groupBy time not working with stream

,
1

Hi,
I want to group data by time using the stream setup.
(The batch and query nodes are not acceptable unless there is a way to wildcard the measurement. - If there is, I would like to know)

The data that is fed via telegraf consists of A 30 second burst every 15 minutes (beginning by the 15 minute mark).

My goal is aggregation of those INCOMPLETE 15 minute windows.

The Influx query over the entire time series works as expected:
select spread(x),spread(y),min(x), min(y),… from (m) group by time(15m)

Re-implementing it using streams fails

//Some UDF Processing per Point FOR ALL measurements: Works
var data = stream
  |from()
    .database('db')
    .retentionPolicy('autogen')
    .groupBy('tag') 
  @udf()
  |eval(lambda: string("tag"))
    .as('tag')
    .tags('tag')
    .keep('x', 'y','z', 'w')  

 //Writing in other database with same measurement: Works as well    
data  
  |influxDBOut()
   .database('dbOut')
   .precision('ms')

I am currently trying and failing to produce groups:

// This fails/ does not produce any group
//Reducing the every duration to less that 30s appears to trigger some groups
var groupWindow = data 
  |groupBy()
  |window()
    .period(15m)
    .every(15m)
  |log()

//This seems to be causing an error. Should that work?
var groupGroupBy = data 
  |groupBy(time(15m))
  |log()

How should I go about grouping by 15 minutes?

Best regards.

2

Hello @Christoph1,
I’m not sure. @Emrys_Landivar any ideas?

3

@Christoph1
When you group, what do you mean? Do you mean you want to take some aggregate over a 15 minute window, like a sql group by, or do you mean you want to window the data into a kapacitor group?

4

Thank you for taking an interest,

I fail to see the difference, in what way is a kapacitor group different from an InfluxQL group by?

I want to form aggregates over a 15 minute windows.
The input arrives unordered, though.

5

Ok, then you don’t necessairly want a groupBy, you want to window the data into 15 minute chunks, then you want to use a sum or count or etc after that.

influxql group by takes a window of data and clumps it together by certain keys for later processing.

6

I think I can answer that with a careful ‘yes’.

As I understand kapacitor, “window” implements the influx “group by”.

But it does not work with irregularly arriving data.
There is no “snapshot every 15 minutes” and continue as whole batch from there