v25.3.0

assets/certs/cert.pem

@@ -1,19 +1,20 @@
 -----BEGIN CERTIFICATE-----
-MIIDCzCCAfOgAwIBAgIUPU4tviz3ZvsMDjCz1NZRT16b0Y4wDQYJKoZIhvcNAQEL
-BQAwFTETMBEGA1UEAwwKcHVzaC5yb2NrczAeFw0yNTAyMDMyMzA5MzRaFw0yNjAy
-MDMyMzA5MzRaMBUxEzARBgNVBAMMCnB1c2gucm9ja3MwggEiMA0GCSqGSIb3DQEB
-AQUAA4IBDwAwggEKAoIBAQCZMkBYD/pYLBv9MiyHTLRT24kQyPeJBtZqryibi1jk
-BT1ZgNl3yo5U6kjj/nYBU/oy7M4OFC0xyaJQ4wpvLHu7xzREqwT9N9WcDcxaahUi
-P8+PsjGyznPrtXa1ASzGAYMNvXyWWp3351UWZHMEs6eY/Y7i8m4+0NwP5h8RNBCF
-KSFS41Ee9rNAMCnQSHZv1vIzKeVYPmYnCVmL7X2kQb+gS6Rvq5sEGLLKMC5QtTwI
-rdkPGpx4xZirIyf8KANbt0sShwUDpiCSuOCtpze08jMzoHLG9Nv97cJQjb/BhiES
-hLL+YjfAUFjq0rQ38zFKLJ87QB9Jym05mY6IadGQLXVXAgMBAAGjUzBRMB0GA1Ud
-DgQWBBQjpowWjrql/Eo2EVjl29xcjuCgkTAfBgNVHSMEGDAWgBQjpowWjrql/Eo2
-EVjl29xcjuCgkTAPBgNVHRMBAf8EBTADAQH/MA0GCSqGSIb3DQEBCwUAA4IBAQAY
-44vqbaf6ewFrZC0f3Kk4A10lC6qjWkcDFfw+JE8nzt+4+xPqp1eWgZKF2rONyAv2
-nG41Xygt19ByancXLU44KB24LX8F1GV5Oo7CGBA+xtoSPc0JulXw9fGclZDC6XiR
-P/+vhGgCHicbfP2O+N00pOifrTtf2tmOT4iPXRRo4TxmPzuCd+ZJTlBhPlKCmICq
-yGdAiEo6HsSiP+M5qVlNx8s57MhQYk5TpgmI6FU4mO7zfDfSatFonlg+aDbrnaqJ
-v/+km02M+oB460GmKwsSTnThHZgLNCLiKqD8bdziiCQjx5u0GjLI6468o+Aehb8l
-l/x9vWTTk/QKq41X5hFk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 -----END CERTIFICATE-----

assets/certs/key.pem

@@ -1,28 +1,28 @@
 -----BEGIN PRIVATE KEY-----
-MIIEvAIBADANBgkqhkiG9w0BAQEFAASCBKYwggSiAgEAAoIBAQCZMkBYD/pYLBv9
-MiyHTLRT24kQyPeJBtZqryibi1jkBT1ZgNl3yo5U6kjj/nYBU/oy7M4OFC0xyaJQ
-4wpvLHu7xzREqwT9N9WcDcxaahUiP8+PsjGyznPrtXa1ASzGAYMNvXyWWp3351UW
-ZHMEs6eY/Y7i8m4+0NwP5h8RNBCFKSFS41Ee9rNAMCnQSHZv1vIzKeVYPmYnCVmL
-7X2kQb+gS6Rvq5sEGLLKMC5QtTwIrdkPGpx4xZirIyf8KANbt0sShwUDpiCSuOCt
-pze08jMzoHLG9Nv97cJQjb/BhiEShLL+YjfAUFjq0rQ38zFKLJ87QB9Jym05mY6I
-adGQLXVXAgMBAAECggEARGCBBq1PBHbfoUH5TQSIAlvdEEBa9+602lZG7jIioVfT
-W7Uem5Ctuan+kcDcY9hbNsqqZ+9KgsvoJmlIGXoF2jjeE/4vUmRO9AHWoc5yk2Be
-4NjcxN3QMLdEfiLBnLlFCOd4CdX1ZxZ6TG3WRpV3a1pVIeeqHGB1sKT6Xd/atcwG
-RvpiXzu0SutGxVb6WE9r6hovZ4fVERCyCRczUGrUH5ICbxf6E7L4u8xjEYR4uEKK
-/8ZkDqrWdRASDAdPPMNqnHUEAho/WnxpNeb6B4lvvv2QWxIS9H1OikF/NzWPgVNS
-oPpvtJgjyo5xdgLm3zE4lcSPNVSrh1TBXuAn9TG4WQKBgQDScPFkUNBqjC5iPMof
-bqDHlhlptrHmiv9LC0lgjEDPgIEQfjLfdCugwDk32QyAcb5B60upDYeqCFDkfV/C
-T536qxevYPjPAjahLPHqMxkWpjvtY6NOTgbbcpVtblU2Fj8R8qbyPNADG31LicU9
-GVPtQ4YcVaMWCYbg5107+9dFWQKBgQC6XK+foKK+81RFdrqaNNgebTWTsANnBcZe
-xl0bj6oL5yY0IzroxHvgcNS7UMriWCu+K2xfkUBdMmxU773VN5JQ5k15ezjgtrvc
-8oAaEsxYP4su12JSTC/zsBANUgrNbFj8++qqKYWt2aQc2O/kbZ4MNfekIVFc8AjM
-2X9PxvxKLwKBgHXL7QO3TQLnVyt8VbQEjBFMzwriznB7i+4o8jkOKVU93IEr8zQr
-5iQElcLSR3I6uUJTALYvsaoXH5jXKVwujwL69LYiNQRDe+r6qqvrUHbiNJdsd8Rk
-XuhGGqj34tD04Pcd+h+MtO+YWqmHBBZwcA9XBeIkebbjPFH2kLT8AwN5AoGAYQy9
-hMJxnkE3hIkk+gNE/OtgeE20J+Vw/ZANkrnJEzPHyGUEW41e+W2oyvdzAFZsSTdx
-037f5ujIU58Z27x53NliRT4vS4693H0Iyws5EUfeIoGVuUflvODWKymraHjhCrXh
-6cV/0R5DAabTnsCbCr7b/MRBC8YQvyUQ0KnOXo8CgYBQYGpvJnSWyvsCjtb6apTP
-drjcBhVd0aSBpLGtDdtUCV4oLl9HPy+cLzcGaqckBqCwEq5DKruhMEf7on56bUMd
-m/3ItFk1TnhysAeJHb3zLqmJ9CKBitpqLlsOE7MEXVNmbTYeXU10Uo9yOfyt1i7T
-su+nT5VtyPkmF/l4wZl5+g==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 -----END PRIVATE KEY-----

changelog.md

@@ -1,5 +1,61 @@
 # Changelog
 
+## 2026-02-14 - 25.3.0 - feat(smart-proxy)
+add background concurrent certificate provisioning with per-domain timeouts and concurrency control
+
+- Add ISmartProxyOptions settings: certProvisionTimeout (ms) and certProvisionConcurrency (default 4)
+- Run certProvisionFunction as fire-and-forget background tasks (stores promise on start/route-update and awaited on stop)
+- Provision certificates in parallel with a concurrency limit using a new ConcurrencySemaphore utility
+- Introduce per-domain timeout handling (default 300000ms) via withTimeout and surface timeout errors as certificate-failed events
+- Refactor provisioning into provisionSingleDomain to isolate domain handling, ACME fallback preserved
+- Run provisioning outside route update mutex so route updates are not blocked by slow provisioning
+
+## 2026-02-14 - 25.2.2 - fix(smart-proxy)
+start metrics polling before certificate provisioning to avoid blocking metrics collection
+
+- Start metrics polling immediately after Rust engine startup so metrics are available without waiting for certificate provisioning.
+- Run certProvisionFunction after startup because ACME/DNS-01 provisioning can hang or be slow and must not block observability.
+- Code change in ts/proxies/smart-proxy/smart-proxy.ts: metricsAdapter.startPolling() moved to run before provisionCertificatesViaCallback().
+
+## 2026-02-14 - 25.2.1 - fix(smartproxy)
+no changes detected in git diff
+
+- The provided diff contains no file changes; no code or documentation updates to release.
+
+## 2026-02-14 - 25.2.0 - feat(metrics)
+add per-IP and HTTP-request metrics, propagate source IP through proxy paths, and expose new metrics to the TS adapter
+
+- Add per-IP tracking and IpMetrics in MetricsCollector (active/total connections, bytes, throughput).
+- Add HTTP request counters and tracking (total_http_requests, http_requests_per_sec, recent counters and tests).
+- Include throughput history (ThroughputSample serialization, retention and snapshotting) and expose history in snapshots.
+- Propagate source IP through HTTP and passthrough code paths: CountingBody.record_bytes and MetricsCollector methods now accept source_ip; connection_opened/closed calls include source IP.
+- Introduce ForwardMetricsCtx to carry metrics context (collector, route_id, source_ip) into passthrough forwarding routines; update ConnectionGuard to include source_ip.
+- TypeScript adapter (rust-metrics-adapter.ts) updated to return per-IP counts, top IPs, per-IP throughput, throughput history mapping, and HTTP request rates/total where available.
+- Numerous unit tests added for per-IP tracking, HTTP request tracking, throughput history and ThroughputTracker.history behavior.
+
+## 2026-02-13 - 25.1.0 - feat(metrics)
+add real-time throughput sampling and byte-counting metrics
+
+- Add CountingBody wrapper to count HTTP request and response bytes and report them to MetricsCollector.
+- Implement lock-free hot-path byte recording and a cold-path sampling API (sample_all) in MetricsCollector with throughput history and configurable retention (default 3600s).
+- Spawn a background sampling task in RustProxy (configurable sample_interval_ms) and tear it down on stop so throughput trackers are regularly sampled.
+- Instrument passthrough TCP forwarding and socket-relay paths to record per-chunk bytes (lock-free) so long-lived connections contribute to throughput measurements.
+- Wrap HTTP request/response bodies with CountingBody in proxy_service to capture bytes_in/bytes_out and report on body completion; connection_closed handling updated accordingly.
+- Expose recent throughput metrics to the TypeScript adapter (throughputRecentIn/Out) and pass metrics settings from the TS SmartProxy into Rust.
+- Add http-body dependency and update Cargo.toml/Cargo.lock entries for the new body wrapper usage.
+- Add unit tests for MetricsCollector throughput tracking and a new end-to-end throughput test (test.throughput.ts).
+- Update test certificates (assets/certs cert.pem and key.pem) used by TLS tests.
+
+## 2026-02-13 - 25.0.0 - BREAKING CHANGE(certs)
+accept a second eventComms argument in certProvisionFunction, add cert provisioning event types, and emit certificate lifecycle events
+
+- Breaking API change: certProvisionFunction signature changed from (domain: string) => Promise<TSmartProxyCertProvisionObject> to (domain: string, eventComms: ICertProvisionEventComms) => Promise<TSmartProxyCertProvisionObject>. Custom provisioners must accept (or safely ignore) the new second argument.
+- New types added and exported: ICertProvisionEventComms, ICertificateIssuedEvent, ICertificateFailedEvent.
+- smart-proxy now constructs an eventComms channel that allows provisioners to log/warn/error and set expiry date and source for the issued event.
+- Emits 'certificate-issued' (domain, expiryDate, source, isRenewal?) on successful provisioning and 'certificate-failed' (domain, error, source) on failures.
+- Updated public exports to include the new types so they are available to consumers.
+- Removed readme.byte-counting-audit.md (documentation file deleted).
+
 ## 2026-02-13 - 24.0.1 - fix(proxy)
 improve proxy robustness: add connect timeouts, graceful shutdown, WebSocket watchdog, and metrics guard
 

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@push.rocks/smartproxy",
-  "version": "24.0.1",
+  "version": "25.3.0",
   "private": false,
   "description": "A powerful proxy package with unified route-based configuration for high traffic management. Features include SSL/TLS support, flexible routing patterns, WebSocket handling, advanced security options, and automatic ACME certificate management.",
   "main": "dist_ts/index.js",

