feat(new transform): Add new incremental_to_absolute transform and change to MetricSet to an LRU cache with optional capacity policies

Cargo.toml

@@ -197,7 +197,7 @@ dnstap-parser = { path = "lib/dnstap-parser", optional = true }
 fakedata = { path = "lib/fakedata", optional = true }
 portpicker = { path = "lib/portpicker" }
 tracing-limit = { path = "lib/tracing-limit" }
-vector-common = { path = "lib/vector-common", default-features = false}
+vector-common = { path = "lib/vector-common", default-features = false }
 vector-lib.workspace = true
 vector-config.workspace = true
 vector-config-common.workspace = true
@@ -384,7 +384,7 @@ seahash = { version = "4.1.0", default-features = false }
 smallvec = { version = "1", default-features = false, features = ["union", "serde"] }
 snap = { version = "1.1.1", default-features = false }
 socket2 = { version = "0.5.10", default-features = false }
-sqlx = { version = "0.8.6", default-features = false, features = ["derive", "postgres", "chrono", "runtime-tokio"], optional=true }
+sqlx = { version = "0.8.6", default-features = false, features = ["derive", "postgres", "chrono", "runtime-tokio"], optional = true }
 stream-cancel = { version = "0.8.2", default-features = false }
 strip-ansi-escapes = { version = "0.2.1", default-features = false }
 syslog = { version = "6.1.1", default-features = false, optional = true }
@@ -691,6 +691,7 @@ transforms-logs = [
 transforms-metrics = [
   "transforms-aggregate",
   "transforms-filter",
+  "transforms-incremental_to_absolute",
   "transforms-log_to_metric",
   "transforms-lua",
   "transforms-metric_to_log",
@@ -703,6 +704,7 @@ transforms-aggregate = []
 transforms-aws_ec2_metadata = ["dep:arc-swap"]
 transforms-dedupe = ["transforms-impl-dedupe"]
 transforms-filter = []
+transforms-incremental_to_absolute = []
 transforms-window = []
 transforms-log_to_metric = []
 transforms-lua = ["dep:mlua", "vector-lib/lua"]

changelog.d/incremental_to_absolute_transform.feature.md

@@ -0,0 +1,6 @@
+Add a new `incremental_to_absolute` transform which converts incremental metrics to absolute metrics. This is useful for
+use cases when sending metrics to a sink is lossy or you want to get a historical record of metrics, in which case
+incremental metrics may be inaccurate since any gaps in metrics sent will result in an inaccurate reading of the ending
+value.
+
+authors: GreyLilac09

lib/vector-core/src/event/metric/data.rs

@@ -58,11 +58,14 @@ impl MetricData {
 
     /// Consumes this metric, returning it as an absolute metric.
     ///
-    /// If the metric was already absolute, nothing is changed.
+    /// The interval_ms is removed. If the metric was already absolute, nothing is changed.
     #[must_use]
     pub fn into_absolute(self) -> Self {
         Self {
-            time: self.time,
+            time: MetricTime {
+                timestamp: self.time.timestamp,
+                interval_ms: None,
+            },
             kind: MetricKind::Absolute,
             value: self.value,
         }

