SoC does not get updated

[daniel-zullo-frequenz]

What happened?

An application kept reporting 100% SoC while the real SoC queried from the components was around 20%. It seems the application kept the outdated SoC as there wasn't any update for this metric. Apparently the SoC was no longer reported when new PV inverters were added to the microgrid (the PV inverters were reporting only Error state). It took around 5 days to find out about the issue when the components low SoC alert triggered. Unfortunately there is no logs available.

What did you expect instead?

The SoC is reported and it is up-to-date

Affected version(s)

Observed in v1.0.0-rc601

Affected part(s)

Microgrid (API, component graph, etc.) (part:microgrid)

Extra information

The SoC was reported correctly once the application was restarted

[llucax]

We observed this issue when the component graph was misconfigured as follows (meter_3 was added to the component graph):

      grid_1
     /        \
meter_2    meter_3

meter_3 was added by mistake and wasn't really a meter at all, so it was reporting no data. Because of this the SDK gave up on calculating the grid load.

This shouldn't affect how the SoC is calculated, but it might be relevant and some unexpected side-effect.

[shsms]

We have tasks that calculate the SoC and stream it. My theory is that the tasks must have crashed because of an exception.

The only real case I can see now is if both soc_upper_bound and soc_lower_bound are equal. I'll add a fix for that.

frequenz-sdk-python/src/frequenz/sdk/timeseries/battery_pool/_metric_calculator.py

Lines 400 to 402 in 2e7a096

	soc_scaled = (
	(soc - soc_lower_bound) / (soc_upper_bound - soc_lower_bound) * 100.0
	)

But I am continuing to look for other ways to reproduce.

[llucax]

We are just speculating about #1081 being the fix for this, right? I guess we can still leave it closed and reopen if it happens again.