Downsample system metrics with "strings"

t481 · June 29, 2021, 5:38pm

Hello,
I’m looking to create a task to downsample collected system metrics into another bucket.

Below is the original CQ code

CREATE CONTINUOUS QUERY cq_system_metrics ON system_metrics_daily BEGIN SELECT mean() AS value INTO system_metrics.two_weeks.:MEASUREMENT FROM system_metrics_daily.two_hours././ GROUP BY time(1h) END

Since upgrading to v2.0 I have not been able to downsample the data. Here is the query I’m running

from(bucket: "system_metrics/two_hours")
	|> range(start: -task.every)
	|> aggregateWindow(every: 1h, fn: mean)
	|> to(bucket: "system_metrics/two_weeks", org: "myorg", timeColumn: "_stop")

This results in an error:

input type for mean aggregate: string

I suspect some of the metrics are “string” such as “Up” etc.

How can I overcome this?

russ_savage · June 29, 2021, 7:47pm

Hi t481, the way to solve this for now is to separate the numerical and string data so that you can apply the correct aggregates. Here’s an example:

github.com

influxdata/community-templates/blob/master/downsampling/default_inputs/downsampling_tasks.yml#L16


      
          every: 24h
          name: 04. downsample_default_1d
          query: |-
              fromBucket = "telegraf"
              toBucket = "telegraf_" + string(v: task.every)
              
              all_data = from(bucket: fromBucket)
                  |> range(start: -task.every)
                  |> filter(fn: (r) =>
                      (r._measurement =~ /^cpu|disk|diskio|mem|processes|swap|system$/))
              numeric_data = all_data
                  |> filter(fn: (r) =>
                      (r._field !~ /^uptime_format$/))
          
          
    numeric_data
                  |> aggregateWindow(every: task.every, fn: mean)
                  |> set(key: "aggregate", value: "mean")
                  |> set(key: "rollup_interval", value: string(v: task.every))
                  |> to(bucket: toBucket)
              numeric_data
                  |> aggregateWindow(every: task.every, fn: min)

Topic		Replies	Views
Downsampling data into another bucket Tasks cq , tasks , downsample	5	1964	June 4, 2021
Manual page CQ not correct mean on string	0	691	December 29, 2018
Downsampling Task fail with error (unsupported input type for mean aggregate) Tasks influxdb , downsample	14	3082	April 7, 2022
Downsample and Downsampling a Downsample	2	978	April 5, 2017
Creating a Downsampling Task with conditions Tasks	4	208	February 16, 2024

Downsample system metrics with "strings"

Related topics