py intentionally?

[fphammerle]

Suggested change

	for monitored_system_unit in state.monitored_system_unit_names:
	await mqtt_client.publish(
	topic=state.get_system_unit_active_state_mqtt_topic(unit_name=monitored_system_unit),
	payload="unknown",
	retain=True,
	)

I am not sure whether we should report "unknown" on the shutdown of systemctl-mqtt, as consumers should always subscribe to the birth/last-will topic systemctl/[hostname]/status (c3533ee) to be notified when systemctl-mqtt becomes available or unavailable (so "unknown" seems redundant to me and somehow misleadingly suggests that systemd reports "unknown").
What do you think?

[avrovulcanxh607]

What I was really aiming for was to clear the topic of stale services. As it stands, if you decided to change what services are monitored the retained topic stays there "forever" - this was an easy way to at least mark out topics that haven't changed since shutdown. Do you think we need to clear out stale topics or leave that as an exercise of due diligence for the user?

[fphammerle]

Sorry, I missed your reply. Yes, I would leave clearing the topic to the user/subscriber as we can't guarantee that "unknown" will be sent (e.g. on sigkill).