fix: replace metrics callback with background scraping to prevent tim… (#2480)

Co-authored-by: Keiven Chang <keivenchang@users.noreply.github.com>

deploy/metrics/prometheus.yml

@@ -42,6 +42,7 @@ scrape_configs:
       - targets: ['host.docker.internal:8080']  # on the "monitoring" network
 
   # Launch via: DYN_SYSTEM_ENABLED=true DYN_SYSTEM_PORT=8081 dynamo.<backend> ...
+  # If you want to update the scrape_interval, you may want to also update component.rs's MAX_DELAY
   - job_name: 'dynamo-backend'
     scrape_interval: 6s
     static_configs:

lib/bindings/python/Cargo.lock

@@ -1134,7 +1134,7 @@ dependencies = [
 
 [[package]]
 name = "dynamo-llm"
-version = "0.4.0"
+version = "0.4.0+post0"
 dependencies = [
  "ahash",
  "akin",
@@ -1202,7 +1202,7 @@ dependencies = [
 
 [[package]]
 name = "dynamo-py3"
-version = "0.4.0"
+version = "0.4.0+post0"
 dependencies = [
  "anyhow",
  "async-openai",
@@ -1229,7 +1229,7 @@ dependencies = [
 
 [[package]]
 name = "dynamo-runtime"
-version = "0.4.0"
+version = "0.4.0+post0"
 dependencies = [
  "anyhow",
  "arc-swap",

lib/runtime/examples/Cargo.lock

@@ -648,7 +648,7 @@ dependencies = [
 
 [[package]]
 name = "dynamo-runtime"
-version = "0.4.0"
+version = "0.4.0+post0"
 dependencies = [
  "anyhow",
  "arc-swap",
@@ -1020,7 +1020,7 @@ checksum = "2304e00983f87ffb38b55b444b5e3b60a884b5d30c0fca7d82fe33449bbe55ea"
 
 [[package]]
 name = "hello_world"
-version = "0.4.0"
+version = "0.4.0+post0"
 dependencies = [
  "dynamo-runtime",
 ]
@@ -2492,7 +2492,7 @@ dependencies = [
 
 [[package]]
 name = "service_metrics"
-version = "0.4.0"
+version = "0.4.0+post0"
 dependencies = [
  "dynamo-runtime",
  "futures",
@@ -2668,7 +2668,7 @@ dependencies = [
 
 [[package]]
 name = "system_metrics"
-version = "0.4.0"
+version = "0.4.0+post0"
 dependencies = [
  "anyhow",
  "dynamo-runtime",

lib/runtime/examples/system_metrics/tests/integration_test.rs

@@ -43,7 +43,7 @@ async fn test_backend_with_metrics() -> Result<()> {
     let distributed = DistributedRuntime::from_settings(runtime.clone()).await?;
 
     // Get the System status server info to find the actual port
-    let system_status_info = distributed.system_status_info();
+    let system_status_info = distributed.system_status_server_info();
     let system_status_port = match system_status_info {
         Some(info) => {
             println!("System status server running on: {}", info.address());

lib/runtime/src/component.rs

@@ -259,6 +259,7 @@ impl Component {
     /// Scrape ServiceSet, which contains NATS stats as well as user defined stats
     /// embedded in data field of ServiceInfo.
     pub async fn scrape_stats(&self, timeout: Duration) -> Result<ServiceSet> {
+        // Debug: scraping stats for component
         let service_name = self.service_name();
         let service_client = self.drt().service_client();
         service_client
@@ -268,9 +269,15 @@ impl Component {
 