readme.byte-counting-audit.md

@@ -1,169 +0,0 @@
-# SmartProxy Byte Counting Audit Report
-
-## Executive Summary
-
-After a comprehensive audit of the SmartProxy codebase, I can confirm that **byte counting is implemented correctly** with no instances of double counting. Each byte transferred through the proxy is counted exactly once in each direction.
-
-## Byte Counting Implementation
-
-### 1. Core Tracking Mechanisms
-
-SmartProxy uses two complementary tracking systems:
-
-1. **Connection Records** (`IConnectionRecord`):
-   - `bytesReceived`: Total bytes received from client
-   - `bytesSent`: Total bytes sent to client
-
-2. **MetricsCollector**:
-   - Global throughput tracking via `ThroughputTracker`
-   - Per-connection byte tracking for route/IP metrics
-   - Called via `recordBytes(connectionId, bytesIn, bytesOut)`
-
-### 2. Where Bytes Are Counted
-
-Bytes are counted in only two files:
-
-#### a) `route-connection-handler.ts`
-- **Line 351**: TLS alert bytes when no SNI is provided
-- **Lines 1286-1301**: Data forwarding callbacks in `setupBidirectionalForwarding()`
-
-#### b) `http-proxy-bridge.ts`  
-- **Line 127**: Initial TLS chunk for HttpProxy connections
-- **Lines 142-154**: Data forwarding callbacks in `setupBidirectionalForwarding()`
-
-## Connection Flow Analysis
-
-### 1. Direct TCP Connection (No TLS)
-
-```
-Client → SmartProxy → Target Server
-```
-
-1. Connection arrives at `RouteConnectionHandler.handleConnection()`
-2. For non-TLS ports, immediately routes via `routeConnection()`
-3. `setupDirectConnection()` creates target connection
-4. `setupBidirectionalForwarding()` handles all data transfer:
-   - `onClientData`: `bytesReceived += chunk.length` + `recordBytes(chunk.length, 0)`
-   - `onServerData`: `bytesSent += chunk.length` + `recordBytes(0, chunk.length)`
-
-**Result**: ✅ Each byte counted exactly once
-
-### 2. TLS Passthrough Connection
-
-```
-Client (TLS) → SmartProxy → Target Server (TLS)
-```
-
-1. Connection waits for initial data to detect TLS
-2. TLS handshake detected, SNI extracted
-3. Route matched, `setupDirectConnection()` called
-4. Initial chunk stored in `pendingData` (NOT counted yet)
-5. On target connect, `pendingData` written to target (still not counted)
-6. `setupBidirectionalForwarding()` counts ALL bytes including initial chunk
-
-**Result**: ✅ Each byte counted exactly once
-
-### 3. TLS Termination via HttpProxy
-
-```
-Client (TLS) → SmartProxy → HttpProxy (localhost) → Target Server
-```
-
-1. TLS connection detected with `tls.mode = "terminate"`
-2. `forwardToHttpProxy()` called:
-   - Initial chunk: `bytesReceived += chunk.length` + `recordBytes(chunk.length, 0)`
-3. Proxy connection created to HttpProxy on localhost
-4. `setupBidirectionalForwarding()` handles subsequent data
-
-**Result**: ✅ Each byte counted exactly once
-
-### 4. HTTP Connection via HttpProxy
-
-```
-Client (HTTP) → SmartProxy → HttpProxy (localhost) → Target Server
-```
-
-1. Connection on configured HTTP port (`useHttpProxy` ports)
-2. Same flow as TLS termination
-3. All byte counting identical to TLS termination
-
-**Result**: ✅ Each byte counted exactly once
-
-### 5. NFTables Forwarding
-
-```
-Client → [Kernel NFTables] → Target Server
-```
-
-1. Connection detected, route matched with `forwardingEngine: 'nftables'`
-2. Connection marked as `usingNetworkProxy = true`
-3. NO application-level forwarding (kernel handles packet routing)
-4. NO byte counting in application layer
-
-**Result**: ✅ No counting (correct - kernel handles everything)
-
-## Special Cases
-
-### PROXY Protocol
-- PROXY protocol headers sent to backend servers are NOT counted in client metrics
-- Only actual client data is counted
-- **Correct behavior**: Protocol overhead is not client data
-
-### TLS Alerts
-- TLS alerts (e.g., for missing SNI) are counted as sent bytes
-- **Correct behavior**: Alerts are actual data sent to the client
-
-### Initial Chunks
-- **Direct connections**: Stored in `pendingData`, counted when forwarded
-- **HttpProxy connections**: Counted immediately upon receipt
-- **Both approaches**: Count each byte exactly once
-
-## Verification Methodology
-
-1. **Code Analysis**: Searched for all instances of:
-   - `bytesReceived +=` and `bytesSent +=`
-   - `recordBytes()` calls
-   - Data forwarding implementations
-
-2. **Flow Tracing**: Followed data path for each connection type from entry to exit
-
-3. **Handler Review**: Examined all forwarding handlers to ensure no additional counting
-
-## Findings
-
-### ✅ No Double Counting Detected
-
-- Each byte is counted exactly once in the direction it flows
-- Connection records and metrics are updated consistently
-- No overlapping or duplicate counting logic found
-
-### Areas of Excellence
-
-1. **Centralized Counting**: All byte counting happens in just two files
-2. **Consistent Pattern**: Uses `setupBidirectionalForwarding()` with callbacks
-3. **Clear Separation**: Forwarding handlers don't interfere with proxy metrics
-
-## Recommendations
-
-1. **Debug Logging**: Add optional debug logging to verify byte counts in production:
-   ```typescript
-   if (settings.debugByteCount) {
-     logger.log('debug', `Bytes counted: ${connectionId} +${bytes} (total: ${record.bytesReceived})`);
-   }
-   ```
-
-2. **Unit Tests**: Create specific tests to ensure byte counting accuracy:
-   - Test initial chunk handling
-   - Test PROXY protocol overhead exclusion
-   - Test HttpProxy forwarding accuracy
-
-3. **Protocol Overhead Tracking**: Consider separately tracking:
-   - PROXY protocol headers
-   - TLS handshake bytes
-   - HTTP headers vs body
-
-4. **NFTables Documentation**: Clearly document that NFTables-forwarded connections are not included in application metrics
-
-## Conclusion
-
-SmartProxy's byte counting implementation is **robust and accurate**. The design ensures that each byte is counted exactly once, with clear separation between connection tracking and metrics collection. No remediation is required.

rust/Cargo.lock

@@ -961,6 +961,7 @@ dependencies = [
  "arc-swap",
  "bytes",
  "dashmap",
+ "http-body",
  "http-body-util",
  "hyper",
  "hyper-util",

rust/Cargo.toml

@@ -29,6 +29,7 @@ serde_json = "1"
 # HTTP proxy engine (hyper-based)
 hyper = { version = "1", features = ["http1", "http2", "server", "client"] }
 hyper-util = { version = "0.1", features = ["tokio", "http1", "http2", "client-legacy", "server-auto"] }
+http-body = "1"
 http-body-util = "0.1"
 bytes = "1"
 

rust/crates/rustproxy-http/Cargo.toml

@@ -14,6 +14,7 @@ rustproxy-metrics = { workspace = true }
 hyper = { workspace = true }
 hyper-util = { workspace = true }
 regex = { workspace = true }
+http-body = { workspace = true }
 http-body-util = { workspace = true }
 bytes = { workspace = true }
 tokio = { workspace = true }

rust/crates/rustproxy-http/src/counting_body.rs

@@ -0,0 +1,126 @@
+//! A body wrapper that counts bytes flowing through and reports them to MetricsCollector.
+
+use std::pin::Pin;
+use std::sync::Arc;
+use std::sync::atomic::{AtomicU64, Ordering};
+use std::task::{Context, Poll};
+
+use bytes::Bytes;
+use http_body::Frame;
+use rustproxy_metrics::MetricsCollector;
+
+/// Wraps any `http_body::Body` and counts data bytes passing through.
+///
+/// When the body is fully consumed or dropped, accumulated byte counts
+/// are reported to the `MetricsCollector`.
+///
+/// The inner body is pinned on the heap to support `!Unpin` types like `hyper::body::Incoming`.
+pub struct CountingBody<B> {
+    inner: Pin<Box<B>>,
+    counted_bytes: AtomicU64,
+    metrics: Arc<MetricsCollector>,
+    route_id: Option<String>,
+    source_ip: Option<String>,
+    /// Whether we count bytes as "in" (request body) or "out" (response body).
+    direction: Direction,
+    /// Whether we've already reported the bytes (to avoid double-reporting on drop).
+    reported: bool,
+}
+
+/// Which direction the bytes flow.
+#[derive(Clone, Copy)]
+pub enum Direction {
+    /// Request body: bytes flowing from client → upstream (counted as bytes_in)
+    In,
+    /// Response body: bytes flowing from upstream → client (counted as bytes_out)
+    Out,
+}
+
+impl<B> CountingBody<B> {
+    /// Create a new CountingBody wrapping an inner body.
+    pub fn new(
+        inner: B,
+        metrics: Arc<MetricsCollector>,
+        route_id: Option<String>,
+        source_ip: Option<String>,
+        direction: Direction,
+    ) -> Self {
+        Self {
+            inner: Box::pin(inner),
+            counted_bytes: AtomicU64::new(0),
+            metrics,
+            route_id,
+            source_ip,
+            direction,
+            reported: false,
+        }
+    }
+
+    /// Report accumulated bytes to the metrics collector.
+    fn report(&mut self) {
+        if self.reported {
+            return;
+        }
+        self.reported = true;
+
+        let bytes = self.counted_bytes.load(Ordering::Relaxed);
+        if bytes == 0 {
+            return;
+        }
+
+        let route_id = self.route_id.as_deref();
+        let source_ip = self.source_ip.as_deref();
+        match self.direction {
+            Direction::In => self.metrics.record_bytes(bytes, 0, route_id, source_ip),
+            Direction::Out => self.metrics.record_bytes(0, bytes, route_id, source_ip),
+        }
+    }
+}
+
+impl<B> Drop for CountingBody<B> {
+    fn drop(&mut self) {
+        self.report();
+    }
+}
+
+// CountingBody is Unpin because inner is Pin<Box<B>> (always Unpin).
+impl<B> Unpin for CountingBody<B> {}
+
+impl<B> http_body::Body for CountingBody<B>
+where
+    B: http_body::Body<Data = Bytes>,
+{
+    type Data = Bytes;
+    type Error = B::Error;
+
+    fn poll_frame(
+        self: Pin<&mut Self>,
+        cx: &mut Context<'_>,
+    ) -> Poll<Option<Result<Frame<Self::Data>, Self::Error>>> {
+        let this = self.get_mut();
+
+        match this.inner.as_mut().poll_frame(cx) {
+            Poll::Ready(Some(Ok(frame))) => {
+                if let Some(data) = frame.data_ref() {
+                    this.counted_bytes.fetch_add(data.len() as u64, Ordering::Relaxed);
+                }
+                Poll::Ready(Some(Ok(frame)))
+            }
+            Poll::Ready(Some(Err(e))) => Poll::Ready(Some(Err(e))),
+            Poll::Ready(None) => {
+                // Body is fully consumed — report now
+                this.report();
+                Poll::Ready(None)
+            }
+            Poll::Pending => Poll::Pending,
+        }
+    }
+
+    fn is_end_stream(&self) -> bool {
+        self.inner.is_end_stream()
+    }
+
+    fn size_hint(&self) -> http_body::SizeHint {
+        self.inner.size_hint()
+    }
+}

rust/crates/rustproxy-http/src/lib.rs

@@ -3,12 +3,14 @@
 //! Hyper-based HTTP proxy service for RustProxy.
 //! Handles HTTP request parsing, route-based forwarding, and response filtering.
 
+pub mod counting_body;
 pub mod proxy_service;
 pub mod request_filter;
 pub mod response_filter;
 pub mod template;
 pub mod upstream_selector;
 
+pub use counting_body::*;
 pub use proxy_service::*;
 pub use template::*;
 pub use upstream_selector::*;

rust/crates/rustproxy-http/src/proxy_service.rs

@@ -21,6 +21,7 @@ use tracing::{debug, error, info, warn};
 use rustproxy_routing::RouteManager;
 use rustproxy_metrics::MetricsCollector;
 
+use crate::counting_body::{CountingBody, Direction};
 use crate::request_filter::RequestFilter;
 use crate::response_filter::ResponseFilter;
 use crate::upstream_selector::UpstreamSelector;
@@ -183,12 +184,14 @@ impl HttpProxyService {
         };
 
         let route_id = route_match.route.id.as_deref();
-        self.metrics.connection_opened(route_id);
+        let ip_str = peer_addr.ip().to_string();
+        self.metrics.record_http_request();
+        self.metrics.connection_opened(route_id, Some(&ip_str));
 
         // Apply request filters (IP check, rate limiting, auth)
         if let Some(ref security) = route_match.route.security {
             if let Some(response) = RequestFilter::apply(security, &req, &peer_addr) {
-                self.metrics.connection_closed(route_id);
+                self.metrics.connection_closed(route_id, Some(&ip_str));
                 return Ok(response);
             }
         }
@@ -196,7 +199,7 @@ impl HttpProxyService {
         // Check for test response (returns immediately, no upstream needed)
         if let Some(ref advanced) = route_match.route.action.advanced {
             if let Some(ref test_response) = advanced.test_response {
-                self.metrics.connection_closed(route_id);
+                self.metrics.connection_closed(route_id, Some(&ip_str));
                 return Ok(Self::build_test_response(test_response));
             }
         }
@@ -204,7 +207,7 @@ impl HttpProxyService {
         // Check for static file serving
         if let Some(ref advanced) = route_match.route.action.advanced {
             if let Some(ref static_files) = advanced.static_files {
-                self.metrics.connection_closed(route_id);
+                self.metrics.connection_closed(route_id, Some(&ip_str));
                 return Ok(Self::serve_static_file(&path, static_files));
             }
         }
@@ -213,7 +216,7 @@ impl HttpProxyService {
         let target = match route_match.target {
             Some(t) => t,
             None => {
-                self.metrics.connection_closed(route_id);
+                self.metrics.connection_closed(route_id, Some(&ip_str));
                 return Ok(error_response(StatusCode::BAD_GATEWAY, "No target available"));
             }
         };
@@ -231,7 +234,7 @@ impl HttpProxyService {
 
         if is_websocket {
             let result = self.handle_websocket_upgrade(
-                req, peer_addr, &upstream, route_match.route, route_id, &upstream_key, cancel,
+                req, peer_addr, &upstream, route_match.route, route_id, &upstream_key, cancel, &ip_str,
             ).await;
             // Note: for WebSocket, connection_ended is called inside
             // the spawned tunnel task when the connection closes.
@@ -279,13 +282,13 @@ impl HttpProxyService {
             Ok(Err(e)) => {
                 error!("Failed to connect to upstream {}:{}: {}", upstream.host, upstream.port, e);
                 self.upstream_selector.connection_ended(&upstream_key);
-                self.metrics.connection_closed(route_id);
+                self.metrics.connection_closed(route_id, Some(&ip_str));
                 return Ok(error_response(StatusCode::BAD_GATEWAY, "Backend unavailable"));
             }
             Err(_) => {
                 error!("Upstream connect timeout for {}:{}", upstream.host, upstream.port);
                 self.upstream_selector.connection_ended(&upstream_key);
-                self.metrics.connection_closed(route_id);
+                self.metrics.connection_closed(route_id, Some(&ip_str));
                 return Ok(error_response(StatusCode::GATEWAY_TIMEOUT, "Backend connect timeout"));
             }
         };
@@ -295,10 +298,10 @@ impl HttpProxyService {
 
         let result = if use_h2 {
             // HTTP/2 backend
-            self.forward_h2(io, parts, body, upstream_headers, &upstream_path, &upstream, route_match.route, route_id).await
+            self.forward_h2(io, parts, body, upstream_headers, &upstream_path, &upstream, route_match.route, route_id, &ip_str).await
         } else {
             // HTTP/1.1 backend (default)
-            self.forward_h1(io, parts, body, upstream_headers, &upstream_path, &upstream, route_match.route, route_id).await
+            self.forward_h1(io, parts, body, upstream_headers, &upstream_path, &upstream, route_match.route, route_id, &ip_str).await
         };
         self.upstream_selector.connection_ended(&upstream_key);
         result
@@ -315,12 +318,13 @@ impl HttpProxyService {
         upstream: &crate::upstream_selector::UpstreamSelection,
         route: &rustproxy_config::RouteConfig,
         route_id: Option<&str>,
+        source_ip: &str,
     ) -> Result<Response<BoxBody<Bytes, hyper::Error>>, hyper::Error> {
         let (mut sender, conn) = match hyper::client::conn::http1::handshake(io).await {
             Ok(h) => h,
             Err(e) => {
                 error!("Upstream handshake failed: {}", e);
-                self.metrics.connection_closed(route_id);
+                self.metrics.connection_closed(route_id, Some(source_ip));
                 return Ok(error_response(StatusCode::BAD_GATEWAY, "Backend handshake failed"));
             }
         };
@@ -345,19 +349,28 @@ impl HttpProxyService {
             }
         }
 
