Sample application of how to read grib2 files and insert the data to influx.

Use at your own risk. I suggest you use this project for inspiration, not as a finished product.

grib2influx

A cli tool to parse forecasts (or other sections) from grib-files and save to influx as timeseries.

Usage

Usage of ./grib2influx:
  -database string
        Database name to use. (default "forecasts")
  -gribfile string
        Gribfile to import. If no gribfile specified, start server mode.
  -influxHost string
        Hostname for influxdb. (default "http://localhost")
  -influxPassword string
        Password for influx.
  -influxUser string
        User for influxdb.
  -influxport int
        Port for influxdb. (default 8086)
  -port int
        Server port no, if servermode. (default 8080)

Testing and an example

To test the program locally you can use docker-compose.

First, build the binary:

$ make

Then start docker-compose:

$ cd docker-compose && docker-compose up -d

Insert data from testdata grib-files:

$ ./grib2influx  -gribfile testdata/gfs.t00z.pgrb2.2p50.f003
$ ./grib2influx  -gribfile testdata/gfs.t00z.pgrb2.2p50.f006
$ ./grib2influx  -gribfile testdata/gfs.t00z.pgrb2.2p50.f009
$ ./grib2influx  -gribfile testdata/gfs.t00z.pgrb2.2p50.f012

Now, go to http://localhost:3000 and login with admin/admin. Add influx as a datasource(this step wil be obsolete when grafana officially adds support for adding datasources in docker) and start playing with your data:

Detailed description

• parse the grib2-file using the griblib, a golang library to read grib2-files
• create points from the data and insert them into influx.
  • series-name is the time of the forecast
  • values are added with the respective category name + coordinates

From one single grib-file you will end up with one series with as many values as categories in the grib-file(wind, temperature etc) multiplied with the number of coordinates: number_of_categories * number_of_coordinates

Walkthrough

fields from the grib-file

• Forecasttime vs valuetime

  • forecasttime: Section1.ReferenceTime + Section1.ReferenceTimeSignificance
  • offset hour in filename for valuetime is used to calculate the valuetime of the gribfile data

• What kind of data is this anyway? Section0.Discipline + Section4.ProductDefinitionTemplate.ParameterCategory

  • use griblib.ReadProductDisciplineParameters to resolve the name

• Calculating the coordinates of the data (Section3) is done manually

• insert 1 batch per message. 1 message contains typically 10 512 data points if the grid covers the whole planet. One file usually contains 77 messages.