Use logical CPU count for num processors

Author: dmitryax

.chloggen/signalfx-cpu-num-processors-logical-count.yaml

@@ -0,0 +1,8 @@
+change_type: breaking
+component: exporter/signalfx
+note: "Translate `cpu.num_processors` from `system.cpu.logical.count`."
+issues: [49296]
+subtext: |
+  The exporter no longer derives `cpu.num_processors` by counting per-core `system.cpu.time` series.
+  Enable `system.cpu.logical.count` in the host_metrics receiver CPU scraper to emit `cpu.num_processors` with the default SignalFx translations.
+change_logs: [user]

exporter/signalfxexporter/README.md

@@ -112,7 +112,7 @@ The following configuration options can also be configured:
   filesystem is not `/`. Example: if the root filesystem is mounted under `/hostfs`, set
   `root_path` to `/hostfs`.
   Note: all components using `root_path` must have the same value; this currently applies
-  to hostmetrics receiver and signalfx exporter.
+  to host_metrics receiver and signalfx exporter.
 - `exclude_properties`: A list of property filters to limit dimension update content.
   Property filters can contain any number of the following fields, supporting (negated)
   string literals, re2 `/regex/`, and [glob](https://github.com/gobwas/glob) syntax values:
@@ -208,7 +208,7 @@ exporters:
 
 The default translation rules defined in [`translation/default_translation_rules.yaml`](./internal/translation/default_translation_rules.yaml) are used by the SignalFx exporter
 to help ensure compatibility with custom charts and dashboards when using the OpenTelemetry Collector.
-The default rules will create the following aggregated metrics from the [`hostmetrics` receiver](https://github.com/open-telemetry/opentelemetry-collector-contrib/blob/main/receiver/hostmetricsreceiver/README.md):
+The default rules will create the following aggregated metrics from the [`host_metrics` receiver](https://github.com/open-telemetry/opentelemetry-collector-contrib/blob/main/receiver/hostmetricsreceiver/README.md):
 
 * cpu.idle
 * cpu.interrupt
@@ -237,6 +237,9 @@ The default rules will create the following aggregated metrics from the [`hostme
 * vmpage_io.swap.in
 * vmpage_io.swap.out
 
+`cpu.num_processors` is renamed from `system.cpu.logical.count`, which must be enabled in the host_metrics receiver
+CPU scraper.
+
 ## Example Config
 
 ```yaml

exporter/signalfxexporter/exporter_test.go

@@ -1647,6 +1647,12 @@ func TestDefaultSystemCPUTimeExcludedAndTranslated(t *testing.T) {
 			dp.Attributes().PutStr("state", state)
 		}
 	}
+	cpuCount := sm.Metrics().AppendEmpty()
+	cpuCount.SetName("system.cpu.logical.count")
+	cpuCountSum := cpuCount.SetEmptySum()
+	cpuCountSum.SetAggregationTemporality(pmetric.AggregationTemporalityCumulative)
+	cpuCountSum.DataPoints().AppendEmpty().SetIntValue(32)
+
 	dps := converter.MetricsToSignalFxV2(md)
 	found := map[string]int64{}
 	for _, dp := range dps {

exporter/signalfxexporter/factory_test.go

@@ -533,6 +533,19 @@ func testGetTranslator(t *testing.T) *translation.MetricTranslator {
 	return tr
 }
 
+func appendLogicalCPUCountDataPoint(dps []*sfxpb.DataPoint, value int64) []*sfxpb.DataPoint {
+	metricType := sfxpb.MetricType_GAUGE
+	timestamp := int64(1597439222299)
+	return append(dps, &sfxpb.DataPoint{
+		Metric:     "system.cpu.logical.count",
+		Timestamp:  timestamp,
+		MetricType: &metricType,
+		Value: sfxpb.Datum{
+			IntValue: &value,
+		},
+	})
+}
+
 func TestDefaultCPUTranslations(t *testing.T) {
 	var pts1 []*sfxpb.DataPoint
 	err := testReadJSON("testdata/json/system.cpu.time.1.json", &pts1)
@@ -541,6 +554,7 @@ func TestDefaultCPUTranslations(t *testing.T) {
 	var pts2 []*sfxpb.DataPoint
 	err = testReadJSON("testdata/json/system.cpu.time.2.json", &pts2)
 	require.NoError(t, err)
+	pts2 = appendLogicalCPUCountDataPoint(pts2, 8)
 
 	tr := testGetTranslator(t)
 	log := zap.NewNop()
@@ -566,6 +580,7 @@ func TestDefaultCPUTranslations(t *testing.T) {
 
 	cpuNumProcessors := m["cpu.num_processors"]
 	require.Len(t, cpuNumProcessors, 1)
+	require.Equal(t, 8, int(*cpuNumProcessors[0].Value.IntValue))
 
 	cpuIdle := m["cpu.idle"]
 	require.Len(t, cpuIdle, 1)

exporter/signalfxexporter/internal/translation/default_translation_rules.yaml

@@ -88,15 +88,10 @@ translation_rules:
     sf_temp.cpu.softirq: int
     sf_temp.cpu.nice: int
 
-# compute cpu.num_processors
-- action: copy_metrics
+# convert cpu.num_processors metric
+- action: rename_metrics
   mapping:
-    sf_temp.cpu.idle: sf_temp.cpu.num_processors
-- action: aggregate_metric
-  metric_name: sf_temp.cpu.num_processors
-  aggregation_method: count
-  without_dimensions:
-  - cpu
+    system.cpu.logical.count: cpu.num_processors
 
 - action: aggregate_metric
   metric_name: sf_temp.cpu.idle
@@ -391,7 +386,6 @@ translation_rules:
     sf_temp.cpu.idle: cpu.idle
     sf_temp.cpu.interrupt: cpu.interrupt
     sf_temp.cpu.nice: cpu.nice
-    sf_temp.cpu.num_processors: cpu.num_processors
     sf_temp.cpu.softirq: cpu.softirq
     sf_temp.cpu.steal: cpu.steal
     sf_temp.cpu.system: cpu.system

exporter/signalfxexporter/testdata/hostmetrics_system_cpu_time_1.yaml

@@ -90,6 +90,16 @@ resourceMetrics:
                   timeUnixNano: "1136214245000000000"
               isMonotonic: true
             unit: s
+          - description: Number of available logical CPUs.
+            name: system.cpu.logical.count
+            sum:
+              aggregationTemporality: 2
+              dataPoints:
+                - asInt: "2"
+                  startTimeUnixNano: "1662132756000000000"
+                  timeUnixNano: "1136214245000000000"
+              isMonotonic: false
+            unit: "{cpu}"
         scope:
           name: github.com/open-telemetry/opentelemetry-collector-contrib/receiver/hostmetricsreceiver/internal/scraper/cpuscraper
           version: latest

exporter/signalfxexporter/testdata/hostmetrics_system_cpu_time_2.yaml

@@ -90,6 +90,16 @@ resourceMetrics:
                   timeUnixNano: "1136214250000000000"
               isMonotonic: true
             unit: s
+          - description: Number of available logical CPUs.
+            name: system.cpu.logical.count
+            sum:
+              aggregationTemporality: 2
+              dataPoints:
+                - asInt: "2"
+                  startTimeUnixNano: "1662132756000000000"
+                  timeUnixNano: "1136214250000000000"
+              isMonotonic: false
+            unit: "{cpu}"
         scope:
           name: github.com/open-telemetry/opentelemetry-collector-contrib/receiver/hostmetricsreceiver/internal/scraper/cpuscraper
           version: latest