+        // Wrap the request body in CountingBody to track bytes_in
+        let counting_req_body = CountingBody::new(
+            body,
+            Arc::clone(&self.metrics),
+            route_id.map(|s| s.to_string()),
+            Some(source_ip.to_string()),
+            Direction::In,
+        );
+
         // Stream the request body through to upstream
-        let upstream_req = upstream_req.body(body).unwrap();
+        let upstream_req = upstream_req.body(counting_req_body).unwrap();
 
         let upstream_response = match sender.send_request(upstream_req).await {
             Ok(resp) => resp,
             Err(e) => {
                 error!("Upstream request failed: {}", e);
-                self.metrics.connection_closed(route_id);
+                self.metrics.connection_closed(route_id, Some(source_ip));
                 return Ok(error_response(StatusCode::BAD_GATEWAY, "Backend request failed"));
             }
         };
 
-        self.build_streaming_response(upstream_response, route, route_id).await
+        self.build_streaming_response(upstream_response, route, route_id, source_ip).await
     }
 
     /// Forward request to backend via HTTP/2 with body streaming.
@@ -371,13 +384,14 @@ impl HttpProxyService {
         upstream: &crate::upstream_selector::UpstreamSelection,
         route: &rustproxy_config::RouteConfig,
         route_id: Option<&str>,
+        source_ip: &str,
     ) -> Result<Response<BoxBody<Bytes, hyper::Error>>, hyper::Error> {
         let exec = hyper_util::rt::TokioExecutor::new();
         let (mut sender, conn) = match hyper::client::conn::http2::handshake(exec, io).await {
             Ok(h) => h,
             Err(e) => {
                 error!("HTTP/2 upstream handshake failed: {}", e);
-                self.metrics.connection_closed(route_id);
+                self.metrics.connection_closed(route_id, Some(source_ip));
                 return Ok(error_response(StatusCode::BAD_GATEWAY, "Backend H2 handshake failed"));
             }
         };
@@ -401,27 +415,41 @@ impl HttpProxyService {
             }
         }
 
+        // Wrap the request body in CountingBody to track bytes_in
+        let counting_req_body = CountingBody::new(
+            body,
+            Arc::clone(&self.metrics),
+            route_id.map(|s| s.to_string()),
+            Some(source_ip.to_string()),
+            Direction::In,
+        );
+
         // Stream the request body through to upstream
-        let upstream_req = upstream_req.body(body).unwrap();
+        let upstream_req = upstream_req.body(counting_req_body).unwrap();
 
         let upstream_response = match sender.send_request(upstream_req).await {
             Ok(resp) => resp,
             Err(e) => {
                 error!("HTTP/2 upstream request failed: {}", e);
-                self.metrics.connection_closed(route_id);
+                self.metrics.connection_closed(route_id, Some(source_ip));
                 return Ok(error_response(StatusCode::BAD_GATEWAY, "Backend H2 request failed"));
             }
         };
 
-        self.build_streaming_response(upstream_response, route, route_id).await
+        self.build_streaming_response(upstream_response, route, route_id, source_ip).await
     }
 
     /// Build the client-facing response from an upstream response, streaming the body.
+    ///
+    /// The response body is wrapped in a `CountingBody` that counts bytes as they
+    /// stream from upstream to client. When the body is fully consumed (or dropped),
+    /// it reports byte counts to the metrics collector and calls `connection_closed`.
     async fn build_streaming_response(
         &self,
         upstream_response: Response<Incoming>,
         route: &rustproxy_config::RouteConfig,
         route_id: Option<&str>,
+        source_ip: &str,
     ) -> Result<Response<BoxBody<Bytes, hyper::Error>>, hyper::Error> {
         let (resp_parts, resp_body) = upstream_response.into_parts();
 
@@ -433,10 +461,23 @@ impl HttpProxyService {
             ResponseFilter::apply_headers(route, headers, None);
         }
 
-        self.metrics.connection_closed(route_id);
+        // Wrap the response body in CountingBody to track bytes_out.
+        // CountingBody will report bytes and we close the connection metric
+        // after the body stream completes (not before it even starts).
+        let counting_body = CountingBody::new(
+            resp_body,
+            Arc::clone(&self.metrics),
+            route_id.map(|s| s.to_string()),
+            Some(source_ip.to_string()),
+            Direction::Out,
+        );
 
-        // Stream the response body directly from upstream to client
-        let body: BoxBody<Bytes, hyper::Error> = BoxBody::new(resp_body);
+        // Close the connection metric now — the HTTP request/response cycle is done
+        // from the proxy's perspective once we hand the streaming body to hyper.
+        // Bytes will still be counted as they flow.
+        self.metrics.connection_closed(route_id, Some(source_ip));
+
+        let body: BoxBody<Bytes, hyper::Error> = BoxBody::new(counting_body);
 
         Ok(response.body(body).unwrap())
     }
@@ -451,6 +492,7 @@ impl HttpProxyService {
         route_id: Option<&str>,
         upstream_key: &str,
         cancel: CancellationToken,
+        source_ip: &str,
     ) -> Result<Response<BoxBody<Bytes, hyper::Error>>, hyper::Error> {
         use tokio::io::{AsyncReadExt, AsyncWriteExt};
 
@@ -466,7 +508,7 @@ impl HttpProxyService {
                     .unwrap_or("");
                 if !allowed_origins.is_empty() && !allowed_origins.iter().any(|o| o == "*" || o == origin) {
                     self.upstream_selector.connection_ended(upstream_key);
-                    self.metrics.connection_closed(route_id);
+                    self.metrics.connection_closed(route_id, Some(source_ip));
                     return Ok(error_response(StatusCode::FORBIDDEN, "Origin not allowed"));
                 }
             }
@@ -483,13 +525,13 @@ impl HttpProxyService {
             Ok(Err(e)) => {
                 error!("WebSocket: failed to connect upstream {}:{}: {}", upstream.host, upstream.port, e);
                 self.upstream_selector.connection_ended(upstream_key);
-                self.metrics.connection_closed(route_id);
+                self.metrics.connection_closed(route_id, Some(source_ip));
                 return Ok(error_response(StatusCode::BAD_GATEWAY, "Backend unavailable"));
             }
             Err(_) => {
                 error!("WebSocket: upstream connect timeout for {}:{}", upstream.host, upstream.port);
                 self.upstream_selector.connection_ended(upstream_key);
-                self.metrics.connection_closed(route_id);
+                self.metrics.connection_closed(route_id, Some(source_ip));
                 return Ok(error_response(StatusCode::GATEWAY_TIMEOUT, "Backend connect timeout"));
             }
         };
@@ -551,7 +593,7 @@ impl HttpProxyService {
         if let Err(e) = upstream_stream.write_all(raw_request.as_bytes()).await {
             error!("WebSocket: failed to send upgrade request to upstream: {}", e);
             self.upstream_selector.connection_ended(upstream_key);
-            self.metrics.connection_closed(route_id);
+            self.metrics.connection_closed(route_id, Some(source_ip));
             return Ok(error_response(StatusCode::BAD_GATEWAY, "Backend write failed"));
         }
 
@@ -562,7 +604,7 @@ impl HttpProxyService {
                 Ok(0) => {
                     error!("WebSocket: upstream closed before completing handshake");
                     self.upstream_selector.connection_ended(upstream_key);
-                    self.metrics.connection_closed(route_id);
+                    self.metrics.connection_closed(route_id, Some(source_ip));
                     return Ok(error_response(StatusCode::BAD_GATEWAY, "Backend closed"));
                 }
                 Ok(_) => {
@@ -576,14 +618,14 @@ impl HttpProxyService {
                     if response_buf.len() > 8192 {
                         error!("WebSocket: upstream response headers too large");
                         self.upstream_selector.connection_ended(upstream_key);
-                        self.metrics.connection_closed(route_id);
+                        self.metrics.connection_closed(route_id, Some(source_ip));
                         return Ok(error_response(StatusCode::BAD_GATEWAY, "Backend response too large"));
                     }
                 }
                 Err(e) => {
                     error!("WebSocket: failed to read upstream response: {}", e);
                     self.upstream_selector.connection_ended(upstream_key);
-                    self.metrics.connection_closed(route_id);
+                    self.metrics.connection_closed(route_id, Some(source_ip));
                     return Ok(error_response(StatusCode::BAD_GATEWAY, "Backend read failed"));
                 }
             }
@@ -601,7 +643,7 @@ impl HttpProxyService {
         if status_code != 101 {
             debug!("WebSocket: upstream rejected upgrade with status {}", status_code);
             self.upstream_selector.connection_ended(upstream_key);
-            self.metrics.connection_closed(route_id);
+            self.metrics.connection_closed(route_id, Some(source_ip));
             return Ok(error_response(
                 StatusCode::from_u16(status_code).unwrap_or(StatusCode::BAD_GATEWAY),
                 "WebSocket upgrade rejected by backend",
@@ -635,6 +677,7 @@ impl HttpProxyService {
 
         let metrics = Arc::clone(&self.metrics);
         let route_id_owned = route_id.map(|s| s.to_string());
+        let source_ip_owned = source_ip.to_string();
         let upstream_selector = self.upstream_selector.clone();
         let upstream_key_owned = upstream_key.to_string();
 
@@ -645,7 +688,7 @@ impl HttpProxyService {
                     debug!("WebSocket: client upgrade failed: {}", e);
                     upstream_selector.connection_ended(&upstream_key_owned);
                     if let Some(ref rid) = route_id_owned {
-                        metrics.connection_closed(Some(rid.as_str()));
+                        metrics.connection_closed(Some(rid.as_str()), Some(&source_ip_owned));
                     }
                     return;
                 }
@@ -750,8 +793,8 @@ impl HttpProxyService {
 
             upstream_selector.connection_ended(&upstream_key_owned);
             if let Some(ref rid) = route_id_owned {
-                metrics.record_bytes(bytes_in, bytes_out, Some(rid.as_str()));
-                metrics.connection_closed(Some(rid.as_str()));
+                metrics.record_bytes(bytes_in, bytes_out, Some(rid.as_str()), Some(&source_ip_owned));
+                metrics.connection_closed(Some(rid.as_str()), Some(&source_ip_owned));
             }
         });
 

rust/crates/rustproxy-metrics/src/collector.rs

@@ -1,6 +1,9 @@
 use dashmap::DashMap;
 use serde::{Deserialize, Serialize};
 use std::sync::atomic::{AtomicU64, Ordering};
+use std::sync::Mutex;
+
+use crate::throughput::{ThroughputSample, ThroughputTracker};
 
 /// Aggregated metrics snapshot.
 #[derive(Debug, Clone, Serialize, Deserialize)]
@@ -12,7 +15,14 @@ pub struct Metrics {
     pub bytes_out: u64,
     pub throughput_in_bytes_per_sec: u64,
     pub throughput_out_bytes_per_sec: u64,
+    pub throughput_recent_in_bytes_per_sec: u64,
+    pub throughput_recent_out_bytes_per_sec: u64,
     pub routes: std::collections::HashMap<String, RouteMetrics>,
+    pub ips: std::collections::HashMap<String, IpMetrics>,
+    pub throughput_history: Vec<ThroughputSample>,
+    pub total_http_requests: u64,
+    pub http_requests_per_sec: u64,
+    pub http_requests_per_sec_recent: u64,
 }
 
 /// Per-route metrics.
@@ -25,6 +35,20 @@ pub struct RouteMetrics {
     pub bytes_out: u64,
     pub throughput_in_bytes_per_sec: u64,
     pub throughput_out_bytes_per_sec: u64,
+    pub throughput_recent_in_bytes_per_sec: u64,
+    pub throughput_recent_out_bytes_per_sec: u64,
+}
+
+/// Per-IP metrics.
+#[derive(Debug, Clone, Serialize, Deserialize)]
+#[serde(rename_all = "camelCase")]
+pub struct IpMetrics {
+    pub active_connections: u64,
+    pub total_connections: u64,
+    pub bytes_in: u64,
+    pub bytes_out: u64,
+    pub throughput_in_bytes_per_sec: u64,
+    pub throughput_out_bytes_per_sec: u64,
 }
 
 /// Statistics snapshot.
@@ -38,7 +62,18 @@ pub struct Statistics {
     pub uptime_seconds: u64,
 }
 
+/// Default retention for throughput samples (1 hour).
+const DEFAULT_RETENTION_SECONDS: usize = 3600;
+
+/// Maximum number of IPs to include in a snapshot (top by active connections).
+const MAX_IPS_IN_SNAPSHOT: usize = 100;
+
 /// Metrics collector tracking connections and throughput.
+///
+/// Design: The hot path (`record_bytes`) is entirely lock-free — it only touches
+/// `AtomicU64` counters. The cold path (`sample_all`, called at 1Hz) drains
+/// those atomics and feeds the throughput trackers under a Mutex. This avoids
+/// contention when `record_bytes` is called per-chunk in the TCP copy loop.
 pub struct MetricsCollector {
     active_connections: AtomicU64,
     total_connections: AtomicU64,
@@ -51,10 +86,38 @@ pub struct MetricsCollector {
     /// Per-route byte counters
     route_bytes_in: DashMap<String, AtomicU64>,
     route_bytes_out: DashMap<String, AtomicU64>,
+
+    // ── Per-IP tracking ──
+    ip_connections: DashMap<String, AtomicU64>,
+    ip_total_connections: DashMap<String, AtomicU64>,
+    ip_bytes_in: DashMap<String, AtomicU64>,
+    ip_bytes_out: DashMap<String, AtomicU64>,
+    ip_pending_tp: DashMap<String, (AtomicU64, AtomicU64)>,
+    ip_throughput: DashMap<String, Mutex<ThroughputTracker>>,
+
+    // ── HTTP request tracking ──
+    total_http_requests: AtomicU64,
+    pending_http_requests: AtomicU64,
+    http_request_throughput: Mutex<ThroughputTracker>,
+
+    // ── Lock-free pending throughput counters (hot path) ──
+    global_pending_tp_in: AtomicU64,
+    global_pending_tp_out: AtomicU64,
+    route_pending_tp: DashMap<String, (AtomicU64, AtomicU64)>,
+
+    // ── Throughput history — only locked during sampling (cold path) ──
+    global_throughput: Mutex<ThroughputTracker>,
+    route_throughput: DashMap<String, Mutex<ThroughputTracker>>,
+    retention_seconds: usize,
 }
 
 impl MetricsCollector {
     pub fn new() -> Self {
+        Self::with_retention(DEFAULT_RETENTION_SECONDS)
+    }
+
+    /// Create a MetricsCollector with a custom retention period for throughput history.
+    pub fn with_retention(retention_seconds: usize) -> Self {
         Self {
             active_connections: AtomicU64::new(0),
             total_connections: AtomicU64::new(0),
@@ -64,11 +127,26 @@ impl MetricsCollector {
             route_total_connections: DashMap::new(),
             route_bytes_in: DashMap::new(),
             route_bytes_out: DashMap::new(),
+            ip_connections: DashMap::new(),
+            ip_total_connections: DashMap::new(),
+            ip_bytes_in: DashMap::new(),
+            ip_bytes_out: DashMap::new(),
+            ip_pending_tp: DashMap::new(),
+            ip_throughput: DashMap::new(),
+            total_http_requests: AtomicU64::new(0),
+            pending_http_requests: AtomicU64::new(0),
+            http_request_throughput: Mutex::new(ThroughputTracker::new(retention_seconds)),
+            global_pending_tp_in: AtomicU64::new(0),
+            global_pending_tp_out: AtomicU64::new(0),
+            route_pending_tp: DashMap::new(),
+            global_throughput: Mutex::new(ThroughputTracker::new(retention_seconds)),
+            route_throughput: DashMap::new(),
+            retention_seconds,
         }
     }
 
