[BUG] - Always on and other "monitor was offline" but it wasn't

Anyone else having this issue?
My Sense was online all day yesterday (friday), not a single drop out, everything worked as usual. But my always on monitor has been off all day? I’ve had Sense open on 2nd monitor since midnight without any issues. Always on bubble is still reporting it’s normal fluctuating 300-1,000w (never understood how the 48hr average can fluctuate so much minute to minute, hour to hour, yet the bar graph remains so flat.). . :roll_eyes:
Main power meter

Submitted Ticket #263297

My other is broke too…

It’s most likely connected to the 12am-1am issue reported in other thread. Notice it’s 12:30 and 12:57 on your time.

Check it after 1AM

Guess it’s a DST issue with Sense, not HA :man_shrugging:t3:


Well this is interesting.

Always on consistent across all hours on Thursday 11/18 as I would expect

Now look at Always on Friday 11/19

Something clearly went South.

It can also be seen in the weekly day by day

Is this a new firmware build :man_shrugging:t3:

Well golly… We may have inadvertently figured out the issue discussed in that thread too…
It’s after 2am and my “Always On” and “Other” monitors are no longer showing offline.
Always on is delayed by 1hr. 12am filled in 1am is now missing.

Sense has been really wonky lately. Out of my 23 natively detected devices only 9 are somewhat accurate anymore.

Everything is lumped into the always on. Other rarely shows usage 1.6 kWh since Sept. 1st…
Some individual days in November show minor usage, but it’s not included in the monthly total.
Must be that new common core math…

1 Like

at about 2am 12am filled in. Sometime between about 2:30am & 2:45am 1am filled in. About 3:05am 2am filled in.
Something really strange is going on…

Appears this always on fluctuation is also really screwing with the stats, making the compare feature utterly useless. Wonder what else.
Comparing my real time always on to other homes always on from a month ago, apples to bananas?

All these captures were taken within a 10 min time frame.

Here’s is what I suspect might be happening, but I don’t know for sure. Usually Always On and Other are CALCULATED in real-time by the Sense monitor working with the Sense mothership to produce the real-time bubbles, etc. The calculations for Other rely on the value of Always On and neither are really direct monitors of actual data, but calculations.

But if the real-time calculation goes askew, Sense has a way to backfill with recalculated data. I have seen this back fill happen for new device models and for other reasons. That might be what you see taking place - and you might not even see the different Always On panels change values simultaneously because they might be refreshed with data separately.

Now why might the Always On calculation go askew ? A few weeks ago I filed a ticket about seeing a different behavior/pattern in my hourly Always On vs what I was calculating / emulating using my Home Assistant data collected from Sense. Turns out Sense had done some tweaking of Always On behind the scenes, but it might have needed some additional adjustment. Blue is Sense hourly, orange is my HA calculation / emulation (with a few HA outages)

BTW-I saw another update / backfill Always On indicator a few days ago. My Always Ons for my smart plugs disappeared for a few days. I filed a ticket, but before anyone at Sense could investigate, those were back again.

Always on and has never made sense to me. I understand Always On is an average over 48hrs and Other is some version of what’s left over. I used to think Other was real time total W - Always On average - some unknown tabulation of knowns = Other.
I totally understand why Always On could (as expected) increase or decrease over a longer period, it’s an average. What I’m confused about is why wouldn’t the average, average itself out over 48hrs and not fluctuate so dramatically within any given 10 minute time period? This isn’t new, If this was an issue with backfilling data one would still think it would average itself out.

Never understood why Sense confuses everyone with “Always On” instead of “Estimated unknown” with a baseline average. Apples and bananas turn the bubble bowl into a fruit salad of oranges. Always on could be useful to know just not how it’s presented.

That’s just one issue, still doesn’t explain the missing other usage in the weekly an monthly views.
I will try and catch my Always On and Other again after midnight to see if this is an every day issue. If anyone else is up at that time additional confirmation would help narrow what @Beachcomber mentioned being discussed in the other thread about the HA issues. I don’t have solar but I highly suspect that is related.

Always On is an an attempt to assess how much power is truly on all the time due to devices that have large Always On on components (many do !) It’s not an average/mean over 48 hours as many people seem to assume - it’s more like a low water mark over the last 48 hours, much much lower than an average would be.

Most other power analysis systems (I use two others), include a similar Always On function. It’s very useful in separating out the origins of your energy usage.

The 12A-1AM data missing/incorrect started after the end of Daylight Savings Time and we “fell back” an hour. @DevOpsTodd and I have had private conversations during the week and it’s clear we are all seeing the issue in some form. Where the bug actually is is debatable.

The Always On issue is a different issue imo.

As I showed above, everything was pretty consistent daily in that data until yesterday, Friday 11/19.

That’s why it’s so confusing it’s a fruit salad.
It’s supposed to be a calculation of vampire loads, devices that consume power when they are not in use. but for nearly everyone it isn’t anything remotely close to that. It’s a garbled mess (low baseline) that includes countless unknown devices that turn on and off regularly. Other is anything undetected above that baseline, but not quite as it’s real time and calculated quite differently. Notice how Other briefly appears before a known device bubble appears, and reflects the known device quite accurately despite the always on baseline being an average of sorts.

Knowing the Average On baseline can be helpful for some picking out vampire loads long term, I get that. What would be much more helpful when actively hunting would be a real time display that showed “unknown” or an “other” that actually showed Real time wattage total minus the total of all the knowns. There is no real time total of all the knowns making it impossible to know exactly how much unknown wattage your home is currently using at that current moment in time. IMO, this makes much more sense and would alleviate this confusion that so many have regarding the always on and other bubble that gets tossed into the “Now” bowl.

Real time Always On and Other bubble is still functioning during this time when Sense reports the Monitor is offline in the historical. If this is also effecting solar (backfilling) that may explain why your real time solar is reporting but your historical is missing in HA. It would depend on how HA grabs that info and back fills with the back filled info from Sense or if HA runs it’s own tabulator for historical.

I agree that it would be useful to show the sum of known waveforms overlaid on the Power Meter to do power sleuthing. But Always On is about as good as it gets if you are taking aim at moderate to large energy vampires, as long as it is working consistently.

I have NO Solar so any backfill of Solar Production is purely fictional

Always On Yesterday

Always On Today


In my theory, you have incomplete backfill for 11/19. Confusing why you would even suggest Solar was involved. Will be interesting to see if / when data for 11/19 changes

It was an answer to a statement

1 Like

Both would have their place. For users that have a large device that is always on, always on is helpful. Though I would Imagine most users large devices cycle, HVAC, heating, motors, etc… Sense is pretty darn good at detecting larger devices. Larger always on devices are the easiest to hunt down. Flipping breakers and watching for a drop/increase in the overall usage is a great method for that. Smaller devices are usually the hardest for Sense to detect. Users have a hard time as well. Going around turning every device on and off looking for a slight increase decrease in overall wattage while praying a detected device doesn’t cycle. Users can’t use always on for that as it’s not real time. This can also lead to users going crazy trying to figure out what devices are always on using so much wattage, not realizing Lights, Tv’s, fans, and all sorts of devices quite often get thrown into the always on pile. None of them are always on or true vampire loads.

Sorry, my fault for only skimming through that thread.

As reported in real time and internet was never down, why would you assume a backfill even be needed?