[ML] DFA explain API throws error when called called with data stream containing counter fields

[jgowdyelastic]

The DFA UI calls the endpoint _ml/data_frame/analytics/_explain for all analysis types. For outlier detection, no fields are supplied and the payload looks something like this:

{
  "description": "",
  "source": {
    "index": "metrics-weather_sensors-dev",
    "query": {
      "match_all": {}
    }
  },
  "analysis": {
    "outlier_detection": {
      "compute_feature_influence": "true",
      "standardization_enabled": "true"
    }
  },
  "max_num_threads": 1
}

If the index is a data stream which contains a counter field, an error is returned:

[memory_usage_estimation_137565] Unable to estimate memory usage as no documents in the source indices [metrics-weather_sensors-dev] contained all the fields selected for analysis. If you are relying on automatic field selection then there are currently mapped fields that do not exist in any indexed documents, and you will have to switch to explicit field selection and include only fields that exist in indexed documents.]: [memory_usage_estimation_137565] Unable to estimate memory usage as no documents in the source indices [metrics-weather_sensors-dev] contained all the fields selected for analysis. If you are relying on automatic field selection then there are currently mapped fields that do not exist in any indexed documents, and you will have to switch to explicit field selection and include only fields that exist in indexed documents.

---

This original description is wrong. The data being used when I originally noticed this had a counter field in the mappings, but it was not populated in the data.
When the counter field is populated, the error above does not happen. I don't know if it means this is still a problem, I would assume unpopulated counter fields should not cause the endpoint to throw an error?

In addition to this, currently the explain endpoint returns is_included: true for counter fields, is this correct?

[elasticsearchmachine]

Pinging @elastic/ml-core (Team:ML)