     /// Record a new connection.
-    pub fn connection_opened(&self, route_id: Option<&str>) {
+    pub fn connection_opened(&self, route_id: Option<&str>, source_ip: Option<&str>) {
         self.active_connections.fetch_add(1, Ordering::Relaxed);
         self.total_connections.fetch_add(1, Ordering::Relaxed);
 
@@ -82,10 +160,21 @@ impl MetricsCollector {
                 .or_insert_with(|| AtomicU64::new(0))
                 .fetch_add(1, Ordering::Relaxed);
         }
+
+        if let Some(ip) = source_ip {
+            self.ip_connections
+                .entry(ip.to_string())
+                .or_insert_with(|| AtomicU64::new(0))
+                .fetch_add(1, Ordering::Relaxed);
+            self.ip_total_connections
+                .entry(ip.to_string())
+                .or_insert_with(|| AtomicU64::new(0))
+                .fetch_add(1, Ordering::Relaxed);
+        }
     }
 
     /// Record a connection closing.
-    pub fn connection_closed(&self, route_id: Option<&str>) {
+    pub fn connection_closed(&self, route_id: Option<&str>, source_ip: Option<&str>) {
         self.active_connections.fetch_sub(1, Ordering::Relaxed);
 
         if let Some(route_id) = route_id {
@@ -96,13 +185,34 @@ impl MetricsCollector {
                 }
             }
         }
+
+        if let Some(ip) = source_ip {
+            if let Some(counter) = self.ip_connections.get(ip) {
+                let val = counter.load(Ordering::Relaxed);
+                if val > 0 {
+                    counter.fetch_sub(1, Ordering::Relaxed);
+                }
+                // Clean up zero-count entries to prevent memory growth
+                if val <= 1 {
+                    drop(counter);
+                    self.ip_connections.remove(ip);
+                }
+            }
+        }
     }
 
-    /// Record bytes transferred.
-    pub fn record_bytes(&self, bytes_in: u64, bytes_out: u64, route_id: Option<&str>) {
+    /// Record bytes transferred (lock-free hot path).
+    ///
+    /// Called per-chunk in the TCP copy loop. Only touches AtomicU64 counters —
+    /// no Mutex is taken. The throughput trackers are fed during `sample_all()`.
+    pub fn record_bytes(&self, bytes_in: u64, bytes_out: u64, route_id: Option<&str>, source_ip: Option<&str>) {
         self.total_bytes_in.fetch_add(bytes_in, Ordering::Relaxed);
         self.total_bytes_out.fetch_add(bytes_out, Ordering::Relaxed);
 
+        // Accumulate into lock-free pending throughput counters
+        self.global_pending_tp_in.fetch_add(bytes_in, Ordering::Relaxed);
+        self.global_pending_tp_out.fetch_add(bytes_out, Ordering::Relaxed);
+
         if let Some(route_id) = route_id {
             self.route_bytes_in
                 .entry(route_id.to_string())
@@ -112,6 +222,123 @@ impl MetricsCollector {
                 .entry(route_id.to_string())
                 .or_insert_with(|| AtomicU64::new(0))
                 .fetch_add(bytes_out, Ordering::Relaxed);
+
+            // Accumulate into per-route pending throughput counters (lock-free)
+            let entry = self.route_pending_tp
+                .entry(route_id.to_string())
+                .or_insert_with(|| (AtomicU64::new(0), AtomicU64::new(0)));
+            entry.0.fetch_add(bytes_in, Ordering::Relaxed);
+            entry.1.fetch_add(bytes_out, Ordering::Relaxed);
+        }
+
+        if let Some(ip) = source_ip {
+            self.ip_bytes_in
+                .entry(ip.to_string())
+                .or_insert_with(|| AtomicU64::new(0))
+                .fetch_add(bytes_in, Ordering::Relaxed);
+            self.ip_bytes_out
+                .entry(ip.to_string())
+                .or_insert_with(|| AtomicU64::new(0))
+                .fetch_add(bytes_out, Ordering::Relaxed);
+
+            // Accumulate into per-IP pending throughput counters (lock-free)
+            let entry = self.ip_pending_tp
+                .entry(ip.to_string())
+                .or_insert_with(|| (AtomicU64::new(0), AtomicU64::new(0)));
+            entry.0.fetch_add(bytes_in, Ordering::Relaxed);
+            entry.1.fetch_add(bytes_out, Ordering::Relaxed);
+        }
+    }
+
+    /// Record an HTTP request (called once per request in the HTTP proxy).
+    pub fn record_http_request(&self) {
+        self.total_http_requests.fetch_add(1, Ordering::Relaxed);
+        self.pending_http_requests.fetch_add(1, Ordering::Relaxed);
+    }
+
+    /// Take a throughput sample on all trackers (cold path, call at 1Hz or configured interval).
+    ///
+    /// Drains the lock-free pending counters and feeds the accumulated bytes
+    /// into the throughput trackers (under Mutex). This is the only place
+    /// the Mutex is locked.
+    pub fn sample_all(&self) {
+        // Drain global pending bytes and feed into the tracker
+        let global_in = self.global_pending_tp_in.swap(0, Ordering::Relaxed);
+        let global_out = self.global_pending_tp_out.swap(0, Ordering::Relaxed);
+        if let Ok(mut tracker) = self.global_throughput.lock() {
+            tracker.record_bytes(global_in, global_out);
+            tracker.sample();
+        }
+
+        // Drain per-route pending bytes; collect into a Vec to avoid holding DashMap shards
+        let mut route_samples: Vec<(String, u64, u64)> = Vec::new();
+        for entry in self.route_pending_tp.iter() {
+            let route_id = entry.key().clone();
+            let pending_in = entry.value().0.swap(0, Ordering::Relaxed);
+            let pending_out = entry.value().1.swap(0, Ordering::Relaxed);
+            route_samples.push((route_id, pending_in, pending_out));
+        }
+
+        // Feed pending bytes into route trackers and sample
+        let retention = self.retention_seconds;
+        for (route_id, pending_in, pending_out) in &route_samples {
+            // Ensure the tracker exists
+            self.route_throughput
+                .entry(route_id.clone())
+                .or_insert_with(|| Mutex::new(ThroughputTracker::new(retention)));
+            // Now get a separate ref and lock it
+            if let Some(tracker_ref) = self.route_throughput.get(route_id) {
+                if let Ok(mut tracker) = tracker_ref.value().lock() {
+                    tracker.record_bytes(*pending_in, *pending_out);
+                    tracker.sample();
+                }
+            }
+        }
+
+        // Also sample any route trackers that had no new pending bytes
+        // (to keep their sample window advancing)
+        for entry in self.route_throughput.iter() {
+            if !self.route_pending_tp.contains_key(entry.key()) {
+                if let Ok(mut tracker) = entry.value().lock() {
+                    tracker.sample();
+                }
+            }
+        }
+
+        // Drain per-IP pending bytes and feed into IP throughput trackers
+        let mut ip_samples: Vec<(String, u64, u64)> = Vec::new();
+        for entry in self.ip_pending_tp.iter() {
+            let ip = entry.key().clone();
+            let pending_in = entry.value().0.swap(0, Ordering::Relaxed);
+            let pending_out = entry.value().1.swap(0, Ordering::Relaxed);
+            ip_samples.push((ip, pending_in, pending_out));
+        }
+        for (ip, pending_in, pending_out) in &ip_samples {
+            self.ip_throughput
+                .entry(ip.clone())
+                .or_insert_with(|| Mutex::new(ThroughputTracker::new(retention)));
+            if let Some(tracker_ref) = self.ip_throughput.get(ip) {
+                if let Ok(mut tracker) = tracker_ref.value().lock() {
+                    tracker.record_bytes(*pending_in, *pending_out);
+                    tracker.sample();
+                }
+            }
+        }
+        // Sample idle IP trackers
+        for entry in self.ip_throughput.iter() {
+            if !self.ip_pending_tp.contains_key(entry.key()) {
+                if let Ok(mut tracker) = entry.value().lock() {
+                    tracker.sample();
+                }
+            }
+        }
+
+        // Drain pending HTTP request count and feed into HTTP throughput tracker
+        let pending_reqs = self.pending_http_requests.swap(0, Ordering::Relaxed);
+        if let Ok(mut tracker) = self.http_request_throughput.lock() {
+            // Use bytes_in field to track request count (each request = 1 "byte")
+            tracker.record_bytes(pending_reqs, 0);
+            tracker.sample();
         }
     }
 
@@ -135,10 +362,22 @@ impl MetricsCollector {
         self.total_bytes_out.load(Ordering::Relaxed)
     }
 
-    /// Get a full metrics snapshot including per-route data.
+    /// Get a full metrics snapshot including per-route and per-IP data.
     pub fn snapshot(&self) -> Metrics {
         let mut routes = std::collections::HashMap::new();
 
+        // Get global throughput (instant = last 1 sample, recent = last 10 samples)
+        let (global_tp_in, global_tp_out, global_recent_in, global_recent_out, throughput_history) =
+            self.global_throughput
+                .lock()
+                .map(|t| {
+                    let (i_in, i_out) = t.instant();
+                    let (r_in, r_out) = t.recent();
+                    let history = t.history(60);
+                    (i_in, i_out, r_in, r_out, history)
+                })
+                .unwrap_or((0, 0, 0, 0, Vec::new()));
+
         // Collect per-route metrics
         for entry in self.route_total_connections.iter() {
             let route_id = entry.key().clone();
@@ -156,24 +395,92 @@ impl MetricsCollector {
                 .map(|c| c.load(Ordering::Relaxed))
                 .unwrap_or(0);
 
+            let (route_tp_in, route_tp_out, route_recent_in, route_recent_out) = self.route_throughput
+                .get(&route_id)
+                .and_then(|entry| entry.value().lock().ok().map(|t| {
+                    let (i_in, i_out) = t.instant();
+                    let (r_in, r_out) = t.recent();
+                    (i_in, i_out, r_in, r_out)
+                }))
+                .unwrap_or((0, 0, 0, 0));
+
             routes.insert(route_id, RouteMetrics {
                 active_connections: active,
                 total_connections: total,
                 bytes_in,
                 bytes_out,
-                throughput_in_bytes_per_sec: 0,
-                throughput_out_bytes_per_sec: 0,
+                throughput_in_bytes_per_sec: route_tp_in,
+                throughput_out_bytes_per_sec: route_tp_out,
+                throughput_recent_in_bytes_per_sec: route_recent_in,
+                throughput_recent_out_bytes_per_sec: route_recent_out,
             });
         }
 
+        // Collect per-IP metrics — only IPs with active connections or total > 0,
+        // capped at top MAX_IPS_IN_SNAPSHOT sorted by active count
+        let mut ip_entries: Vec<(String, u64, u64, u64, u64, u64, u64)> = Vec::new();
+        for entry in self.ip_total_connections.iter() {
+            let ip = entry.key().clone();
+            let total = entry.value().load(Ordering::Relaxed);
+            let active = self.ip_connections
+                .get(&ip)
+                .map(|c| c.load(Ordering::Relaxed))
+                .unwrap_or(0);
+            let bytes_in = self.ip_bytes_in
+                .get(&ip)
+                .map(|c| c.load(Ordering::Relaxed))
+                .unwrap_or(0);
+            let bytes_out = self.ip_bytes_out
+                .get(&ip)
+                .map(|c| c.load(Ordering::Relaxed))
+                .unwrap_or(0);
+            let (tp_in, tp_out) = self.ip_throughput
+                .get(&ip)
+                .and_then(|entry| entry.value().lock().ok().map(|t| t.instant()))
+                .unwrap_or((0, 0));
+            ip_entries.push((ip, active, total, bytes_in, bytes_out, tp_in, tp_out));
+        }
+        // Sort by active connections descending, then cap
+        ip_entries.sort_by(|a, b| b.1.cmp(&a.1));
+        ip_entries.truncate(MAX_IPS_IN_SNAPSHOT);
+
+        let mut ips = std::collections::HashMap::new();
+        for (ip, active, total, bytes_in, bytes_out, tp_in, tp_out) in ip_entries {
+            ips.insert(ip, IpMetrics {
+                active_connections: active,
+                total_connections: total,
+                bytes_in,
+                bytes_out,
+                throughput_in_bytes_per_sec: tp_in,
+                throughput_out_bytes_per_sec: tp_out,
+            });
+        }
+
+        // HTTP request rates
+        let (http_rps, http_rps_recent) = self.http_request_throughput
+            .lock()
+            .map(|t| {
+                let (instant, _) = t.instant();
+                let (recent, _) = t.recent();
+                (instant, recent)
+            })
+            .unwrap_or((0, 0));
+
         Metrics {
             active_connections: self.active_connections(),
             total_connections: self.total_connections(),
             bytes_in: self.total_bytes_in(),
             bytes_out: self.total_bytes_out(),
-            throughput_in_bytes_per_sec: 0,
-            throughput_out_bytes_per_sec: 0,
+            throughput_in_bytes_per_sec: global_tp_in,
+            throughput_out_bytes_per_sec: global_tp_out,
+            throughput_recent_in_bytes_per_sec: global_recent_in,
+            throughput_recent_out_bytes_per_sec: global_recent_out,
             routes,
+            ips,
+            throughput_history,
+            total_http_requests: self.total_http_requests.load(Ordering::Relaxed),
+            http_requests_per_sec: http_rps,
+            http_requests_per_sec_recent: http_rps_recent,
         }
     }
 }
@@ -198,10 +505,10 @@ mod tests {
     #[test]
     fn test_connection_opened_increments() {
         let collector = MetricsCollector::new();
-        collector.connection_opened(None);
+        collector.connection_opened(None, None);
         assert_eq!(collector.active_connections(), 1);
         assert_eq!(collector.total_connections(), 1);
-        collector.connection_opened(None);
+        collector.connection_opened(None, None);
         assert_eq!(collector.active_connections(), 2);
         assert_eq!(collector.total_connections(), 2);
     }
@@ -209,10 +516,10 @@ mod tests {
     #[test]
     fn test_connection_closed_decrements() {
         let collector = MetricsCollector::new();
-        collector.connection_opened(None);
-        collector.connection_opened(None);
+        collector.connection_opened(None, None);
+        collector.connection_opened(None, None);
         assert_eq!(collector.active_connections(), 2);
-        collector.connection_closed(None);
+        collector.connection_closed(None, None);
         assert_eq!(collector.active_connections(), 1);
         // total_connections should stay at 2
         assert_eq!(collector.total_connections(), 2);
@@ -221,23 +528,23 @@ mod tests {
     #[test]
     fn test_route_specific_tracking() {
         let collector = MetricsCollector::new();
-        collector.connection_opened(Some("route-a"));
-        collector.connection_opened(Some("route-a"));
-        collector.connection_opened(Some("route-b"));
+        collector.connection_opened(Some("route-a"), None);
+        collector.connection_opened(Some("route-a"), None);
+        collector.connection_opened(Some("route-b"), None);
 
         assert_eq!(collector.active_connections(), 3);
         assert_eq!(collector.total_connections(), 3);
 