     /// Add Prometheus metrics for this component's service stats.
     ///
-    /// Uses a channel to synchronize with the spawned async task, ensuring
-    /// metrics are updated before the callback returns.
-    pub fn add_metrics_callback(&self) -> Result<()> {
+    /// Starts a background task that scrapes stats every ~4.7s and updates metrics.
+    /// The thinking was that it should be a little bit shorter than the Prometheus polling interval.
+    /// Currently Prometheus polls every 6 seconds, and I wanted every poll to be fresh, so this is set
+    /// as an arbitrary 4.7 seconds plus 0.3 seconds if it times out. It's a bit of a hand-wavey decision.
+    pub fn start_scraping_metrics(&self) -> Result<()> {
+        const NATS_TIMEOUT_AND_INITIAL_DELAY_MS: std::time::Duration =
+            std::time::Duration::from_millis(300);
+        const MAX_DELAY_MS: std::time::Duration = std::time::Duration::from_millis(4700);
+
         let component_metrics = ComponentNatsPrometheusMetrics::new(self)?;
 
         let component_clone = self.clone();
@@ -281,59 +288,40 @@ impl Component {
             self.service_name()
         ); // it happens that in component, hierarchy and service name are the same
 
-        // Register a metrics callback that scrapes component statistics
-        let metrics_callback = Arc::new(move || {
-            // Timeout for scraping metrics from components (in milliseconds)
-            // This value is also used by KV Router metrics aggregator (300ms) and other components
-            const METRICS_SCRAPE_TIMEOUT_MS: u64 = 300;
-
-            // Get the current Tokio runtime handle
-            let handle = tokio::runtime::Handle::try_current()
-                .map_err(|err| anyhow::anyhow!("No Tokio runtime handle available: {}", err))?;
-
+        // Start a background task that scrapes stats every 5 seconds
         let m = component_metrics.clone();
         let c = component_clone.clone();
 
-            // Create a channel to synchronize with the spawned task
-            let (tx, rx) = std::sync::mpsc::channel::<anyhow::Result<()>>();
+        // Use std::thread for the background task to avoid runtime context issues
+        std::thread::spawn(move || {
+            // Use the existing secondary runtime from drt for background metrics scraping
+            let rt = c.drt().runtime().secondary();
+
+            // Run the background scraping loop
+            rt.block_on(async {
+                let timeout = NATS_TIMEOUT_AND_INITIAL_DELAY_MS;
+                let mut delay = NATS_TIMEOUT_AND_INITIAL_DELAY_MS;
 
-            let timeout = std::time::Duration::from_millis(METRICS_SCRAPE_TIMEOUT_MS);
-            handle.spawn(async move {
-                let result = match c.scrape_stats(timeout).await {
+                loop {
+                    match c.scrape_stats(timeout).await {
                         Ok(service_set) => {
                             m.update_from_service_set(&service_set);
-                        Ok(())
                         }
                         Err(err) => {
-                        // Reset metrics on failure
+                            tracing::error!(
+                                "Background scrape failed for {}: {}",
+                                c.service_name(),
+                                err
+                            );
                             m.reset_to_zeros();
-                        Err(anyhow::anyhow!("Failed to scrape stats: {}", err))
+                            // Double delay on failure, capped at MAX_DELAY
+                            delay = std::cmp::min(delay * 2, MAX_DELAY_MS);
                         }
-                };
-
-                // Send the result back to the waiting thread
-                // If send fails, the receiver has already given up waiting
-                let _ = tx.send(result);
-            });
-
-            // Wait for the spawned task to complete (with a timeout to prevent hanging)
-            // Add 100ms buffer to the scrape timeout to account for processing overhead
-            let recv_timeout = std::time::Duration::from_millis(METRICS_SCRAPE_TIMEOUT_MS + 100);
-            match rx.recv_timeout(recv_timeout) {
-                Ok(result) => result, // Return the actual result from scraping
-                Err(std::sync::mpsc::RecvTimeoutError::Timeout) => {
-                    component_metrics.reset_to_zeros();
-                    Err(anyhow::anyhow!("Metrics collection timed out"))
-                }
-                Err(std::sync::mpsc::RecvTimeoutError::Disconnected) => {
-                    component_metrics.reset_to_zeros();
-                    Err(anyhow::anyhow!("Metrics collection task failed"))
                     }
+                    tokio::time::sleep(delay).await;
                 }
             });
-
-        self.drt()
-            .register_metrics_callback(hierarchies, metrics_callback);
+        });
 
