Value mapping with derived labels

There are these types of derived labels:

• A mapping derived label creates a new label name and pulls values from some other label on the time series. Value mapping applies to mapping derived labels.
• A constructed derived label creates a label name and values where one didn't previously exist. By definition, Chronosphere already creates values for a constructed derived label, so value mapping doesn't apply to constructed derived labels.

The following examples start with this set of time series:

{__name__="grpc_metric", service="labrador", instance="blah"}
{__name__="grpc_metric", service="golden", instance="blah"}
{__name__="envoy_metric", microservice="Labrador", instance="blah"}
{__name__="envoy_metric", microservice="GOLDEN", instance="blah"}

Standardize label names

Different teams or services provide source data, some of which might not conform to the intended labeling standard. For cleaner reviews, you can map all of the provided into a standardized space. Nothing can be added to the target space without explicit mapping, which preserves and augments the original labeling.

For example:

Top-level value mappings

Top-level value mappings are mappings that apply to all name mappings, and not to a specific name mapping. This MappingLabel includes top-level value mappings in the value_mappings section.

label_name: standard_service
mapping_label:
  name_mappings:
     - filters:
        - name: __name__
          value_glob: grpc_*
       source_label: service
     - filters:
        - name: __name__
          value_glob: envoy_*
       source_label: microservice
  value_mappings:
    - source_value_globs:
       - labrador
       - Labrador
      target_value: labrador-retriever
    - source_value_globs:
       - golden
       - GOLDEN
      target_value: golden-retriever

If you were to use the query grpc_metric{standard_service="labrador-retriever"}, the resulting time series is:

{__name__="grpc_metric", service="labrador", standard_service="labrador-retriever", instance="blah"}

For another query envoy_metric{standard_service="labrador-retriever"}, the resultingtime series is:

{__name__="envoy_metric", microservice="Labrador", standard_service="labrador-retriever", instance="blah"}

If you query using an non-existent target value such as grpc_metric{standard_service="labrador"}, it returns this time series:

{__name__="grpc_metric", service="labrador", standard_service="labrador-retriever", instance="blah"}

Value mappings inside a name mapping

Value mappings match to one specific name mapping.

For example:

label_name: standard_service
  mapping_label:
    name_mappings:
       - filters:
          - name: __name__
            value_glob: grpc_*
         source_label: service
         value_mappings:
          - source_value_globs:
             - labrador
             - Labrador
            target_value: labrador-retriever
       - filters:
          - name: __name__
            value_glob: envoy_*
          source_label: microservice

If we query grpc_metric{standard_service="labrador-retriever"}, with this configuration, the target value labrador-retriever applies only to metrics that match __name__:grpc_*.

The resulting time series is:

{__name__="grpc_metric", service="labrador", standard_service="labrador-retriever", instance="blah"}

Consider the query envoy_metric{standard_service="labrador-retriever"}. According to the configuration, the value mapping for labrador-retriever applies only to metrics that match __name__:grpc_*.

This query returns no results because:

• standard_service="labrador-retriever" isn't a physical label-value pair on envoy_metric.
• There is no top-level value mapping where labrador-retriever is the target value. There's also no name mapping specific value mapping where labrador-retriever is the target value for metrics that match __name__:envoy_*.

Use value mapping inside name mapping with top level mapping together

You can use Value Mappings and name mapping level value mappings in the same configuration. For example:

label_name: standard_service
  mapping_label:
    name_mappings:
       - filters:
          - name: __name__
            value_glob: grpc_*
         source_label: service
         value_mappings:
          - source_value_glob:
             - labrador
             - Labrador
            target_value: labrador-retriever
       - filters:
          - name: __name__
            value_glob: envoy_*
          source_label: microservice
    value_mappings:
    - source_value_globs:
       - labrador
       - Labrador
      target_value: golden-retriever

In this example, target value labrador-retriever applies specifically to metrics that matches __name__:grpc_* because it matches to that specific name mapping, where value labrador-retriever applies to all name mappings.

For the query grpc_metric{standard_service="labrador-retriever"} results in the time series:

{__name__="grpc_service", service="labrador", standard_service="labrador-retriever", instance="blah"}

Similarly, for the query envoy_metric{standard_service="golden-retriever"}, the resulting time series is:

{__name__="envoy_metric", microservice="GOLDEN", standard_service="golden-retriever", instance="blah"}

However, querying envoy_metric{standard_service="labrador-retriever"} returns no values, because:

• standard_service="labrador-retriever" isn't a physical label-value pair on envoy_metric.
• There is no top-level value mapping where labrador-retriever is the target value. There's also no name mapping that has a value mapping where labrador-retriever is the target value for metrics that match __name__:envoy_*.

If there is a top-level value mapping with the same target value as a name mapping specific value mapping, like this situation:

label_name: standard_service
  mapping_label:
    name_mappings:
       - filters:
          - name: __name__
            value_glob: grpc_*
         source_label: service
         value_mappings:
          - source_value_filters:
             - labrador
             - Labrador
            target_value: golden-retriever
       - filters:
          - name: __name__
            value_glob: envoy_*
          source_label: microservice
    value_mappings:
     - source_value_globs:
        - golden
        - GOLDEN
       target_value: golden-retriever

In this case, both value mappings have the same target value of golden-retriever. The value mappings that are name mapping specific apply over the top-level value mappings.