-        collector.connection_closed(Some("route-a"));
+        collector.connection_closed(Some("route-a"), None);
         assert_eq!(collector.active_connections(), 2);
     }
 
     #[test]
     fn test_record_bytes() {
         let collector = MetricsCollector::new();
-        collector.record_bytes(100, 200, Some("route-a"));
-        collector.record_bytes(50, 75, Some("route-a"));
-        collector.record_bytes(25, 30, None);
+        collector.record_bytes(100, 200, Some("route-a"), None);
+        collector.record_bytes(50, 75, Some("route-a"), None);
+        collector.record_bytes(25, 30, None, None);
 
         let total_in = collector.total_bytes_in.load(Ordering::Relaxed);
         let total_out = collector.total_bytes_out.load(Ordering::Relaxed);
@@ -248,4 +555,114 @@ mod tests {
         let route_in = collector.route_bytes_in.get("route-a").unwrap();
         assert_eq!(route_in.load(Ordering::Relaxed), 150);
     }
+
+    #[test]
+    fn test_throughput_tracking() {
+        let collector = MetricsCollector::with_retention(60);
+
+        // Open a connection so the route appears in the snapshot
+        collector.connection_opened(Some("route-a"), None);
+
+        // Record some bytes
+        collector.record_bytes(1000, 2000, Some("route-a"), None);
+        collector.record_bytes(500, 750, None, None);
+
+        // Take a sample (simulates the 1Hz tick)
+        collector.sample_all();
+
+        // Check global throughput
+        let snapshot = collector.snapshot();
+        assert_eq!(snapshot.throughput_in_bytes_per_sec, 1500);
+        assert_eq!(snapshot.throughput_out_bytes_per_sec, 2750);
+
+        // Check per-route throughput
+        let route_a = snapshot.routes.get("route-a").unwrap();
+        assert_eq!(route_a.throughput_in_bytes_per_sec, 1000);
+        assert_eq!(route_a.throughput_out_bytes_per_sec, 2000);
+    }
+
+    #[test]
+    fn test_throughput_zero_before_sampling() {
+        let collector = MetricsCollector::with_retention(60);
+        collector.record_bytes(1000, 2000, None, None);
+
+        // Without sampling, throughput should be 0
+        let snapshot = collector.snapshot();
+        assert_eq!(snapshot.throughput_in_bytes_per_sec, 0);
+        assert_eq!(snapshot.throughput_out_bytes_per_sec, 0);
+    }
+
+    #[test]
+    fn test_per_ip_tracking() {
+        let collector = MetricsCollector::with_retention(60);
+
+        collector.connection_opened(Some("route-a"), Some("1.2.3.4"));
+        collector.connection_opened(Some("route-a"), Some("1.2.3.4"));
+        collector.connection_opened(Some("route-b"), Some("5.6.7.8"));
+
+        // Check IP active connections (drop DashMap refs immediately to avoid deadlock)
+        assert_eq!(
+            collector.ip_connections.get("1.2.3.4").unwrap().load(Ordering::Relaxed),
+            2
+        );
+        assert_eq!(
+            collector.ip_connections.get("5.6.7.8").unwrap().load(Ordering::Relaxed),
+            1
+        );
+
+        // Record bytes per IP
+        collector.record_bytes(100, 200, Some("route-a"), Some("1.2.3.4"));
+        collector.record_bytes(300, 400, Some("route-b"), Some("5.6.7.8"));
+        collector.sample_all();
+
+        let snapshot = collector.snapshot();
+        assert_eq!(snapshot.ips.len(), 2);
+        let ip1_metrics = snapshot.ips.get("1.2.3.4").unwrap();
+        assert_eq!(ip1_metrics.active_connections, 2);
+        assert_eq!(ip1_metrics.bytes_in, 100);
+
+        // Close connections
+        collector.connection_closed(Some("route-a"), Some("1.2.3.4"));
+        assert_eq!(
+            collector.ip_connections.get("1.2.3.4").unwrap().load(Ordering::Relaxed),
+            1
+        );
+
+        // Close last connection for IP — should be cleaned up
+        collector.connection_closed(Some("route-a"), Some("1.2.3.4"));
+        assert!(collector.ip_connections.get("1.2.3.4").is_none());
+    }
+
+    #[test]
+    fn test_http_request_tracking() {
+        let collector = MetricsCollector::with_retention(60);
+
+        collector.record_http_request();
+        collector.record_http_request();
+        collector.record_http_request();
+
+        assert_eq!(collector.total_http_requests.load(Ordering::Relaxed), 3);
+
+        collector.sample_all();
+
+        let snapshot = collector.snapshot();
+        assert_eq!(snapshot.total_http_requests, 3);
+        assert_eq!(snapshot.http_requests_per_sec, 3);
+    }
+
+    #[test]
+    fn test_throughput_history_in_snapshot() {
+        let collector = MetricsCollector::with_retention(60);
+
+        for i in 1..=5 {
+            collector.record_bytes(i * 100, i * 200, None, None);
+            collector.sample_all();
+        }
+
+        let snapshot = collector.snapshot();
+        assert_eq!(snapshot.throughput_history.len(), 5);
+        // History should be chronological (oldest first)
+        assert_eq!(snapshot.throughput_history[0].bytes_in, 100);
+        assert_eq!(snapshot.throughput_history[4].bytes_in, 500);
+    }
 }

rust/crates/rustproxy-metrics/src/throughput.rs

@@ -1,8 +1,10 @@
+use serde::{Deserialize, Serialize};
 use std::sync::atomic::{AtomicU64, Ordering};
 use std::time::{Instant, SystemTime, UNIX_EPOCH};
 
 /// A single throughput sample.
-#[derive(Debug, Clone, Copy)]
+#[derive(Debug, Clone, Copy, Serialize, Deserialize)]
+#[serde(rename_all = "camelCase")]
 pub struct ThroughputSample {
     pub timestamp_ms: u64,
     pub bytes_in: u64,
@@ -106,6 +108,27 @@ impl ThroughputTracker {
         self.throughput(10)
     }
 
+    /// Return the last N samples in chronological order (oldest first).
+    pub fn history(&self, window_seconds: usize) -> Vec<ThroughputSample> {
+        let window = window_seconds.min(self.count);
+        if window == 0 {
+            return Vec::new();
+        }
+        let mut result = Vec::with_capacity(window);
+        for i in 0..window {
+            let idx = if self.write_index >= i + 1 {
+                self.write_index - i - 1
+            } else {
+                self.capacity - (i + 1 - self.write_index)
+            };
+            if idx < self.samples.len() {
+                result.push(self.samples[idx]);
+            }
+        }
+        result.reverse(); // Return oldest-first (chronological)
+        result
+    }
+
     /// How long this tracker has been alive.
     pub fn uptime(&self) -> std::time::Duration {
         self.created_at.elapsed()
@@ -170,4 +193,40 @@ mod tests {
         std::thread::sleep(std::time::Duration::from_millis(10));
         assert!(tracker.uptime().as_millis() >= 10);
     }
+
+    #[test]
+    fn test_history_returns_chronological() {
+        let mut tracker = ThroughputTracker::new(60);
+        for i in 1..=5 {
+            tracker.record_bytes(i * 100, i * 200);
+            tracker.sample();
+        }
+        let history = tracker.history(5);
+        assert_eq!(history.len(), 5);
+        // First sample should have 100 bytes_in, last should have 500
+        assert_eq!(history[0].bytes_in, 100);
+        assert_eq!(history[4].bytes_in, 500);
+    }
+
+    #[test]
+    fn test_history_wraps_around() {
+        let mut tracker = ThroughputTracker::new(3); // Small capacity
+        for i in 1..=5 {
+            tracker.record_bytes(i * 100, i * 200);
+            tracker.sample();
+        }
+        // Only last 3 should be retained
+        let history = tracker.history(10); // Ask for more than available
+        assert_eq!(history.len(), 3);
+        assert_eq!(history[0].bytes_in, 300);
+        assert_eq!(history[1].bytes_in, 400);
+        assert_eq!(history[2].bytes_in, 500);
+    }
+
+    #[test]
+    fn test_history_empty() {
+        let tracker = ThroughputTracker::new(60);
+        let history = tracker.history(10);
+        assert!(history.is_empty());
+    }
 }

rust/crates/rustproxy-passthrough/src/forwarder.rs

@@ -5,13 +5,14 @@ use std::sync::Arc;
 use std::sync::atomic::{AtomicU64, Ordering};
 use tracing::debug;
 
-use super::connection_record::ConnectionRecord;
+use rustproxy_metrics::MetricsCollector;
 
-/// Statistics for a forwarded connection.
-#[derive(Debug, Default)]
-pub struct ForwardStats {
-    pub bytes_in: AtomicU64,
-    pub bytes_out: AtomicU64,
+/// Context for forwarding metrics, replacing the growing tuple pattern.
+#[derive(Clone)]
+pub struct ForwardMetricsCtx {
+    pub collector: Arc<MetricsCollector>,
+    pub route_id: Option<String>,
+    pub source_ip: Option<String>,
 }
 
 /// Perform bidirectional TCP forwarding between client and backend.
@@ -68,6 +69,10 @@ pub async fn forward_bidirectional(
 
 /// Perform bidirectional TCP forwarding with inactivity and max lifetime timeouts.
 ///
+/// When `metrics` is provided, bytes are reported to the MetricsCollector
+/// per-chunk (lock-free) as they flow through the copy loops, enabling
+/// real-time throughput sampling for long-lived connections.
+///
 /// Returns (bytes_from_client, bytes_from_backend) when the connection closes or times out.
 pub async fn forward_bidirectional_with_timeouts(
     client: TcpStream,
@@ -76,10 +81,14 @@ pub async fn forward_bidirectional_with_timeouts(
     inactivity_timeout: std::time::Duration,
     max_lifetime: std::time::Duration,
     cancel: CancellationToken,
+    metrics: Option<ForwardMetricsCtx>,
 ) -> std::io::Result<(u64, u64)> {
     // Send initial data (peeked bytes) to backend
     if let Some(data) = initial_data {
         backend.write_all(data).await?;
+        if let Some(ref ctx) = metrics {
+            ctx.collector.record_bytes(data.len() as u64, 0, ctx.route_id.as_deref(), ctx.source_ip.as_deref());
+        }
     }
 
     let (mut client_read, mut client_write) = client.into_split();
@@ -90,6 +99,7 @@ pub async fn forward_bidirectional_with_timeouts(
 
     let la1 = Arc::clone(&last_activity);
     let initial_len = initial_data.map_or(0u64, |d| d.len() as u64);
+    let metrics_c2b = metrics.clone();
     let c2b = tokio::spawn(async move {
         let mut buf = vec![0u8; 65536];
         let mut total = initial_len;
@@ -103,12 +113,16 @@ pub async fn forward_bidirectional_with_timeouts(
             }
             total += n as u64;
             la1.store(start.elapsed().as_millis() as u64, Ordering::Relaxed);
+            if let Some(ref ctx) = metrics_c2b {
+                ctx.collector.record_bytes(n as u64, 0, ctx.route_id.as_deref(), ctx.source_ip.as_deref());
+            }
         }
         let _ = backend_write.shutdown().await;
         total
     });
 