         Ok(())
     }
@@ -587,7 +575,7 @@ impl Namespace {
         // Register the metrics callback for this component.
         // If registration fails, log a warning but do not propagate the error,
         // as metrics are not mission critical and should not block component creation.
-        if let Err(err) = component.add_metrics_callback() {
+        if let Err(err) = component.start_scraping_metrics() {
             tracing::warn!(
                 "Failed to add metrics callback for component '{}': {}",
                 component.service_name(),

lib/runtime/src/distributed.rs

@@ -247,7 +247,8 @@ impl DistributedRuntime {
         self.instance_sources.clone()
     }
 
-    /// Add a Prometheus metric to a specific hierarchy's registry
+    /// Add a Prometheus metric to a specific hierarchy's registry. Note that it is possible
+    /// to register the same metric name multiple times, as long as the labels are different.
     pub fn add_prometheus_metric(
         &self,
         hierarchy: &str,
@@ -257,16 +258,6 @@ impl DistributedRuntime {
         let mut registries = self.hierarchy_to_metricsregistry.write().unwrap();
         let entry = registries.entry(hierarchy.to_string()).or_default();
 
-        // If a metric with this name already exists for the hierarchy, warn and skip registration
-        if entry.has_metric_named(metric_name) {
-            tracing::warn!(
-                hierarchy = ?hierarchy,
-                metric_name = ?metric_name,
-                "Metric already exists in registry; skipping registration"
-            );
-            return Ok(());
-        }
-
         // Try to register the metric and provide better error information
         match entry.prometheus_registry.register(prometheus_metric) {
             Ok(_) => Ok(()),

lib/runtime/src/metrics.rs

@@ -1089,11 +1089,9 @@ mod test_metricsregistry_prometheus_fmt_outputs {
         let endpoint_output =
             super::test_helpers::remove_nats_lines(&endpoint_output_raw).join("\n");
 
-        let expected_endpoint_output = format!(
-            r#"# HELP dynamo_component_testcounter A test counter
+        let expected_endpoint_output = r#"# HELP dynamo_component_testcounter A test counter
 # TYPE dynamo_component_testcounter counter
-dynamo_component_testcounter{{dynamo_component="comp345",dynamo_endpoint="ep345",dynamo_namespace="ns345"}} 123.456789"#
-        );
+dynamo_component_testcounter{dynamo_component="comp345",dynamo_endpoint="ep345",dynamo_namespace="ns345"} 123.456789"#.to_string();
 
         assert_eq!(
             endpoint_output, expected_endpoint_output,
@@ -1120,14 +1118,12 @@ dynamo_component_testcounter{{dynamo_component="comp345",dynamo_endpoint="ep345"
         let component_output =
             super::test_helpers::remove_nats_lines(&component_output_raw).join("\n");
 
-        let expected_component_output = format!(
-            r#"# HELP dynamo_component_testcounter A test counter
+        let expected_component_output = r#"# HELP dynamo_component_testcounter A test counter
 # TYPE dynamo_component_testcounter counter
-dynamo_component_testcounter{{dynamo_component="comp345",dynamo_endpoint="ep345",dynamo_namespace="ns345"}} 123.456789
+dynamo_component_testcounter{dynamo_component="comp345",dynamo_endpoint="ep345",dynamo_namespace="ns345"} 123.456789
 # HELP dynamo_component_testgauge A test gauge
 # TYPE dynamo_component_testgauge gauge
-dynamo_component_testgauge{{dynamo_component="comp345",dynamo_namespace="ns345"}} 50000"#
-        );
+dynamo_component_testgauge{dynamo_component="comp345",dynamo_namespace="ns345"} 50000"#.to_string();
 
         assert_eq!(
             component_output, expected_component_output,
@@ -1153,17 +1149,15 @@ dynamo_component_testgauge{{dynamo_component="comp345",dynamo_namespace="ns345"}
         let namespace_output =
             super::test_helpers::remove_nats_lines(&namespace_output_raw).join("\n");
 
