I have several values i want to view over al longer time. If I use the history panel and choose the entity with a long time span (months in some cases) the view is supplemented from the long term statistic.
When I make a history graph card and set the “hours_to_show” to more than 250, the extra is simply cut off. This looks deliberate. If so, it ought to get mentioned in the documentation.
Add on question. Can I just add the two lines
recorder:
purge_keep_days: 60
to my configuration.yaml and reboot without breaking anything?
What is the state class of the sensors? Long term statistics are only stored for sensors with the state class measurement, total and total_increasing. For others the cut off will be at around 250 hours because 10 days is the default.
Thanks for trying to help, but as I clearly said I did use the history panel and it works. Where it doesn’t is in the history graph card inside another dashboard. Fair enough, but the documentation should say so.
But neither the history card nor the statistics graph card will show them if there aren’t any!
I agree, the docs are unclear. History graph card doesn’t mention long term statistics at all. Statistics graph card says “statistics are gathered every 5 minutes and also hourly for sensors that support it” (not helpful). History dashboard mentions long term statistics sampling, but says nothing about which sensors will have long term statistics (there is a link to the Data Science Portal). You have to dig around in the Developer Docs to find anything useful.)
Thank you. One can pare down the statistics to show nothing but a line like history does. I wasn’t aware of that. Not sure yet yet, whether I like the rounding or not, but for longer times it makes no difference anyway.
It is the history card
that I expected to behave the same as the history panel. Mentioning the ten day limit may be helpful here.
Thank you
Yeah just must be an old version? This feature was added to the history card much later than it was added to the history panel, so there’s a range of old versions where this might be true.
Define “old”. I am in Docker, where updating is a pain. But I have done it recently and am on version 2024.9.3, which I consider quite new.
I did try out HAOS on a test system but found a couple of quite serious drawbacks. (Can’t boot from harddisk on a Raspberry Pi3B and can’t use WLAN channel 13. Both work flawlessly using Raspian and Docker.)
Yes. That of course is the very page I got those entries from. What it does not say there is what happens if I add “recorder:” to a configuration.yaml that didn’t have one before. Do a lose all my old data? With entities this is just what happens in comparable cases.
That’s why I ask here before meddling with that setting.
Look, if the history panel and the statistics card both show the long term statistics, they are there and they get written. (They were not initially, but I added “state_call: measurement” quite some time ago and they are present from that moment on.)
That is seven major versions behind the current version. It will be eight versions behind in a few days. So no, that is not “quite new” and is the reason it is not working for you.