     let la2 = Arc::clone(&last_activity);
+    let metrics_b2c = metrics;
     let b2c = tokio::spawn(async move {
         let mut buf = vec![0u8; 65536];
         let mut total = 0u64;
@@ -122,6 +136,9 @@ pub async fn forward_bidirectional_with_timeouts(
             }
             total += n as u64;
             la2.store(start.elapsed().as_millis() as u64, Ordering::Relaxed);
+            if let Some(ref ctx) = metrics_b2c {
+                ctx.collector.record_bytes(0, n as u64, ctx.route_id.as_deref(), ctx.source_ip.as_deref());
+            }
         }
         let _ = client_write.shutdown().await;
         total
@@ -173,153 +190,3 @@ pub async fn forward_bidirectional_with_timeouts(
     watchdog.abort();
     Ok((bytes_in, bytes_out))
 }
-
-/// Forward bidirectional with a callback for byte counting.
-pub async fn forward_bidirectional_with_stats(
-    client: TcpStream,
-    backend: TcpStream,
-    initial_data: Option<&[u8]>,
-    stats: Arc<ForwardStats>,
-) -> std::io::Result<()> {
-    let (bytes_in, bytes_out) = forward_bidirectional(client, backend, initial_data).await?;
-    stats.bytes_in.fetch_add(bytes_in, Ordering::Relaxed);
-    stats.bytes_out.fetch_add(bytes_out, Ordering::Relaxed);
-    Ok(())
-}
-
-/// Perform bidirectional TCP forwarding with inactivity / lifetime timeouts,
-/// updating a `ConnectionRecord` with byte counts and activity timestamps
-/// in real time for zombie detection.
-///
-/// When `record` is `None`, this behaves identically to
-/// `forward_bidirectional_with_timeouts`.
-///
-/// The record's `client_closed` / `backend_closed` flags are set when the
-/// respective copy loop terminates, giving the zombie scanner visibility
-/// into half-open connections.
-pub async fn forward_bidirectional_with_record(
-    client: TcpStream,
-    mut backend: TcpStream,
-    initial_data: Option<&[u8]>,
-    inactivity_timeout: std::time::Duration,
-    max_lifetime: std::time::Duration,
-    cancel: CancellationToken,
-    record: Option<Arc<ConnectionRecord>>,
-) -> std::io::Result<(u64, u64)> {
-    // Send initial data (peeked bytes) to backend
-    if let Some(data) = initial_data {
-        backend.write_all(data).await?;
-        if let Some(ref r) = record {
-            r.record_bytes_in(data.len() as u64);
-        }
-    }
-
-    let (mut client_read, mut client_write) = client.into_split();
-    let (mut backend_read, mut backend_write) = backend.into_split();
-
-    let last_activity = Arc::new(AtomicU64::new(0));
-    let start = std::time::Instant::now();
-
-    let la1 = Arc::clone(&last_activity);
-    let initial_len = initial_data.map_or(0u64, |d| d.len() as u64);
-    let rec1 = record.clone();
-    let c2b = tokio::spawn(async move {
-        let mut buf = vec![0u8; 65536];
-        let mut total = initial_len;
-        loop {
-            let n = match client_read.read(&mut buf).await {
-                Ok(0) | Err(_) => break,
-                Ok(n) => n,
-            };
-            if backend_write.write_all(&buf[..n]).await.is_err() {
-                break;
-            }
-            total += n as u64;
-            let now_ms = start.elapsed().as_millis() as u64;
-            la1.store(now_ms, Ordering::Relaxed);
-            if let Some(ref r) = rec1 {
-                r.record_bytes_in(n as u64);
-            }
-        }
-        let _ = backend_write.shutdown().await;
-        // Mark client side as closed
-        if let Some(ref r) = rec1 {
-            r.client_closed.store(true, Ordering::Relaxed);
-        }
-        total
-    });
-
-    let la2 = Arc::clone(&last_activity);
-    let rec2 = record.clone();
-    let b2c = tokio::spawn(async move {
-        let mut buf = vec![0u8; 65536];
-        let mut total = 0u64;
-        loop {
-            let n = match backend_read.read(&mut buf).await {
-                Ok(0) | Err(_) => break,
-                Ok(n) => n,
-            };
-            if client_write.write_all(&buf[..n]).await.is_err() {
-                break;
-            }
-            total += n as u64;
-            let now_ms = start.elapsed().as_millis() as u64;
-            la2.store(now_ms, Ordering::Relaxed);
-            if let Some(ref r) = rec2 {
-                r.record_bytes_out(n as u64);
-            }
-        }
-        let _ = client_write.shutdown().await;
-        // Mark backend side as closed
-        if let Some(ref r) = rec2 {
-            r.backend_closed.store(true, Ordering::Relaxed);
-        }
-        total
-    });
-
-    // Watchdog: inactivity, max lifetime, and cancellation
-    let la_watch = Arc::clone(&last_activity);
-    let c2b_handle = c2b.abort_handle();
-    let b2c_handle = b2c.abort_handle();
-    let watchdog = tokio::spawn(async move {
-        let check_interval = std::time::Duration::from_secs(5);
-        let mut last_seen = 0u64;
-        loop {
-            tokio::select! {
-                _ = cancel.cancelled() => {
-                    debug!("Connection cancelled by shutdown");
-                    c2b_handle.abort();
-                    b2c_handle.abort();
-                    break;
-                }
-                _ = tokio::time::sleep(check_interval) => {
-                    // Check max lifetime
-                    if start.elapsed() >= max_lifetime {
-                        debug!("Connection exceeded max lifetime, closing");
-                        c2b_handle.abort();
-                        b2c_handle.abort();
-                        break;
-                    }
-
-                    // Check inactivity
-                    let current = la_watch.load(Ordering::Relaxed);
-                    if current == last_seen {
-                        let elapsed_since_activity = start.elapsed().as_millis() as u64 - current;
-                        if elapsed_since_activity >= inactivity_timeout.as_millis() as u64 {
-                            debug!("Connection inactive for {}ms, closing", elapsed_since_activity);
-                            c2b_handle.abort();
-                            b2c_handle.abort();
-                            break;
-                        }
-                    }
-                    last_seen = current;
-                }
-            }
-        }
-    });
-
-    let bytes_in = c2b.await.unwrap_or(0);
-    let bytes_out = b2c.await.unwrap_or(0);
-    watchdog.abort();
-    Ok((bytes_in, bytes_out))
-}

rust/crates/rustproxy-passthrough/src/tcp_listener.rs

@@ -20,14 +20,16 @@ use crate::connection_tracker::ConnectionTracker;
 struct ConnectionGuard {
     metrics: Arc<MetricsCollector>,
     route_id: Option<String>,
+    source_ip: Option<String>,
     disarmed: bool,
 }
 
 impl ConnectionGuard {
-    fn new(metrics: Arc<MetricsCollector>, route_id: Option<&str>) -> Self {
+    fn new(metrics: Arc<MetricsCollector>, route_id: Option<&str>, source_ip: Option<&str>) -> Self {
         Self {
             metrics,
             route_id: route_id.map(|s| s.to_string()),
+            source_ip: source_ip.map(|s| s.to_string()),
             disarmed: false,
         }
     }
@@ -42,7 +44,7 @@ impl ConnectionGuard {
 impl Drop for ConnectionGuard {
     fn drop(&mut self) {
         if !self.disarmed {
-            self.metrics.connection_closed(self.route_id.as_deref());
+            self.metrics.connection_closed(self.route_id.as_deref(), self.source_ip.as_deref());
         }
     }
 }
@@ -395,6 +397,9 @@ impl TcpListenerManager {
 
         stream.set_nodelay(true)?;
 
+        // Extract source IP once for all metric calls
+        let ip_str = peer_addr.ip().to_string();
+
         // === Fast path: try port-only matching before peeking at data ===
         // This handles "server-speaks-first" protocols where the client
         // doesn't send initial data (e.g., SMTP, greeting-based protocols).
@@ -460,8 +465,8 @@ impl TcpListenerManager {
                             }
                         }
 
-                        metrics.connection_opened(route_id);
-                        let _fast_guard = ConnectionGuard::new(Arc::clone(&metrics), route_id);
+                        metrics.connection_opened(route_id, Some(&ip_str));
+                        let _fast_guard = ConnectionGuard::new(Arc::clone(&metrics), route_id, Some(&ip_str));
 
                         let connect_timeout = std::time::Duration::from_millis(conn_config.connection_timeout_ms);
                         let inactivity_timeout = std::time::Duration::from_millis(conn_config.socket_timeout_ms);
@@ -496,17 +501,25 @@ impl TcpListenerManager {
                             let mut backend_w = backend;
                             backend_w.write_all(header.as_bytes()).await?;
 
-                            let (bytes_in, bytes_out) = forwarder::forward_bidirectional_with_timeouts(
+                            let (_bytes_in, _bytes_out) = forwarder::forward_bidirectional_with_timeouts(
                                 stream, backend_w, None,
                                 inactivity_timeout, max_lifetime, cancel,
+                                Some(forwarder::ForwardMetricsCtx {
+                                    collector: Arc::clone(&metrics),
+                                    route_id: route_id.map(|s| s.to_string()),
+                                    source_ip: Some(ip_str.clone()),
+                                }),
                             ).await?;
-                            metrics.record_bytes(bytes_in, bytes_out, route_id);
                         } else {
-                            let (bytes_in, bytes_out) = forwarder::forward_bidirectional_with_timeouts(
+                            let (_bytes_in, _bytes_out) = forwarder::forward_bidirectional_with_timeouts(
                                 stream, backend, None,
                                 inactivity_timeout, max_lifetime, cancel,
+                                Some(forwarder::ForwardMetricsCtx {
+                                    collector: Arc::clone(&metrics),
+                                    route_id: route_id.map(|s| s.to_string()),
+                                    source_ip: Some(ip_str.clone()),
+                                }),
                             ).await?;
-                            metrics.record_bytes(bytes_in, bytes_out, route_id);
                         }
 
                         return Ok(());
@@ -646,8 +659,8 @@ impl TcpListenerManager {
         }
 
         // Track connection in metrics — guard ensures connection_closed on all exit paths
-        metrics.connection_opened(route_id);
-        let _conn_guard = ConnectionGuard::new(Arc::clone(&metrics), route_id);
+        metrics.connection_opened(route_id, Some(&ip_str));
+        let _conn_guard = ConnectionGuard::new(Arc::clone(&metrics), route_id, Some(&ip_str));
 
         // Check if this is a socket-handler route that should be relayed to TypeScript
         if route_match.route.action.action_type == RouteActionType::SocketHandler {
@@ -661,6 +674,7 @@ impl TcpListenerManager {
                     stream, n, port, peer_addr,
                     &route_match, domain.as_deref(), is_tls,
                     &relay_socket_path,
+                    &metrics, route_id,
                 ).await;
             } else {
                 debug!("Socket-handler route matched but no relay path configured");
@@ -751,11 +765,15 @@ impl TcpListenerManager {
                 let mut actual_buf = vec![0u8; n];
                 stream.read_exact(&mut actual_buf).await?;
 
-                let (bytes_in, bytes_out) = forwarder::forward_bidirectional_with_timeouts(
+                let (_bytes_in, _bytes_out) = forwarder::forward_bidirectional_with_timeouts(
                     stream, backend, Some(&actual_buf),
                     inactivity_timeout, max_lifetime, cancel,
+                    Some(forwarder::ForwardMetricsCtx {
+                        collector: Arc::clone(&metrics),
+                        route_id: route_id.map(|s| s.to_string()),
+                        source_ip: Some(ip_str.clone()),
+                    }),
                 ).await?;
-                metrics.record_bytes(bytes_in, bytes_out, route_id);
                 Ok(())
             }
             Some(rustproxy_config::TlsMode::Terminate) => {
@@ -812,12 +830,15 @@ impl TcpListenerManager {
                     let (tls_read, tls_write) = tokio::io::split(buf_stream);
                     let (backend_read, backend_write) = tokio::io::split(backend);
 
-                    let (bytes_in, bytes_out) = Self::forward_bidirectional_split_with_timeouts(
+                    let (_bytes_in, _bytes_out) = Self::forward_bidirectional_split_with_timeouts(
                         tls_read, tls_write, backend_read, backend_write,
                         inactivity_timeout, max_lifetime,
+                        Some(forwarder::ForwardMetricsCtx {
+                            collector: Arc::clone(&metrics),
+                            route_id: route_id.map(|s| s.to_string()),
+                            source_ip: Some(ip_str.clone()),
+                        }),
                     ).await;
-
-                    metrics.record_bytes(bytes_in, bytes_out, route_id);
                 }
                 Ok(())
             }
@@ -825,7 +846,7 @@ impl TcpListenerManager {
                 let route_tls = route_match.route.action.tls.as_ref();
                 Self::handle_tls_terminate_reencrypt(
                     stream, n, &domain, &target_host, target_port,
-                    peer_addr, &tls_configs, &metrics, route_id, &conn_config, route_tls,
+                    peer_addr, &tls_configs, Arc::clone(&metrics), route_id, &conn_config, route_tls,
                 ).await
             }
             None => {
@@ -862,11 +883,15 @@ impl TcpListenerManager {
                     let mut actual_buf = vec![0u8; n];
                     stream.read_exact(&mut actual_buf).await?;
 
-                    let (bytes_in, bytes_out) = forwarder::forward_bidirectional_with_timeouts(
+                    let (_bytes_in, _bytes_out) = forwarder::forward_bidirectional_with_timeouts(
                         stream, backend, Some(&actual_buf),
                         inactivity_timeout, max_lifetime, cancel,
+                        Some(forwarder::ForwardMetricsCtx {
+                            collector: Arc::clone(&metrics),
+                            route_id: route_id.map(|s| s.to_string()),
+                            source_ip: Some(ip_str.clone()),
+                        }),
                     ).await?;
-                    metrics.record_bytes(bytes_in, bytes_out, route_id);
                     Ok(())
                 }
             }
@@ -892,6 +917,8 @@ impl TcpListenerManager {
         domain: Option<&str>,
         is_tls: bool,
         relay_path: &str,
+        metrics: &MetricsCollector,
+        route_id: Option<&str>,
     ) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
         use tokio::io::{AsyncReadExt, AsyncWriteExt};
         use tokio::net::UnixStream;
@@ -932,12 +959,22 @@ impl TcpListenerManager {
         unix_stream.write_all(&initial_buf).await?;
 
         // Bidirectional relay between TCP client and Unix socket handler
+        let initial_len = initial_buf.len() as u64;
         match tokio::io::copy_bidirectional(&mut stream, &mut unix_stream).await {
             Ok((c2s, s2c)) => {
+                // Include initial data bytes that were forwarded before copy_bidirectional
+                let total_in = c2s + initial_len;
                 debug!("Socket handler relay complete for {}: {} bytes in, {} bytes out",
-                    route_key, c2s, s2c);
+                    route_key, total_in, s2c);
+                let ip = peer_addr.ip().to_string();
+                metrics.record_bytes(total_in, s2c, route_id, Some(&ip));
             }
             Err(e) => {
+                // Still record the initial data even on error
+                if initial_len > 0 {
+                    let ip = peer_addr.ip().to_string();
+                    metrics.record_bytes(initial_len, 0, route_id, Some(&ip));
+                }
                 debug!("Socket handler relay ended for {}: {}", route_key, e);
             }
         }
@@ -954,7 +991,7 @@ impl TcpListenerManager {
         target_port: u16,
         peer_addr: std::net::SocketAddr,
         tls_configs: &HashMap<String, TlsCertConfig>,
-        metrics: &MetricsCollector,
+        metrics: Arc<MetricsCollector>,
         route_id: Option<&str>,
         conn_config: &ConnectionConfig,
         route_tls: Option<&rustproxy_config::RouteTls>,
@@ -1019,12 +1056,16 @@ impl TcpListenerManager {
             }
         };
 
-        let (bytes_in, bytes_out) = Self::forward_bidirectional_split_with_timeouts(
+        let (_bytes_in, _bytes_out) = Self::forward_bidirectional_split_with_timeouts(
             client_read, client_write, backend_read, backend_write,
             inactivity_timeout, max_lifetime,
+            Some(forwarder::ForwardMetricsCtx {
+                collector: metrics,
+                route_id: route_id.map(|s| s.to_string()),
+                source_ip: Some(peer_addr.ip().to_string()),
+            }),
         ).await;
 
-        metrics.record_bytes(bytes_in, bytes_out, route_id);
         Ok(())
     }
 
