feat(host_metrics source): add temperature metrics collector

Author: somaz94

changelog.d/21389_host_metrics_temperature.feature.md

@@ -0,0 +1,11 @@
+The `host_metrics` source can now collect hardware temperature readings via a
+new `temperature` collector. When enabled, it emits `temperature_celsius`,
+`temperature_max_celsius`, and `temperature_critical_celsius` gauges, each
+tagged with the `component` label of the sensor it was read from.
+
+The collector is opt-in: add `temperature` to the `collectors` list to enable
+it. Components that do not report a given value (for example a missing critical
+threshold) are skipped, and environments without temperature sensors simply
+produce no metrics.
+
+authors: somaz94

src/sources/host_metrics/mod.rs

@@ -40,6 +40,7 @@ mod network;
 mod process;
 #[cfg(target_os = "linux")]
 mod tcp;
+mod temperature;
 
 /// Collector types.
 #[serde_as]
@@ -78,6 +79,9 @@ pub enum Collector {
 
     /// Metrics related to TCP connections.
     TCP,
+
+    /// Metrics related to component temperatures.
+    Temperature,
 }
 
 /// Filtering configuration.
@@ -186,7 +190,7 @@ pub fn default_namespace() -> Option<String> {
     Some(String::from("host"))
 }
 
-const fn example_collectors() -> [&'static str; 9] {
+const fn example_collectors() -> [&'static str; 10] {
     [
         "cgroups",
         "cpu",
@@ -197,6 +201,7 @@ const fn example_collectors() -> [&'static str; 9] {
         "memory",
         "network",
         "tcp",
+        "temperature",
     ]
 }
 
@@ -420,6 +425,9 @@ impl HostMetrics {
         if self.config.has_collector(Collector::TCP) {
             self.tcp_metrics(&mut buffer).await;
         }
+        if self.config.has_collector(Collector::Temperature) {
+            self.temperature_metrics(&mut buffer).await;
+        }
 
         let metrics = buffer.metrics;
         self.events_received.emit(CountByteSize(

src/sources/host_metrics/temperature.rs

@@ -0,0 +1,66 @@
+use sysinfo::Components;
+use vector_lib::metric_tags;
+
+use super::HostMetrics;
+
+const COMPONENT: &str = "component";
+const TEMPERATURE_CELSIUS: &str = "temperature_celsius";
+const TEMPERATURE_MAX_CELSIUS: &str = "temperature_max_celsius";
+const TEMPERATURE_CRITICAL_CELSIUS: &str = "temperature_critical_celsius";
+
+impl HostMetrics {
+    pub async fn temperature_metrics(&self, output: &mut super::MetricsBuffer) {
+        output.name = "temperature";
+        let components = Components::new_with_refreshed_list();
+        for component in &components {
+            let label = component.label();
+            let tags = || metric_tags!(COMPONENT => label);
+            if let Some(temperature) = component.temperature() {
+                output.gauge(TEMPERATURE_CELSIUS, temperature as f64, tags());
+            }
+            if let Some(max) = component.max() {
+                output.gauge(TEMPERATURE_MAX_CELSIUS, max as f64, tags());
+            }
+            if let Some(critical) = component.critical() {
+                output.gauge(TEMPERATURE_CRITICAL_CELSIUS, critical as f64, tags());
+            }
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::{
+        super::{HostMetrics, HostMetricsConfig, MetricsBuffer, tests::all_gauges},
+        COMPONENT,
+    };
+
+    #[tokio::test]
+    async fn generates_temperature_metrics() {
+        let mut buffer = MetricsBuffer::new(None);
+        HostMetrics::new(HostMetricsConfig::default())
+            .temperature_metrics(&mut buffer)
+            .await;
+        let metrics = buffer.metrics;
+
+        // Temperature sensors are not exposed in many environments (containers,
+        // virtual machines, CI runners), so the component list can legitimately
+        // be empty. When metrics are produced, they must all be gauges named
+        // `temperature*` and carry the `component` tag.
+        assert!(all_gauges(&metrics));
+        for metric in &metrics {
+            assert!(
+                metric.name().starts_with("temperature"),
+                "unexpected metric name: {}",
+                metric.name()
+            );
+            assert!(
+                metric
+                    .tags()
+                    .expect("temperature metric is missing tags")
+                    .contains_key(COMPONENT),
+                "temperature metric is missing the `component` tag"
+            );
+        }
+    }
+}

website/cue/reference/components/sources/generated/host_metrics.cue

@@ -80,17 +80,18 @@ generated: components: sources: host_metrics: configuration: {
 
 						Only available on Linux.
 						"""
-					cpu:        "Metrics related to CPU utilization."
-					disk:       "Metrics related to disk I/O utilization."
-					filesystem: "Metrics related to filesystem space utilization."
-					host:       "Metrics related to the host."
-					load:       "Metrics related to the system load average."
-					memory:     "Metrics related to memory utilization."
-					network:    "Metrics related to network utilization."
-					process:    "Metrics related to Process utilization."
-					tcp:        "Metrics related to TCP connections."
+					cpu:         "Metrics related to CPU utilization."
+					disk:        "Metrics related to disk I/O utilization."
+					filesystem:  "Metrics related to filesystem space utilization."
+					host:        "Metrics related to the host."
+					load:        "Metrics related to the system load average."
+					memory:      "Metrics related to memory utilization."
+					network:     "Metrics related to network utilization."
+					process:     "Metrics related to Process utilization."
+					tcp:         "Metrics related to TCP connections."
+					temperature: "Metrics related to component temperatures."
 				}
-				examples: ["cgroups", "cpu", "disk", "filesystem", "load", "host", "memory", "network", "tcp"]
+				examples: ["cgroups", "cpu", "disk", "filesystem", "load", "host", "memory", "network", "tcp", "temperature"]
 			}
 		}
 	}

website/cue/reference/components/sources/host_metrics.cue

@@ -193,6 +193,11 @@ components: sources: host_metrics: {
 			}
 		}
 
+		// Host temperature
+		temperature_celsius: _host & _temperature_gauge & {description: "The current temperature reported by a hardware component, in degrees Celsius."}
+		temperature_max_celsius: _host & _temperature_gauge & {description: "The highest temperature recorded for a hardware component, in degrees Celsius."}
+		temperature_critical_celsius: _host & _temperature_gauge & {description: "The temperature at which a hardware component is considered critical, in degrees Celsius."}
+
 		// Helpers
 		_host: {
 			default_namespace: "host"
@@ -307,5 +312,17 @@ components: sources: host_metrics: {
 				}
 			}
 		}
+
+		_temperature_gauge: {
+			type: "gauge"
+			tags: _host_metrics_tags & {
+				collector: examples: ["temperature"]
+				component: {
+					description: "The label of the hardware component the temperature was read from."
+					required:    true
+					examples: ["Core 0", "coretemp Package id 0", "nvme Composite"]
+				}
+			}
+		}
 	}
 }