How to modify time series data

grafana

#1

Hi,

I have been playing with the format of the data and in this process some wrong temperatures have been uploaded to grafana (temperatures at arround 5000°C :o| ).

  • Is it possible to edit the time series to modify some parts of it directly in grafana or do I have to export then edit then import?

Thanks for your help


#2

Dear Lorenzo,

thanks for writing in. Due to its nature as timeseries database, InfluxDB makes it difficult to edit data. However, we will see what we can do and try to get back to you soon.

Cheers,
Andreas.


#3

Dear Lorenzo,

TLDR; Updating values is not possible in a InfluxDB-native way, but it is certainly doable when bringing in some additional tooling.

Issue

There are some resources around (see below) which show us InfluxDB does not support updating values natively. Bummer!

Workaround

Instead, you would have to export the data (favorably in line protocol format), edit the raw export file and reimport the data again - as you expected already. It seems to be possible to overwrite existing data, though.

Resources

Tools

Cheers,
Andreas.


P.S.: If you can tell us in which database measurement we can find the faulty data and how you would like to edit the value of its temperature field, we will see what we can do for you.


#4

Couldn’t we leave the data in the DB but flag it as inappropriate? For the short run the whole dataset?


#5

Thanks for your suggestions.
As I pursued my exporation with Grafana I found it’s easy to import data into it. Therefore the option of exporting and re-importing again sounds quite straight forward, and I can do it myself. Which is great.


#6

Cool, we didn’t use that feature of Grafana before, i believe. We are happy to hear how this works in general and whether it was helpful for resolving the issue you have been running into on your own.


#7

As influxdb is capable of storing / handling boolean as a field value type, this idea is something, we should deploy someday. But it all depends on whether

  • recording boolean during write of the whole set or
  • exporting / manipulation / re-importing it all subsequently a previously recorded time series
  • or even both! ;)

This is something I want to see e.g. a ‘beekeper present / not present’ flag done with, besides masking unplausible values.