@@ -1058,6 +1099,9 @@ impl TcpListenerManager {
     }
 
     /// Forward bidirectional between two split streams with inactivity and lifetime timeouts.
+    ///
+    /// When `metrics` is provided, bytes are reported per-chunk (lock-free) for
+    /// real-time throughput measurement.
     async fn forward_bidirectional_split_with_timeouts<R1, W1, R2, W2>(
         mut client_read: R1,
         mut client_write: W1,
@@ -1065,6 +1109,7 @@ impl TcpListenerManager {
         mut backend_write: W2,
         inactivity_timeout: std::time::Duration,
         max_lifetime: std::time::Duration,
+        metrics: Option<forwarder::ForwardMetricsCtx>,
     ) -> (u64, u64)
     where
         R1: tokio::io::AsyncRead + Unpin + Send + 'static,
@@ -1080,6 +1125,7 @@ impl TcpListenerManager {
         let start = std::time::Instant::now();
 
         let la1 = Arc::clone(&last_activity);
+        let metrics_c2b = metrics.clone();
         let c2b = tokio::spawn(async move {
             let mut buf = vec![0u8; 65536];
             let mut total = 0u64;
@@ -1096,12 +1142,16 @@ impl TcpListenerManager {
                     start.elapsed().as_millis() as u64,
                     Ordering::Relaxed,
                 );
+                if let Some(ref ctx) = metrics_c2b {
+                    ctx.collector.record_bytes(n as u64, 0, ctx.route_id.as_deref(), ctx.source_ip.as_deref());
+                }
             }
             let _ = backend_write.shutdown().await;
             total
         });
 
         let la2 = Arc::clone(&last_activity);
+        let metrics_b2c = metrics;
         let b2c = tokio::spawn(async move {
             let mut buf = vec![0u8; 65536];
             let mut total = 0u64;
@@ -1118,6 +1168,9 @@ impl TcpListenerManager {
                     start.elapsed().as_millis() as u64,
                     Ordering::Relaxed,
                 );
+                if let Some(ref ctx) = metrics_b2c {
+                    ctx.collector.record_bytes(0, n as u64, ctx.route_id.as_deref(), ctx.source_ip.as_deref());
+                }
             }
             let _ = client_write.shutdown().await;
             total

rust/crates/rustproxy/src/lib.rs

@@ -71,6 +71,7 @@ pub struct RustProxy {
     cert_manager: Option<Arc<tokio::sync::Mutex<CertManager>>>,
     challenge_server: Option<challenge_server::ChallengeServer>,
     renewal_handle: Option<tokio::task::JoinHandle<()>>,
+    sampling_handle: Option<tokio::task::JoinHandle<()>>,
     nft_manager: Option<NftManager>,
     started: bool,
     started_at: Option<Instant>,
@@ -100,14 +101,19 @@ impl RustProxy {
         let cert_manager = Self::build_cert_manager(&options)
             .map(|cm| Arc::new(tokio::sync::Mutex::new(cm)));
 
+        let retention = options.metrics.as_ref()
+            .and_then(|m| m.retention_seconds)
+            .unwrap_or(3600) as usize;
+
         Ok(Self {
             options,
             route_table: ArcSwap::from(Arc::new(route_manager)),
             listener_manager: None,
-            metrics: Arc::new(MetricsCollector::new()),
+            metrics: Arc::new(MetricsCollector::with_retention(retention)),
             cert_manager,
             challenge_server: None,
             renewal_handle: None,
+            sampling_handle: None,
             nft_manager: None,
             started: false,
             started_at: None,
@@ -276,6 +282,21 @@ impl RustProxy {
         self.started = true;
         self.started_at = Some(Instant::now());
 
+        // Start the throughput sampling task
+        let metrics = Arc::clone(&self.metrics);
+        let interval_ms = self.options.metrics.as_ref()
+            .and_then(|m| m.sample_interval_ms)
+            .unwrap_or(1000);
+        self.sampling_handle = Some(tokio::spawn(async move {
+            let mut interval = tokio::time::interval(
+                std::time::Duration::from_millis(interval_ms)
+            );
+            loop {
+                interval.tick().await;
+                metrics.sample_all();
+            }
+        }));
+
         // Apply NFTables rules for routes using nftables forwarding engine
         self.apply_nftables_rules(&self.options.routes.clone()).await;
 
@@ -478,6 +499,11 @@ impl RustProxy {
 
         info!("Stopping RustProxy...");
 
+        // Stop sampling task
+        if let Some(handle) = self.sampling_handle.take() {
+            handle.abort();
+        }
+
         // Stop renewal timer
         if let Some(handle) = self.renewal_handle.take() {
             handle.abort();