-        let expected_namespace_output = format!(
-            r#"# HELP dynamo_component_testcounter A test counter
+        let expected_namespace_output = r#"# HELP dynamo_component_testcounter A test counter
 # TYPE dynamo_component_testcounter counter
-dynamo_component_testcounter{{dynamo_component="comp345",dynamo_endpoint="ep345",dynamo_namespace="ns345"}} 123.456789
+dynamo_component_testcounter{dynamo_component="comp345",dynamo_endpoint="ep345",dynamo_namespace="ns345"} 123.456789
 # HELP dynamo_component_testgauge A test gauge
 # TYPE dynamo_component_testgauge gauge
-dynamo_component_testgauge{{dynamo_component="comp345",dynamo_namespace="ns345"}} 50000
+dynamo_component_testgauge{dynamo_component="comp345",dynamo_namespace="ns345"} 50000
 # HELP dynamo_component_testintcounter A test int counter
 # TYPE dynamo_component_testintcounter counter
-dynamo_component_testintcounter{{dynamo_namespace="ns345"}} 12345"#
-        );
+dynamo_component_testintcounter{dynamo_namespace="ns345"} 12345"#.to_string();
 
         assert_eq!(
             namespace_output, expected_namespace_output,
@@ -1186,7 +1180,7 @@ dynamo_component_testintcounter{{dynamo_namespace="ns345"}} 12345"#
             .create_intgaugevec(
                 "testintgaugevec",
                 "A test int gauge vector",
-                &["instance", "service", "status"],
+                &["instance", "status"],
                 &[("service", "api")],
             )
             .unwrap();
@@ -1226,37 +1220,42 @@ dynamo_component_testintcounter{{dynamo_namespace="ns345"}} 12345"#
         let filtered_drt_output =
             super::test_helpers::remove_nats_lines(&drt_output_raw).join("\n");
 