src/transforms/incremental_to_absolute.rs

@@ -0,0 +1,219 @@
+use std::{collections::HashMap, future::ready, pin::Pin, time::Duration};
+
+use futures::{Stream, StreamExt};
+use vector_lib::config::LogNamespace;
+use vector_lib::configurable::configurable_component;
+
+use crate::sinks::util::buffer::metrics::{MetricSet, TtlPolicy};
+use crate::{
+    config::{DataType, Input, OutputId, TransformConfig, TransformContext, TransformOutput},
+    event::Event,
+    schema,
+    transforms::{TaskTransform, Transform},
+};
+
+/// Configuration for the `incremental_to_absolute` transform.
+#[configurable_component(transform(
+    "incremental_to_absolute",
+    "Convert incremental metrics to absolute."
+))]
+#[derive(Clone, Debug, Default)]
+#[serde(deny_unknown_fields)]
+pub struct IncrementalToAbsoluteConfig {
+    /// The amount of time, in seconds, that incremental metrics will persist in the internal
+    /// metrics cache after having not been updated before they expire and are removed.
+    /// Once removed, incremental counters are reset to 0.
+    #[serde(skip_serializing_if = "crate::serde::is_default")]
+    #[configurable(metadata(docs::common = false, docs::required = true, docs::examples = "120"))]
+    pub expire_metrics_secs: u64,
+}
+
+impl_generate_config_from_default!(IncrementalToAbsoluteConfig);
+
+#[async_trait::async_trait]
+#[typetag::serde(name = "incremental_to_absolute")]
+impl TransformConfig for IncrementalToAbsoluteConfig {
+    async fn build(&self, _context: &TransformContext) -> crate::Result<Transform> {
+        IncrementalToAbsolute::new(self).map(Transform::event_task)
+    }
+
+    fn input(&self) -> Input {
+        Input::metric()
+    }
+
+    fn outputs(
+        &self,
+        _: vector_lib::enrichment::TableRegistry,
+        _: &[(OutputId, schema::Definition)],
+        _: LogNamespace,
+    ) -> Vec<TransformOutput> {
+        vec![TransformOutput::new(DataType::Metric, HashMap::new())]
+    }
+}
+#[derive(Debug)]
+pub struct IncrementalToAbsolute {
+    data: MetricSet,
+}
+
+impl IncrementalToAbsolute {
+    pub fn new(config: &IncrementalToAbsoluteConfig) -> crate::Result<Self> {
+        Ok(Self {
+            data: MetricSet::with_ttl_policy(TtlPolicy::new(Duration::from_secs(
+                config.expire_metrics_secs,
+            ))),
+        })
+    }
+    pub fn transform_one(&mut self, event: Event) -> Option<Event> {
+        if let Some(metric) = self.data.make_absolute(event.as_metric().clone()) {
+            // Create a new Event from the metric
+            Some(Event::Metric(metric))
+        } else {
+            None
+        }
+    }
+}
+
+impl TaskTransform<Event> for IncrementalToAbsolute {
+    fn transform(
+        self: Box<Self>,
+        task: Pin<Box<dyn Stream<Item = Event> + Send>>,
+    ) -> Pin<Box<dyn Stream<Item = Event> + Send>>
+    where
+        Self: 'static,
+    {
+        let mut inner = self;
+        Box::pin(task.filter_map(move |v| ready(inner.transform_one(v))))
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use futures_util::SinkExt;
+    use similar_asserts::assert_eq;
+    use std::sync::Arc;
+    use vector_lib::config::ComponentKey;
+
+    use super::*;
+    use crate::event::{
+        metric::{MetricKind, MetricValue},
+        Metric,
+    };
+
+    fn make_metric(name: &'static str, kind: MetricKind, value: MetricValue) -> Event {
+        let mut event = Event::Metric(Metric::new(name, kind, value))
+            .with_source_id(Arc::new(ComponentKey::from("in")))
+            .with_upstream_id(Arc::new(OutputId::from("transform")));
+
+        event.metadata_mut().set_source_type("unit_test_stream");
+
+        event
+    }
+
+    async fn assert_metric_eq(
+        tx: &mut futures::channel::mpsc::Sender<Event>,
+        mut out_stream: impl Stream<Item = Event> + Unpin,
+        metric: Event,
+        expected_metric: Event,
+    ) {
+        tx.send(metric).await.unwrap();
+        if let Some(out_event) = out_stream.next().await {
+            let result = out_event;
+            assert_eq!(result, expected_metric);
+        } else {
+            panic!("Unexpectedly received None in output stream");
+        }
+    }
+
+    #[tokio::test]
+    async fn test_incremental_to_absolute() {
+        let config = toml::from_str::<IncrementalToAbsoluteConfig>(
+            r#"
+expire_metrics_secs = 9999
+"#,
+        )
+        .unwrap();
+        let incremental_to_absolute = IncrementalToAbsolute::new(&config)
+            .map(Transform::event_task)
+            .unwrap();
+        let incremental_to_absolute = incremental_to_absolute.into_task();
+        let (mut tx, rx) = futures::channel::mpsc::channel(10);
+        let mut out_stream = incremental_to_absolute.transform_events(Box::pin(rx));
+
+        let inc_counter_1 = make_metric(
+            "incremental_counter",
+            MetricKind::Incremental,
+            MetricValue::Counter { value: 10.0 },
+        );
+        let expected_inc_counter_1 = make_metric(
+            "incremental_counter",
+            MetricKind::Absolute,
+            MetricValue::Counter { value: 10.0 },
+        );
+        assert_metric_eq(
+            &mut tx,
+            &mut out_stream,
+            inc_counter_1,
+            expected_inc_counter_1,
+        )
+        .await;
+
+        let inc_counter_2 = make_metric(
+            "incremental_counter",
+            MetricKind::Incremental,
+            MetricValue::Counter { value: 10.0 },
+        );
+        let expected_inc_counter_2 = make_metric(
+            "incremental_counter",
+            MetricKind::Absolute,
+            MetricValue::Counter { value: 20.0 },
+        );
+        assert_metric_eq(
+            &mut tx,
+            &mut out_stream,
+            inc_counter_2,
+            expected_inc_counter_2,
+        )
+        .await;
+
+        let inc_counter_3 = make_metric(
+            "incremental_counter",
+            MetricKind::Incremental,
+            MetricValue::Counter { value: 10.0 },
+        );
+        let expected_inc_counter_3 = make_metric(
+            "incremental_counter",
+            MetricKind::Absolute,
+            MetricValue::Counter { value: 30.0 },
+        );
+        assert_metric_eq(
+            &mut tx,
+            &mut out_stream,
+            inc_counter_3,
+            expected_inc_counter_3,
+        )
+        .await;
+
+        // Absolute counters and gauges are emitted unchanged
+        let gauge = make_metric(
+            "gauge",
+            MetricKind::Absolute,
+            MetricValue::Gauge { value: 42.0 },
+        );
+        let expected_gauge = gauge.clone();
+        assert_metric_eq(&mut tx, &mut out_stream, gauge, expected_gauge).await;
+
+        let absolute_counter = make_metric(
+            "absolute_counter",
+            MetricKind::Absolute,
+            MetricValue::Counter { value: 42.0 },
+        );
+        let absolute_counter_expected = absolute_counter.clone();
+        assert_metric_eq(
+            &mut tx,
+            &mut out_stream,
+            absolute_counter,
+            absolute_counter_expected,
+        )
+        .await;
+    }
+}

