Problem with aggregation of Performance Counters within same category

[githubmui]

• Version: 6.0.0.beta1
• Beat: MetricBeat
• Operating System: Windows 7
• Module: Windows
• Steps to reproduce: Configure some performance counters from category memory

- module: windows
  metricsets: ["perfmon"]
  period: 10s
  perfmon.counters:

    - instance_label: "memory.instance"
      instance_name: ""
      measurement_label: "memory.AB"
      query: '\Memory\Available Bytes'  
      
    - instance_label: "memory.instance"
      instance_name: ""
      measurement_label: "memory.CB"
      query: '\Memory\Committed Bytes'  

    - instance_label: "memory.instance"
      instance_name: ""
      measurement_label: "memory.PS"
      query: '\Memory\Pages/sec'  

• Observed result: Every performance counter will be output as a separate JSON record.
• Expected result: All counter values with same "measurement_label"-namespace, with same instance_label and same instance_name will be aggregated to one single JSON record, like ...

{
  "@timestamp": "2017-08-18T03:38:45.395Z",
  "@metadata": {
    "beat": "metricbeat",
    "type": "doc"
  },
  "metricset": {
    "name": "perfmon",
    "rtt": 1000,
    "module": "windows"
  },
  "windows": {
    "perfmon": {
      "memory": {
        "instance": "",
        "AB": 10,
        "CB": 1.356897670,
        "PS": 9.2346e+10,      
      }
    }
  },
  "beat": {
    "name": "WINHOST",
    "hostname": "WINHOST",
    "version": "6.0.0-beta1"
  }
}

[synhershko]

Would also love to hear any insight on this. FWIW, the documentation suggests an aggregation occurs also across categories (which is the desired behavior in almost every case I've seen): https://www.elastic.co/guide/en/beats/metricbeat/current/metricbeat-metricset-windows-perfmon.html

[ruflin]

@synhershko Have a look at #6584

[andrewkroh]

And the documentation that you linked to has been changed on master. https://www.elastic.co/guide/en/beats/metricbeat/master/metricbeat-metricset-windows-perfmon.html

[andyhoffman12]

I'd absolutely love to see this. FWIW I would appreciate the ability to see an entire snapshot of the performance information inside one record though any aggregation is a win in my book. In my instance I'm using metricbeat to locate poor performing workstations. This means I'm often looking for a combination of CPU/Memory/Disk/Process stats at the same time. Looking at it from this aspect allows me to determine if the system is under powered or if a user is simply asking to much from the machine.

[andyhoffman12]

Looks like there's a pull request that will resolve this: #8688

[narph]

@githubmui, sounds like perfmon.group_measurements_by_instance might solve your scenario, will close the issue for now, feel free to reopen it if there are still questions regarding this.