-        let expected_drt_output = format!(
-            r#"# HELP dynamo_component_testcounter A test counter
+        let expected_drt_output = r#"# HELP dynamo_component_testcounter A test counter
 # TYPE dynamo_component_testcounter counter
-dynamo_component_testcounter{{dynamo_component="comp345",dynamo_endpoint="ep345",dynamo_namespace="ns345"}} 123.456789
+dynamo_component_testcounter{dynamo_component="comp345",dynamo_endpoint="ep345",dynamo_namespace="ns345"} 123.456789
 # HELP dynamo_component_testcountervec A test counter vector
 # TYPE dynamo_component_testcountervec counter
-dynamo_component_testcountervec{{method="GET",service="api",status="200"}} 10
-dynamo_component_testcountervec{{method="POST",service="api",status="201"}} 5
+dynamo_component_testcountervec{dynamo_component="comp345",dynamo_endpoint="ep345",dynamo_namespace="ns345",method="GET",service="api",status="200"} 10
+dynamo_component_testcountervec{dynamo_component="comp345",dynamo_endpoint="ep345",dynamo_namespace="ns345",method="POST",service="api",status="201"} 5
 # HELP dynamo_component_testgauge A test gauge
 # TYPE dynamo_component_testgauge gauge
-dynamo_component_testgauge{{dynamo_component="comp345",dynamo_namespace="ns345"}} 50000
+dynamo_component_testgauge{dynamo_component="comp345",dynamo_namespace="ns345"} 50000
 # HELP dynamo_component_testhistogram A test histogram
 # TYPE dynamo_component_testhistogram histogram
-dynamo_component_testhistogram_bucket{{le="1"}} 0
-dynamo_component_testhistogram_bucket{{le="2.5"}} 2
-dynamo_component_testhistogram_bucket{{le="5"}} 3
-dynamo_component_testhistogram_bucket{{le="10"}} 3
-dynamo_component_testhistogram_bucket{{le="+Inf"}} 3
-dynamo_component_testhistogram_sum 7.5
-dynamo_component_testhistogram_count 3
+dynamo_component_testhistogram_bucket{dynamo_component="comp345",dynamo_namespace="ns345",le="0.005"} 0
+dynamo_component_testhistogram_bucket{dynamo_component="comp345",dynamo_namespace="ns345",le="0.01"} 0
+dynamo_component_testhistogram_bucket{dynamo_component="comp345",dynamo_namespace="ns345",le="0.025"} 0
+dynamo_component_testhistogram_bucket{dynamo_component="comp345",dynamo_namespace="ns345",le="0.05"} 0
+dynamo_component_testhistogram_bucket{dynamo_component="comp345",dynamo_namespace="ns345",le="0.1"} 0
+dynamo_component_testhistogram_bucket{dynamo_component="comp345",dynamo_namespace="ns345",le="0.25"} 0
+dynamo_component_testhistogram_bucket{dynamo_component="comp345",dynamo_namespace="ns345",le="0.5"} 0
+dynamo_component_testhistogram_bucket{dynamo_component="comp345",dynamo_namespace="ns345",le="1"} 1
+dynamo_component_testhistogram_bucket{dynamo_component="comp345",dynamo_namespace="ns345",le="2.5"} 2
+dynamo_component_testhistogram_bucket{dynamo_component="comp345",dynamo_namespace="ns345",le="5"} 3
+dynamo_component_testhistogram_bucket{dynamo_component="comp345",dynamo_namespace="ns345",le="10"} 3
+dynamo_component_testhistogram_bucket{dynamo_component="comp345",dynamo_namespace="ns345",le="+Inf"} 3
+dynamo_component_testhistogram_sum{dynamo_component="comp345",dynamo_namespace="ns345"} 7.5
+dynamo_component_testhistogram_count{dynamo_component="comp345",dynamo_namespace="ns345"} 3
 # HELP dynamo_component_testintcounter A test int counter
 # TYPE dynamo_component_testintcounter counter
-dynamo_component_testintcounter{{dynamo_namespace="ns345"}} 12345
+dynamo_component_testintcounter{dynamo_namespace="ns345"} 12345
 # HELP dynamo_component_testintgauge A test int gauge
 # TYPE dynamo_component_testintgauge gauge
-dynamo_component_testintgauge 42
+dynamo_component_testintgauge{dynamo_namespace="ns345"} 42
 # HELP dynamo_component_testintgaugevec A test int gauge vector
 # TYPE dynamo_component_testintgaugevec gauge
-dynamo_component_testintgaugevec{{instance="server1",service="api",status="active"}} 10
-dynamo_component_testintgaugevec{{instance="server2",service="api",status="inactive"}} 0"#
-        );
+dynamo_component_testintgaugevec{dynamo_namespace="ns345",instance="server1",service="api",status="active"} 10
+dynamo_component_testintgaugevec{dynamo_namespace="ns345",instance="server2",service="api",status="inactive"} 0"#.to_string();
 
         assert_eq!(
             filtered_drt_output, expected_drt_output,
@@ -1480,7 +1479,7 @@ mod test_metricsregistry_nats {
                 input: SingleIn<String>,
             ) -> Result<ManyOut<Annotated<String>>, Error> {
                 let (data, ctx) = input.into_parts();
-                let response = format!("{}", data);
+                let response = data.to_string();
                 let stream = stream::iter(vec![Annotated::from_data(response)]);
                 Ok(ResponseStream::new(Box::pin(stream), ctx.context()))
             }
@@ -1505,7 +1504,7 @@ mod test_metricsregistry_nats {
         let drt_output = drt.prometheus_metrics_fmt().unwrap();
         let parsed_metrics: Vec<_> = drt_output
             .lines()
-            .filter_map(|line| super::test_helpers::parse_prometheus_metric(line))
+            .filter_map(super::test_helpers::parse_prometheus_metric)
             .collect();
 
