[processor/metricsgeneration] Metrics should be calculated when second metric value is 0

[crobert-1]

Component(s)

processor/metricsgeneration

Describe the issue you're reporting

Bug
The metrics generation processor is used to create new metrics by using calculations on existing metrics. As explained in the configuration example here, the config specifies the first metric, second metric, and operation type for calculations.

Existing logic for this processor simply skips the calculation when the value of the second metric's datapoint is 0 (reference). This means that even when it's a valid operation, and the result may be relevant for users, no metric is created.

Example showing bug
Relevant example configuration:

experimental_metricsgeneration:
  rules:
    - name: available_capacity
      unit: bytes
      type: calculate
      metric1: total_capacity
      metric2: used_capacity
      operation: subtract

This configuration should create a new metric called available_capacity, which will be the result of total_capacity - used_capacity. If the value of used_capacity is 0, the processor won't create a new metric called available_capacity, when I believe it should.

Proposed Solution
I think the solution here is to allow the calculation to be done and to create the resulting metric.

Open Question
How should we handle the divide and percent operations for the case of dividing by 0? We can either add a data point with a value of 0 or not add any data point at all. Since dividing by 0 is an invalid operation, I believe the right thing to do would be log an error and not add a data point in that case. I'm open to feedback from others though.

[github-actions]

Pinging code owners:

• processor/metricsgeneration: @Aneurysm9

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

[dmitryax]

Since dividing by 0 is an invalid operation, I believe the right thing to do would be log an error and not add a data point in that case.

I agree

[crobert-1]

Had a discussion with Dmitrii offline, we also agreed that if no valid data points are generated for the new metric, the new metric should not be created. I'll add a note in the README to clarify this for users.