Line interpolation (or connection of points) is not working in TimeSeries when aggregating

I am using Grafana 9.0.2 and since the upgrade from v8 I don’t see any difference between lines/points in the time series. The points are not connected anymore.

Here are the two options with basically the same output. Does anyone know why?

OK I just found out it’s caused by the last line |> aggregateWindow(every: v.windowPeriod, fn: mean) but I need some aggregation, otherwise I get “too many data points” when I use large time windows.

What is the recommended way to do this kind of aggregation? I am still very confused about flux (used SQL in Grafana 8).

depends. aggregation by minute, hour, day,months? what are your requirements.

I was using the query above because it adapted automatically to the interval which is selected in the dashboard layout, so I don’t want to hardcode it. In principle the answer to your question is simply v.windowPeriod which is set automatically when the time interval is changed to 1 hour or 1 day etc.

but you said :point_up_2:

You cant have your cake and eat it too :wink:

I said it was working before upgrading (the lines were connected) and now they are not :wink:

But my query looked very different since I used SQL, still, there was some aggregation so that I could display really long time intervals without reaching the maximum points to be displayed.

Are you saying that there is no way to have dynamic time windows and connected lines? I really need different line styles but if they are shown as points, the line style obviously has no effect.

ok I tried it locally and weird enough, I see your issue. what happens if you added this to the end?

|> yield(name: "mean")

but there seems to definitely be an issue as the time span reduces. bug! :bug:

No difference :confused:

and could you add this option to your aggregateWindow function ?
, createEmpty: false

And you can check this link too (there is work around to get fully connected graph) :

1 Like

Ah thanks, createEmpty: false did the trick. Do I need to yield the mean? I don’t see any effect other than the last datapoint vanishing :wink:

1 Like

no in my understanding, yield is usefull for query with multiple output,

but added in simple query as a good practice

1 Like