         println!("=== Initial DRT metrics output ===");
@@ -1517,17 +1516,17 @@ mod test_metricsregistry_nats {
             // DRT NATS metrics (ordered to match DRT_NATS_METRICS)
             (build_metric_name(nats::CONNECTION_STATE), 1.0, 1.0), // Should be connected
             (build_metric_name(nats::CONNECTS), 1.0, 1.0),         // Should have 1 connection
-            (build_metric_name(nats::IN_TOTAL_BYTES), 300.0, 500.0), // ~75% to ~125% of 417
-            (build_metric_name(nats::IN_MESSAGES), 0.0, 0.0),      // No messages yet
-            (build_metric_name(nats::OUT_OVERHEAD_BYTES), 500.0, 700.0), // ~75% to ~125% of 612 (includes endpoint creation overhead)
-            (build_metric_name(nats::OUT_MESSAGES), 0.0, 0.0),           // No messages yet
+            (build_metric_name(nats::IN_TOTAL_BYTES), 400.0, 1500.0), // Wide range around 923
+            (build_metric_name(nats::IN_MESSAGES), 0.0, 5.0),      // Wide range around 2
+            (build_metric_name(nats::OUT_OVERHEAD_BYTES), 700.0, 2500.0), // Wide range around 1633
+            (build_metric_name(nats::OUT_MESSAGES), 0.0, 5.0),     // Wide range around 2
             // Component NATS metrics (ordered to match COMPONENT_NATS_METRICS)
             (build_metric_name(nats::AVG_PROCESSING_MS), 0.0, 0.0), // No processing yet
             (build_metric_name(nats::TOTAL_ERRORS), 0.0, 0.0),      // No errors yet
             (build_metric_name(nats::TOTAL_REQUESTS), 0.0, 0.0),    // No requests yet
             (build_metric_name(nats::TOTAL_PROCESSING_MS), 0.0, 0.0), // No processing yet
-            (build_metric_name(nats::ACTIVE_SERVICES), 0.0, 0.0),   // No services yet
-            (build_metric_name(nats::ACTIVE_ENDPOINTS), 0.0, 0.0),  // No endpoints yet
+            (build_metric_name(nats::ACTIVE_SERVICES), 0.0, 2.0), // Service may not be fully active yet
+            (build_metric_name(nats::ACTIVE_ENDPOINTS), 0.0, 2.0), // Endpoint may not be fully active yet
         ];
 
         for (metric_name, min_value, max_value) in &initial_expected_metric_values {
@@ -1576,7 +1575,6 @@ mod test_metricsregistry_nats {
                     );
                 }
             }
-            sleep(Duration::from_millis(100)).await;
         }
         println!("✓ Sent messages and received responses successfully");
 
@@ -1592,42 +1590,46 @@ mod test_metricsregistry_nats {
 
         let final_parsed_metrics: Vec<_> = super::test_helpers::extract_metrics(&final_drt_output)
             .iter()
-            .filter_map(|line| super::test_helpers::parse_prometheus_metric(line))
+            .filter_map(|line| super::test_helpers::parse_prometheus_metric(line.as_str()))
             .collect();
 