src/transforms/mod.rs

@@ -15,6 +15,8 @@ pub mod aws_ec2_metadata;
 mod exclusive_route;
 #[cfg(feature = "transforms-filter")]
 pub mod filter;
+#[cfg(feature = "transforms-incremental_to_absolute")]
+pub mod incremental_to_absolute;
 #[cfg(feature = "transforms-log_to_metric")]
 pub mod log_to_metric;
 #[cfg(feature = "transforms-lua")]

website/cue/reference/components.cue

@@ -192,19 +192,20 @@ components: {
 		}
 
 		if Args.kind == "transform" {
-			aggregate?:       #FeaturesAggregate
-			convert?:         #FeaturesConvert
-			enrich?:          #FeaturesEnrich
-			filter?:          #FeaturesFilter
-			parse?:           #FeaturesParse
-			program?:         #FeaturesProgram
-			proxy?:           #FeaturesProxy
-			reduce?:          #FeaturesReduce
-			route?:           #FeaturesRoute
-			exclusive_route?: #FeaturesExclusiveRoute
-			sanitize?:        #FeaturesSanitize
-			shape?:           #FeaturesShape
-			window?:          #FeaturesWindow
+			aggregate?:       				#FeaturesAggregate
+			convert?:         				#FeaturesConvert
+			enrich?:          				#FeaturesEnrich
+			filter?:          				#FeaturesFilter
+			parse?:           				#FeaturesParse
+			program?:         				#FeaturesProgram
+			proxy?:           				#FeaturesProxy
+			reduce?:          				#FeaturesReduce
+			route?:           				#FeaturesRoute
+			exclusive_route?: 				#FeaturesExclusiveRoute
+			sanitize?:        				#FeaturesSanitize
+			shape?:           				#FeaturesShape
+			window?:          				#FeaturesWindow
+			incremental_to_absolute?: #FeaturesIncrementalToAbsolute
 		}
 
 		if Args.kind == "sink" {
@@ -338,6 +339,8 @@ components: {
 
 	#FeaturesWindow: {}
 
+	#FeaturesIncrementalToAbsolute: {}
+
 	#FeaturesSend: {
 		_args: {
 			egress_method: string

website/cue/reference/components/transforms/generated/incremental_to_absolute.cue

@@ -0,0 +1,14 @@
+package metadata
+
+generated: components: transforms: incremental_to_absolute: configuration: expire_metrics_secs: {
+	common: false
+	description: """
+		The amount of time, in seconds, that incremental metrics will persist in the internal
+		metrics cache after having not been updated before they expire and are removed.
+		Once removed, incremental counters are reset to 0.
+		"""
+	required: true
+	type: uint: examples: [
+		"120",
+	]
+}