Strange graph behavior for temperatures

---

Hi,

Since a week or so (or that is when I started to notice it), I have very strange graph behavior for all my temperature measurement, but also some other values whereas the measurements itself are correct.
My pimatic and plugins are all up-to-date.
Edit: I’m running on an RPi3 on Debian Stretch (9) but still with node V4.9.1

In the config I have the following for humidity and temperature:

        {
          "deviceId": "*",
          "attributeName": "temperature",
          "type": "number",
          "expire": "1y"
        },
        {
          "deviceId": "*",
          "attributeName": "humidity",
          "type": "number",
          "expire": "1y"
        },

My general temperature from an esp module for my central heating is measured every 30 seconds and has now over a 100.000 points. (But in the database it says: “5 minutes”??)
esp12-kamer-temperatuur temperature number (continuous) 5min 1y 100912
Despite that I get the following graph for 24 hours. The last 2 hours are correct, but where is the rest?

and the following for 7 days:

This is completely ridiculous.

In the 3rd graph you see a external RF sensor with temp, humidty and battery level with the following in the database:

garagebuitensensor	lowBattery	boolean (discrete)	0	7d	7710
garagebuitensensor	temperature	number (continuous)	5min	1y	84676
garagebuitensensor	humidity	number (continuous)	5min	1y	84668

I get this graph for the last 24 hours:

As you can see: battery level functions OK, but temperature and humidity not.
And I get this graph for the last 7 days:

So I stopped pimatic, made a copy of the db (you never know) and within sqlite3 I did:

pragma integrity_check;
VACUUM;

Seemingly nothing is wrong. But obviously something is wrong.
Is this a bug? Am I doing something wrong?

As I suddenly realised I’m still on node 4.9.1 I decided to upgrade.

I stopped pimatic, did the “sudo n stable” which gave me node V10.16.0, did the sudo npm unlink in node-modules/pimatic, tar-gnuzipped the pimatic-app folder and moved it out of the way, reinstalled pimatic, copied certificates, database and config.json back, and started pimatic.

The graph behaviour is still exactly the same (with a 1 hour break due to the node upgrade and the pimatic (re)installation).

So the node upgrade did not solve it.

Something is really wrong with my database.

I did an export from my living room temperature to csv select id, strftime('%Y-%m-%d %H:%M:%S', time / 1000,'unixepoch') as time,value from attributeValueNumber where deviceAttributeid=100; and opened that in a spreadsheet.

I have a 57 values for 28-07-2019 (relates to above graph), nothing for 27-07-2019, 13 for 26-07-2019, 13 for 22-07-2019 (nothing for dates in between),
3 for 12-07
3 for 06-07 (06 July)
8 for 23-06-2019
etc… etc…

I guess this is my database only, but I would like to ask others to test it as well in case something is acting weird “somewhere” in the software/config.

@harry-van-der-wolf said in Strange graph behavior for temperatures:

But in the database it says: “5 minutes”??

not all points are plotted. so its averaged.

and the other thing: my first guess was that it is exactly plotted what’s in the database. the question is: why are values missing? did you change your system time? are the sensor values transmitted with a timestamp that is / was wrong?

I make nightly backups and my backups go one month back. Also in those backups the data is missing.
Although I did not find errors I did a dump of the database and created a new database using the dump and attached that to pimatic. It was 2MB smaller, but that could also be an optimize step during import.
In the past 1 1/2 day not one single point is missing.
I’m now thinking more and more that it is a local database issue. I know I had 2 system crashes/hangs in the past 2 months or so, but never got errors. Still there must have been some errors I assume.