[processor/transform] Aggregate metrics grouping by label

[gabrielgiussi]

Component(s)

processor/transform

Is your feature request related to a problem? Please describe.

The metricstransform already provides this but there is no way to do it in the transform processor.
e.g.

- include: process.runtime.jvm.memory.usage
        action: insert
        new_name: jvm_memory_bytes_used
        operations:
          - action: aggregate_labels
            label_set: [ type ]
            aggregation_type: sum

With this we can go from

process_runtime_jvm_memory_usage{job="unknown_service:java",pool="CodeHeap 'non-nmethods'",type="non_heap"} 1.382272e+06
process_runtime_jvm_memory_usage{job="unknown_service:java",pool="CodeHeap 'non-profiled nmethods'",type="non_heap"} 9.991936e+06
process_runtime_jvm_memory_usage{job="unknown_service:java",pool="CodeHeap 'profiled nmethods'",type="non_heap"} 2.968128e+07
process_runtime_jvm_memory_usage{job="unknown_service:java",pool="Compressed Class Space",type="non_heap"} 2.2830672e+07
process_runtime_jvm_memory_usage{job="unknown_service:java",pool="Eden Space",type="heap"} 1.5339216e+07
process_runtime_jvm_memory_usage{job="unknown_service:java",pool="Metaspace",type="non_heap"} 1.40415936e+08
process_runtime_jvm_memory_usage{job="unknown_service:java",pool="Survivor Space",type="heap"} 330760
process_runtime_jvm_memory_usage{job="unknown_service:java",pool="Tenured Gen",type="heap"} 6.3913088e+07

to this

jvm_memory_bytes_used{type="heap",job="unknown_service:java"} 7.9583064e+07
jvm_memory_bytes_used{type="non_heap",job="unknown_service:java"} 2.04302096e+08

Describe the solution you'd like

A function to aggregate metric values grouping by a set of attributes

- aggregate([attributes["type"]],sum) where metric.name == "process_runtime_jvm_memory_usage"

In the example I'm also renaming the metric but that could be done with a different function in a next step using set(metric.name,"jvm_memory_bytes_used")

(Providing this just as an example, I don't know which could be the best syntax for this)

Describe alternatives you've considered

No response

Additional context

No response

[github-actions]

Pinging code owners:

• processor/transform: @TylerHelmuth @kentquirk @bogdandrutu @evan-bradley

See Adding Labels via Comments if you do not have permissions to add labels yourself.

[github-actions]

This issue has been inactive for 60 days. It will be closed in 60 days if there is no activity. To ping code owners by adding a component label, see Adding Labels via Comments, or if you are unsure of which component this issue relates to, please ping @open-telemetry/collector-contrib-triagers. If this issue is still relevant, please ping the code owners or leave a comment explaining why it is still relevant. Otherwise, please close it.

Pinging code owners:

• processor/transform: @TylerHelmuth @kentquirk @bogdandrutu @evan-bradley

See Adding Labels via Comments if you do not have permissions to add labels yourself.

[seankhliao]

If this were implemented, I think it should just be aggregate("min/max/sum/mean") where name == "metric_name" (This would be run on the metric context.)
Common uses of aggregation would include both with an explicit label list (sum by (labels) (metric) and with a label exclusion list (sum without (labels) (metric).
It would be simpler to leave attribute/label removal to other statements, and have aggregate just merge together matching data points
This would be run on the metric context