[receiver/hostmetrics] The value of process.cpu.utilization may exceed 1

[BinaryFissionGames]

Component(s)

receiver/hostmetrics

What happened?

Description

The process.cpu.utilization metric is expected to be a value between 0 and 1, but it can be greater than 1.

My guess here is that process.cpu.utilization does not properly divide by the number of system cores, so the value can actually be between 0 - ${num_cores}

Steps to Reproduce

1. Run a process with heavy load on a multi-core machine
2. Scrape with the hostmetric receiver
3. Note that process metrics may exceed 1

Expected Result

No metrics exceed 1 (in fact, I'd expect the sum of all processes to be < 1)

Actual Result

I'm getting a metric of 9.5 (this process was taking ~950% cpu in my activity monitor)

{
  "resource": {
    "attributes": [
      { "key": "process.pid", "value": { "intValue": "89633" } },
      { "key": "process.parent_pid", "value": { "intValue": "89617" } },
      {
        "key": "process.executable.name",
        "value": { "stringValue": "load" }
      },
      {
        "key": "process.executable.path",
        "value": {
          "stringValue": "/var/folders/qn/13392c0n4rl4j4cplc65_66h0000gn/T/go-build2476508403/b001/exe/load"
        }
      },
      {
        "key": "process.command",
        "value": {
          "stringValue": "/var/folders/qn/13392c0n4rl4j4cplc65_66h0000gn/T/go-build2476508403/b001/exe/load"
        }
      },
      {
        "key": "process.command_line",
        "value": {
          "stringValue": "/var/folders/qn/13392c0n4rl4j4cplc65_66h0000gn/T/go-build2476508403/b001/exe/load"
        }
      },
      {
        "key": "process.owner",
        "value": { "stringValue": "brandonjohnson" }
      },
      { "key": "host.name", "value": { "stringValue": "Brandons-MBP" } },
      { "key": "os.type", "value": { "stringValue": "darwin" } }
    ]
  },
  "scopeMetrics": [
    {
      "scope": {
        "name": "otelcol/hostmetricsreceiver/process",
        "version": "v1.44.0"
      },
      "metrics": [
        {
          "name": "process.cpu.utilization",
          "description": "Percentage of total CPU time used by the process since last scrape, expressed as a value between 0 and 1. On the first scrape, no data point is emitted for this metric.",
          "unit": "1",
          "gauge": {
            "dataPoints": [
              {
                "attributes": [
                  { "key": "state", "value": { "stringValue": "user" } }
                ],
                "startTimeUnixNano": "1708529361162000000",
                "timeUnixNano": "1708529423825350000",
                "asDouble": 9.532629437655167
              },
              {
                "attributes": [
                  { "key": "state", "value": { "stringValue": "system" } }
                ],
                "startTimeUnixNano": "1708529361162000000",
                "timeUnixNano": "1708529423825350000",
                "asDouble": 0.07040537266051189
              },
              {
                "attributes": [
                  { "key": "state", "value": { "stringValue": "wait" } }
                ],
                "startTimeUnixNano": "1708529361162000000",
                "timeUnixNano": "1708529423825350000",
                "asDouble": 0
              }
            ]
          }
        }
      ]
    }
  ],
  "schemaUrl": "https://opentelemetry.io/schemas/1.9.0"
}

Collector version

v0.94.0

Environment information

Environment

OS: macOS 14.2.1
Compiler(if manually compiled): go 1.22

OpenTelemetry Collector configuration

No response

Log output

No response

Additional context

I used this quick go program to quickly generate load:

package main

import "time"

func main() {
	for i := 0; i < 100; i++ {
		go func() {
			v := 0
			for {
				v++
			}
		}()
	}

	time.Sleep(3 * time.Minute)
}

[github-actions]

Pinging code owners:

• receiver/hostmetrics: @dmitryax @braydonk

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

[crobert-1]

Removing needs triage based on comments in PR.