+        println!("\n=== Waiting 1 second for metrics to stabilize ===");
+        sleep(Duration::from_secs(1)).await;
+        println!("✓ Wait complete, checking final metrics...");
+
         let post_expected_metric_values = [
-            // DRT NATS metrics (ordered to match DRT_NATS_METRICS)
-            (build_metric_name(nats::CONNECTION_STATE), 1.0, 1.0), // Should remain connected
-            (build_metric_name(nats::CONNECTS), 1.0, 1.0),         // Should remain 1 connection
-            (build_metric_name(nats::IN_TOTAL_BYTES), 22000.0, 28000.0), // ~75% to ~125% of 24977 (10 messages × 2000 bytes + overhead)
-            (build_metric_name(nats::IN_MESSAGES), 10.0, 12.0), // Allow small drift (callback may run twice)
-            (build_metric_name(nats::OUT_OVERHEAD_BYTES), 2076.0, 3461.0), // ~75% to ~125% of 2769 (synchronous metrics collection overhead)
-            (build_metric_name(nats::OUT_MESSAGES), 10.0, 12.0), // Allow small drift (callback may run twice)
-            // Component NATS metrics (ordered to match COMPONENT_NATS_METRICS)
-            (build_metric_name(nats::AVG_PROCESSING_MS), 0.0, 1.0), // Should be low processing time
-            (build_metric_name(nats::TOTAL_ERRORS), 0.0, 0.0),      // Should have no errors
-            (build_metric_name(nats::TOTAL_REQUESTS), 0.0, 0.0), // NATS metrics don't track work handler requests
-            (build_metric_name(nats::TOTAL_PROCESSING_MS), 0.0, 5.0), // Should be low total processing time
-            (build_metric_name(nats::ACTIVE_SERVICES), 0.0, 0.0), // NATS metrics don't track work handler services
-            (build_metric_name(nats::ACTIVE_ENDPOINTS), 0.0, 0.0), // NATS metrics don't track work handler endpoints
-            // Work handler metrics with ranges
-            (build_metric_name(work_handler::REQUESTS_TOTAL), 10.0, 10.0), // Exact count (10 messages)
+            // DRT NATS metrics
+            (build_metric_name(nats::CONNECTION_STATE), 1.0, 1.0), // Connected
+            (build_metric_name(nats::CONNECTS), 1.0, 1.0),         // 1 connection
+            (build_metric_name(nats::IN_TOTAL_BYTES), 20000.0, 32000.0), // Wide range around 26117
+            (build_metric_name(nats::IN_MESSAGES), 8.0, 20.0),     // Wide range around 16
+            (build_metric_name(nats::OUT_OVERHEAD_BYTES), 2500.0, 8000.0), // Wide range around 5524
+            (build_metric_name(nats::OUT_MESSAGES), 8.0, 20.0),    // Wide range around 16
+            // Component NATS metrics
+            (build_metric_name(nats::AVG_PROCESSING_MS), 0.0, 1.0), // Low processing time
+            (build_metric_name(nats::TOTAL_ERRORS), 0.0, 0.0),      // No errors
+            (build_metric_name(nats::TOTAL_REQUESTS), 0.0, 0.0),    // No work handler requests
+            (build_metric_name(nats::TOTAL_PROCESSING_MS), 0.0, 5.0), // Low total processing time
+            (build_metric_name(nats::ACTIVE_SERVICES), 0.0, 2.0), // Service may not be fully active
+            (build_metric_name(nats::ACTIVE_ENDPOINTS), 0.0, 2.0), // Endpoint may not be fully active
+            // Work handler metrics
+            (build_metric_name(work_handler::REQUESTS_TOTAL), 10.0, 10.0), // 10 messages
             (
                 build_metric_name(work_handler::REQUEST_BYTES_TOTAL),
                 21000.0,
                 26000.0,
-            ), // ~75% to ~125% of 23520 (10 × 2000 bytes + overhead)
+            ), // ~75-125% of 23520
             (
                 build_metric_name(work_handler::RESPONSE_BYTES_TOTAL),
                 18000.0,
                 23000.0,
-            ), // ~75% to ~125% of 20660 (10 × 2000 bytes + overhead, but response size varies)
-            // Additional component metrics
+            ), // ~75-125% of 20660
             (
                 build_metric_name(work_handler::CONCURRENT_REQUESTS),
                 0.0,
                 1.0,
-            ), // Should be 0 or very low
+            ), // 0 or very low
+            // Histograms have _{count,sum} suffixes
             (
                 format!(
                     "{}_count",
@@ -1635,15 +1637,15 @@ mod test_metricsregistry_nats {
                 ),
                 10.0,
                 10.0,
-            ), // Exact count (10 messages)
+            ), // 10 messages
             (
                 format!(
                     "{}_sum",
                     build_metric_name(work_handler::REQUEST_DURATION_SECONDS)
                 ),
-                0.001,
-                0.999,
-            ), // Processing time sum (10 messages)
+                0.0001,
+                1.0,
+            ), // Processing time sum (wide range)
         ];
 
         println!("\n=== Checking Post-Activity All Metrics (NATS + Work Handler) ===");