[URGENT] Aggregate within window gives unexpected data points

InfluxDB 2
1

Hi,

We have been a customer of InfluxDB Cloud v2 for a few years now and have been using it for air quality IoT.

We have encountered an issue with data aggregation that is affecting many of our customers; since we have ~1,000 devices, and customers check daily or monthly numbers, this adds up very quickly.

What we’re trying to do
We want to get averages of data within a selected time window. E.g., a customer wants to see the 8-hour average of PM2.5 dust particles in the air collected from 8th March 2025 at 07:15 India Standard Time (IST) to 9th March 2025 at 07:15 IST.

Expected results
We expect to see 3 data points within the 24-hour period – one for each 8-hour window, namely:

  1. 8th March 2025 at 15:15 IST
  2. 8th March 2025 at 23:15 IST
  3. 9th March 2025 at 07:15 IST

Actual results

We receive the following 4 data points:

  1. 8th March 2025 at 10:30 IST
  2. 8th March 2025 at 18:30 IST
  3. 9th March 2025 at 02:30 IST
  4. 9th March 2025 at 07:15 IST

What we have tried so far

We tried using the function *timedMovingAverage(every: 8h, period: 8h)* …

  import "timezone"
  option location = timezone.location(name: "Asia/Kolkata")

  from(bucket:"EVIN")
    |> range(start: 1741418100, stop: 1741504500)
    |> filter(fn: (r) => r._measurement == "Pollutant")
    |> filter(fn: (r) => r._field == "PM2_5")
    |> filter(fn: (r) => r.DeviceID == "AirSENCE-062409E640561")
    |> timedMovingAverage(every: 8h, period: 8h)
    |> yield()

Additionally, we have unsuccessfully tried to replace timedMovingAverage with:

    // Alternate #1 (2 variants - with and w/o timeSrc)
    |> aggregateWindow(every: 8h, fn: mean, createEmpty: false [, timeSrc: "_start"])

and with

    // Alternate #2
    |> window(every: 8h)
    |> reduce(fn: (r, accumulator) => ({<get average>}), identity: {count: 0.0, avg: 0.0} )

However, none of these functions work as expected!

Could you please help us create an appropriate query to use a user-defined time window for aggregate operations?

Regards,
George

2

FYI: I’ve reviewed other community topics and tried to implement their suggested solutions, (e.g., using timeShift), most notably this topic:

3

Hello @airsence
Did that work for you?
How are you fairing now?

Ps I recommend checking out InfluxDB 3 Core and Enterprise Python Processing Engine it’s pretty cool.
We’re just getting started but you’re able to quickly build so much especially paired with an LLM.