test/test.throughput.ts

@@ -0,0 +1,699 @@
+import { expect, tap } from '@git.zone/tstest/tapbundle';
+import { SmartProxy } from '../ts/index.js';
+import type { IRouteConfig } from '../ts/index.js';
+import * as net from 'net';
+import * as http from 'http';
+import * as tls from 'tls';
+import * as https from 'https';
+import * as fs from 'fs';
+import * as path from 'path';
+import { fileURLToPath } from 'url';
+
+const __filename = fileURLToPath(import.meta.url);
+const __dirname = path.dirname(__filename);
+
+// ────────────────────────────────────────────────────────────────────────────
+// Port assignments (unique to avoid conflicts with other tests)
+// ────────────────────────────────────────────────────────────────────────────
+const TCP_ECHO_PORT = 47500;
+const HTTP_ECHO_PORT = 47501;
+const TLS_ECHO_PORT = 47502;
+const PROXY_TCP_PORT = 47510;
+const PROXY_HTTP_PORT = 47511;
+const PROXY_TLS_PASS_PORT = 47512;
+const PROXY_TLS_TERM_PORT = 47513;
+const PROXY_SOCKET_PORT = 47514;
+const PROXY_MULTI_A_PORT = 47515;
+const PROXY_MULTI_B_PORT = 47516;
+const PROXY_TP_HTTP_PORT = 47517;
+
+// ────────────────────────────────────────────────────────────────────────────
+// Test certificates
+// ────────────────────────────────────────────────────────────────────────────
+const CERT_PEM = fs.readFileSync(path.join(__dirname, '..', 'assets', 'certs', 'cert.pem'), 'utf8');
+const KEY_PEM = fs.readFileSync(path.join(__dirname, '..', 'assets', 'certs', 'key.pem'), 'utf8');
+
+// ────────────────────────────────────────────────────────────────────────────
+// Backend servers
+// ────────────────────────────────────────────────────────────────────────────
+let tcpEchoServer: net.Server;
+let httpEchoServer: http.Server;
+let tlsEchoServer: tls.Server;
+
+// Helper: force-poll the metrics adapter
+async function pollMetrics(proxy: SmartProxy): Promise<void> {
+  await (proxy as any).metricsAdapter.poll();
+}
+
+// ════════════════════════════════════════════════════════════════════════════
+// Setup: backend servers
+// ════════════════════════════════════════════════════════════════════════════
+tap.test('setup - TCP echo server', async () => {
+  tcpEchoServer = net.createServer((socket) => {
+    socket.on('data', (data) => socket.write(data));
+    socket.on('error', () => {});
+  });
+  await new Promise<void>((resolve) => {
+    tcpEchoServer.listen(TCP_ECHO_PORT, () => {
+      console.log(`TCP echo server on port ${TCP_ECHO_PORT}`);
+      resolve();
+    });
+  });
+});
+
+tap.test('setup - HTTP echo server', async () => {
+  httpEchoServer = http.createServer((req, res) => {
+    let body = '';
+    req.on('data', (chunk) => (body += chunk));
+    req.on('end', () => {
+      res.writeHead(200, { 'Content-Type': 'text/plain' });
+      res.end(`echo:${body}`);
+    });
+  });
+  await new Promise<void>((resolve) => {
+    httpEchoServer.listen(HTTP_ECHO_PORT, () => {
+      console.log(`HTTP echo server on port ${HTTP_ECHO_PORT}`);
+      resolve();
+    });
+  });
+});
+
+tap.test('setup - TLS echo server', async () => {
+  tlsEchoServer = tls.createServer(
+    { cert: CERT_PEM, key: KEY_PEM },
+    (socket) => {
+      socket.on('data', (data) => socket.write(data));
+      socket.on('error', () => {});
+    },
+  );
+  await new Promise<void>((resolve) => {
+    tlsEchoServer.listen(TLS_ECHO_PORT, () => {
+      console.log(`TLS echo server on port ${TLS_ECHO_PORT}`);
+      resolve();
+    });
+  });
+});
+
+// ════════════════════════════════════════════════════════════════════════════
+// Group 1: TCP Forward (plain TCP passthrough — no domain, no TLS)
+// ════════════════════════════════════════════════════════════════════════════
+tap.test('TCP forward - real-time byte tracking', async (tools) => {
+  const proxy = new SmartProxy({
+    routes: [
+      {
+        id: 'tcp-forward',
+        name: 'tcp-forward',
+        match: { ports: PROXY_TCP_PORT },
+        action: {
+          type: 'forward',
+          targets: [{ host: 'localhost', port: TCP_ECHO_PORT }],
+        },
+      },
+    ],
+    metrics: { enabled: true, sampleIntervalMs: 100, retentionSeconds: 60 },
+  });
+  await proxy.start();
+
+  // Connect and send data
+  const client = new net.Socket();
+  await new Promise<void>((resolve, reject) => {
+    client.connect(PROXY_TCP_PORT, 'localhost', () => resolve());
+    client.on('error', reject);
+  });
+
+  let received = 0;
+  client.on('data', (data) => (received += data.length));
+
+  // Send 10 KB in chunks over 1 second
+  const chunk = Buffer.alloc(1024, 'A');
+  for (let i = 0; i < 10; i++) {
+    client.write(chunk);
+    await tools.delayFor(100);
+  }
+
+  // Wait for echo data and sampling to accumulate
+  await tools.delayFor(500);
+
+  // === Key assertion: metrics visible WHILE the connection is still open ===
+  // Before this change, TCP bytes were only reported after connection close.
+  // Now bytes are reported per-chunk in real-time.
+  await pollMetrics(proxy);
+
+  const mDuring = proxy.getMetrics();
+  const bytesInDuring = mDuring.totals.bytesIn();
+  const bytesOutDuring = mDuring.totals.bytesOut();
+  console.log(`TCP forward (during) — bytesIn: ${bytesInDuring}, bytesOut: ${bytesOutDuring}`);
+  expect(bytesInDuring).toBeGreaterThan(0);
+  expect(bytesOutDuring).toBeGreaterThan(0);
+
+  // Check that throughput is non-zero during active TCP traffic
+  const tpDuring = mDuring.throughput.recent();
+  console.log(`TCP forward (during) — recent throughput: in=${tpDuring.in}, out=${tpDuring.out}`);
+  expect(tpDuring.in + tpDuring.out).toBeGreaterThan(0);
+
+  // Close connection
+  client.destroy();
+  await tools.delayFor(500);
+
+  // Final check
+  await pollMetrics(proxy);
+  const m = proxy.getMetrics();
+  const bytesIn = m.totals.bytesIn();
+  const bytesOut = m.totals.bytesOut();
+  console.log(`TCP forward (final) — bytesIn: ${bytesIn}, bytesOut: ${bytesOut}`);
+  expect(bytesIn).toBeGreaterThanOrEqual(bytesInDuring);
+  expect(bytesOut).toBeGreaterThanOrEqual(bytesOutDuring);
+
+  // Check per-route tracking
+  const byRoute = m.throughput.byRoute();
+  console.log('TCP forward — throughput byRoute:', Array.from(byRoute.entries()));
+
+  // ── v25.2.0: Per-IP tracking (TCP connections) ──
+  const byIP = m.connections.byIP();
+  console.log('TCP forward — connections byIP:', Array.from(byIP.entries()));
+  expect(byIP.size).toBeGreaterThan(0);
+
+  const topIPs = m.connections.topIPs(10);
+  console.log('TCP forward — topIPs:', topIPs);
+  expect(topIPs.length).toBeGreaterThan(0);
+  expect(topIPs[0].ip).toBeTruthy();
+
+  // ── v25.2.0: Throughput history ──
+  const history = m.throughput.history(10);
+  console.log('TCP forward — throughput history length:', history.length);
+  expect(history.length).toBeGreaterThan(0);
+  expect(history[0].timestamp).toBeGreaterThan(0);
+
+  await proxy.stop();
+  await tools.delayFor(200);
+});
+
+// ════════════════════════════════════════════════════════════════════════════
+// Group 2: HTTP Forward (plain HTTP proxy)
+// ════════════════════════════════════════════════════════════════════════════
+tap.test('HTTP forward - byte totals tracking', async (tools) => {
+  const proxy = new SmartProxy({
+    routes: [
+      {
+        id: 'http-forward',
+        name: 'http-forward',
+        match: { ports: PROXY_HTTP_PORT },
+        action: {
+          type: 'forward',
+          targets: [{ host: 'localhost', port: HTTP_ECHO_PORT }],
+        },
+      },
+    ],
+    metrics: { enabled: true, sampleIntervalMs: 100, retentionSeconds: 60 },
+  });
+  await proxy.start();
+  await tools.delayFor(300);
+
+  // Send 10 HTTP requests with 1 KB body each
+  for (let i = 0; i < 10; i++) {
+    const body = 'X'.repeat(1024);
+    await new Promise<void>((resolve, reject) => {
+      const req = http.request(
+        {
+          hostname: 'localhost',
+          port: PROXY_HTTP_PORT,
+          path: '/echo',
+          method: 'POST',
+          headers: { 'Content-Type': 'text/plain', 'Content-Length': String(body.length) },
+        },
+        (res) => {
+          let data = '';
+          res.on('data', (chunk) => (data += chunk));
+          res.on('end', () => resolve());
+        },
+      );
+      req.on('error', reject);
+      req.setTimeout(5000, () => {
+        req.destroy();
+        reject(new Error('HTTP request timeout'));
+      });
+      req.end(body);
+    });
+  }
+
+  // Wait for sampling + poll
+  await tools.delayFor(500);
+  await pollMetrics(proxy);
+
+  const m = proxy.getMetrics();
+  const bytesIn = m.totals.bytesIn();
+  const bytesOut = m.totals.bytesOut();
+  console.log(`HTTP forward — bytesIn: ${bytesIn}, bytesOut: ${bytesOut}`);
+
+  // Both directions should have bytes (CountingBody tracks request + response)
+  expect(bytesIn).toBeGreaterThan(0);
+  expect(bytesOut).toBeGreaterThan(0);
+
+  // ── v25.2.0: Per-IP tracking (HTTP connections) ──
+  const byIP = m.connections.byIP();
+  console.log('HTTP forward — connections byIP:', Array.from(byIP.entries()));
+  expect(byIP.size).toBeGreaterThan(0);
+
+  const topIPs = m.connections.topIPs(10);
+  console.log('HTTP forward — topIPs:', topIPs);
+  expect(topIPs.length).toBeGreaterThan(0);
+  expect(topIPs[0].ip).toBeTruthy();
+
+  // ── v25.2.0: HTTP request counting ──
+  const totalReqs = m.requests.total();
+  const rps = m.requests.perSecond();
+  console.log(`HTTP forward — requests total: ${totalReqs}, perSecond: ${rps}`);
+  expect(totalReqs).toBeGreaterThan(0);
+
+  await proxy.stop();
+  await tools.delayFor(200);
+});
+
+// ════════════════════════════════════════════════════════════════════════════
+// Group 3: TLS Passthrough (SNI-based, Rust passes encrypted data through)
+// ════════════════════════════════════════════════════════════════════════════
+tap.test('TLS passthrough - byte totals tracking', async (tools) => {
+  const proxy = new SmartProxy({
+    routes: [
+      {
+        id: 'tls-passthrough',
+        name: 'tls-passthrough',
+        match: { ports: PROXY_TLS_PASS_PORT, domains: 'localhost' },
+        action: {
+          type: 'forward',
+          tls: { mode: 'passthrough' },
+          targets: [{ host: 'localhost', port: TLS_ECHO_PORT }],
+        },
+      },
+    ],
+    metrics: { enabled: true, sampleIntervalMs: 100, retentionSeconds: 60 },
+  });
+  await proxy.start();
+  await tools.delayFor(300);
+
+  // Connect via TLS through the proxy (SNI: localhost)
+  const tlsClient = tls.connect(
+    {
+      host: 'localhost',
+      port: PROXY_TLS_PASS_PORT,
+      servername: 'localhost',
+      rejectUnauthorized: false,
+    },
+  );
+
+  await new Promise<void>((resolve, reject) => {
+    tlsClient.on('secureConnect', () => resolve());
+    tlsClient.on('error', reject);
+  });
+
+  // Send some data
+  const data = Buffer.alloc(2048, 'B');
+  tlsClient.write(data);
+
+  // Wait for echo
+  let received = 0;
+  tlsClient.on('data', (chunk) => (received += chunk.length));
+  await tools.delayFor(1000);
+
+  console.log(`TLS passthrough — received ${received} bytes back`);
+  expect(received).toBeGreaterThan(0);
+
+  tlsClient.destroy();
+  await tools.delayFor(500);
+
+  await pollMetrics(proxy);
+
+  const m = proxy.getMetrics();
+  const bytesIn = m.totals.bytesIn();
+  const bytesOut = m.totals.bytesOut();
+  console.log(`TLS passthrough — bytesIn: ${bytesIn}, bytesOut: ${bytesOut}`);
+
+  // TLS passthrough tracks encrypted bytes flowing through
+  expect(bytesIn).toBeGreaterThan(0);
+  expect(bytesOut).toBeGreaterThan(0);
+
+  await proxy.stop();
+  await tools.delayFor(200);
+});
+
+// ════════════════════════════════════════════════════════════════════════════
+// Group 4: TLS Terminate + HTTP (Rust terminates TLS, forwards to HTTP backend)
+// ════════════════════════════════════════════════════════════════════════════
+tap.test('TLS terminate + HTTP forward - byte totals tracking', async (tools) => {
+  const proxy = new SmartProxy({
+    routes: [
+      {
+        id: 'tls-terminate',
+        name: 'tls-terminate',
+        match: { ports: PROXY_TLS_TERM_PORT, domains: 'localhost' },
+        action: {
+          type: 'forward',
+          tls: {
+            mode: 'terminate',
+            certificate: {
+              cert: CERT_PEM,
+              key: KEY_PEM,
+            },
+          },
+          targets: [{ host: 'localhost', port: HTTP_ECHO_PORT }],
+        },
+      },
+    ],
+    metrics: { enabled: true, sampleIntervalMs: 100, retentionSeconds: 60 },
+    disableDefaultCert: true,
+  });
+  await proxy.start();
+  await tools.delayFor(300);
+
+  // Send HTTPS request through the proxy
+  const body = 'Z'.repeat(2048);
+  await new Promise<void>((resolve, reject) => {
+    const req = https.request(
+      {
+        hostname: 'localhost',
+        port: PROXY_TLS_TERM_PORT,
+        path: '/echo',
+        method: 'POST',
+        headers: { 'Content-Type': 'text/plain', 'Content-Length': String(body.length) },
+        rejectUnauthorized: false,
+      },
+      (res) => {
+        let data = '';
+        res.on('data', (chunk) => (data += chunk));
+        res.on('end', () => {
+          console.log(`TLS terminate — response: ${data.slice(0, 50)}...`);
+          resolve();
+        });
+      },
+    );
+    req.on('error', reject);
+    req.setTimeout(5000, () => {
+      req.destroy();
+      reject(new Error('HTTPS request timeout'));
+    });
+    req.end(body);
+  });
+
+  await tools.delayFor(500);
+  await pollMetrics(proxy);
+
+  const m = proxy.getMetrics();
+  const bytesIn = m.totals.bytesIn();
+  const bytesOut = m.totals.bytesOut();
+  console.log(`TLS terminate — bytesIn: ${bytesIn}, bytesOut: ${bytesOut}`);
+
+  // TLS terminate: request body (bytesIn) and response body (bytesOut) via CountingBody
+  expect(bytesIn).toBeGreaterThan(0);
+  expect(bytesOut).toBeGreaterThan(0);
+
+  await proxy.stop();
+  await tools.delayFor(200);
+});
+
+// ════════════════════════════════════════════════════════════════════════════
+// Group 5: Socket Handler (JS callback handling)
+// ════════════════════════════════════════════════════════════════════════════
+tap.test('Socket handler - byte totals tracking', async (tools) => {
+  const proxy = new SmartProxy({
+    routes: [
+      {
+        id: 'socket-handler',
+        name: 'socket-handler',
+        match: { ports: PROXY_SOCKET_PORT },
+        action: {
+          type: 'socket-handler',
+          socketHandler: (socket, _context) => {
+            socket.on('data', (data) => socket.write(data)); // echo
+            socket.on('error', () => {});
+          },
+        },
+      },
+    ],
+    metrics: { enabled: true, sampleIntervalMs: 100, retentionSeconds: 60 },
+  });
+  await proxy.start();
+  await tools.delayFor(300);
+
+  // Connect and send data
+  const client = new net.Socket();
+  await new Promise<void>((resolve, reject) => {
+    client.connect(PROXY_SOCKET_PORT, 'localhost', () => resolve());
+    client.on('error', reject);
+  });
+
+  const data = Buffer.alloc(4096, 'C');
+  client.write(data);
+
+  let received = 0;
+  client.on('data', (chunk) => (received += chunk.length));
+  await tools.delayFor(500);
+
+  console.log(`Socket handler — received ${received} bytes back`);
+
+  client.destroy();
+  await tools.delayFor(500);
+
+  await pollMetrics(proxy);
+
+  const m = proxy.getMetrics();
+  const bytesIn = m.totals.bytesIn();
+  const bytesOut = m.totals.bytesOut();
+  console.log(`Socket handler — bytesIn: ${bytesIn}, bytesOut: ${bytesOut}`);
+
+  // Socket handler relay now records bytes after copy_bidirectional completes
+  expect(bytesIn).toBeGreaterThan(0);
+  expect(bytesOut).toBeGreaterThan(0);
+
+  await proxy.stop();
+  await tools.delayFor(200);
+});
+
+// ════════════════════════════════════════════════════════════════════════════
+// Group 6: Multi-route throughput isolation
+// ════════════════════════════════════════════════════════════════════════════
+tap.test('Multi-route throughput isolation', async (tools) => {
+  const proxy = new SmartProxy({
+    routes: [
+      {
+        id: 'route-alpha',
+        name: 'route-alpha',
+        match: { ports: PROXY_MULTI_A_PORT },
+        action: {
+          type: 'forward',
+          targets: [{ host: 'localhost', port: TCP_ECHO_PORT }],
+        },
+      },
+      {
+        id: 'route-beta',
+        name: 'route-beta',
+        match: { ports: PROXY_MULTI_B_PORT },
+        action: {
+          type: 'forward',
+          targets: [{ host: 'localhost', port: TCP_ECHO_PORT }],
+        },
+      },
+    ],
+    metrics: { enabled: true, sampleIntervalMs: 100, retentionSeconds: 60 },
+  });
+  await proxy.start();
+  await tools.delayFor(300);
+
+  // Send different amounts to each route
+  // Route alpha: 8 KB
+  const clientA = new net.Socket();
+  await new Promise<void>((resolve, reject) => {
+    clientA.connect(PROXY_MULTI_A_PORT, 'localhost', () => resolve());
+    clientA.on('error', reject);
+  });
+  clientA.on('data', () => {}); // drain
+  for (let i = 0; i < 8; i++) {
+    clientA.write(Buffer.alloc(1024, 'A'));
+    await tools.delayFor(50);
+  }
+
+  // Route beta: 2 KB
+  const clientB = new net.Socket();
+  await new Promise<void>((resolve, reject) => {
+    clientB.connect(PROXY_MULTI_B_PORT, 'localhost', () => resolve());
+    clientB.on('error', reject);
+  });
+  clientB.on('data', () => {}); // drain
+  for (let i = 0; i < 2; i++) {
+    clientB.write(Buffer.alloc(1024, 'B'));
+    await tools.delayFor(50);
+  }
+
+  await tools.delayFor(500);
+
+  // Close both
+  clientA.destroy();
+  clientB.destroy();
+  await tools.delayFor(500);
+
+  await pollMetrics(proxy);
+
+  const m = proxy.getMetrics();
+
+  // Check per-route throughput exists for both
+  const byRoute = m.throughput.byRoute();
+  console.log('Multi-route — throughput byRoute:', Array.from(byRoute.entries()));
+
+  // Check per-route connection counts
+  const connByRoute = m.connections.byRoute();
+  console.log('Multi-route — connections byRoute:', Array.from(connByRoute.entries()));
+
+  // Both routes should have tracked data
+  const totalIn = m.totals.bytesIn();
+  const totalOut = m.totals.bytesOut();
+  console.log(`Multi-route — total bytesIn: ${totalIn}, bytesOut: ${totalOut}`);
+
+  expect(totalIn).toBeGreaterThan(0);
+  expect(totalOut).toBeGreaterThan(0);
+
+  await proxy.stop();
+  await tools.delayFor(200);
+});
+
+// ════════════════════════════════════════════════════════════════════════════
+// Group 7: Throughput sampling over time (HTTP-based for real-time tracking)
+//
+// Uses HTTP proxy path where CountingBody reports bytes incrementally
+// as each request/response body completes. This allows the sampling task
+// to capture non-zero throughput during active traffic.
+// ════════════════════════════════════════════════════════════════════════════
+tap.test('Throughput sampling - values appear during active HTTP traffic', async (tools) => {
+  const proxy = new SmartProxy({
+    routes: [
+      {
+        id: 'sampling-test',
+        name: 'sampling-test',
+        match: { ports: PROXY_TP_HTTP_PORT },
+        action: {
+          type: 'forward',
+          targets: [{ host: 'localhost', port: HTTP_ECHO_PORT }],
+        },
+      },
+    ],
+    metrics: { enabled: true, sampleIntervalMs: 100, retentionSeconds: 60 },
+  });
+  await proxy.start();
+  await tools.delayFor(300);
+
+  // Send HTTP requests continuously for ~2 seconds
+  let sending = true;
+  let requestCount = 0;
+  const sendLoop = (async () => {
+    while (sending) {
+      const body = 'D'.repeat(5120); // 5 KB per request
+      try {
+        await new Promise<void>((resolve, reject) => {
+          const req = http.request(
+            {
+              hostname: 'localhost',
+              port: PROXY_TP_HTTP_PORT,
+              path: '/echo',
+              method: 'POST',
+              headers: { 'Content-Type': 'text/plain', 'Content-Length': String(body.length) },
+            },
+            (res) => {
+              res.on('data', () => {});
+              res.on('end', () => resolve());
+            },
+          );
+          req.on('error', reject);
+          req.setTimeout(3000, () => {
+            req.destroy();
+            reject(new Error('timeout'));
+          });
+          req.end(body);
+        });
+        requestCount++;
+      } catch {
+        // Ignore errors during shutdown
+        break;
+      }
+    }
+  })();
+
+  // After 1.5 seconds of active traffic, check throughput
+  await tools.delayFor(1500);
+
+  await pollMetrics(proxy);
+
+  const m = proxy.getMetrics();
+  const tp = m.throughput.instant();
+  const totalIn = m.totals.bytesIn();
+  const totalOut = m.totals.bytesOut();
+  console.log(`Sampling test — after 1.5s of traffic: instant in=${tp.in}, out=${tp.out}`);
+  console.log(`Sampling test — totals: bytesIn=${totalIn}, bytesOut=${totalOut}, requests=${requestCount}`);
+
+  // Totals should definitely be non-zero after 1.5s of HTTP requests
+  expect(totalIn + totalOut).toBeGreaterThan(0);
+
+  // Throughput instant should be non-zero during active traffic.
+  // The sampling interval is 100ms, so we've had ~15 samples by now.
+  // Each sample captures bytes from completed HTTP request/response bodies.
+  // Note: this can occasionally be 0 if sample boundaries don't align, so we
+  // also check that at least the throughput was non-zero for *some* recent window.
+  const tpRecent = m.throughput.recent();
+  console.log(`Sampling test — recent throughput: in=${tpRecent.in}, out=${tpRecent.out}`);
+  expect(tpRecent.in + tpRecent.out).toBeGreaterThan(0);
+
+  // ── v25.2.0: Per-IP tracking ──
+  const byIP = m.connections.byIP();
+  console.log('Sampling test — connections byIP:', Array.from(byIP.entries()));
+  expect(byIP.size).toBeGreaterThan(0);
+
+  const topIPs = m.connections.topIPs(10);
+  console.log('Sampling test — topIPs:', topIPs);
+  expect(topIPs.length).toBeGreaterThan(0);
+  expect(topIPs[0].ip).toBeTruthy();
+  expect(topIPs[0].count).toBeGreaterThanOrEqual(0);
+
+  // ── v25.2.0: Throughput history ──
+  const history = m.throughput.history(10);
+  console.log(`Sampling test — throughput history: ${history.length} points`);
+  if (history.length > 0) {
+    console.log('  first:', history[0], 'last:', history[history.length - 1]);
+  }
+  expect(history.length).toBeGreaterThan(0);
+  expect(history[0].timestamp).toBeGreaterThan(0);
+
+  // ── v25.2.0: Per-IP throughput ──
+  const tpByIP = m.throughput.byIP();
+  console.log('Sampling test — throughput byIP:', Array.from(tpByIP.entries()));
+
+  // ── v25.2.0: HTTP request counting ──
+  const totalReqs = m.requests.total();
+  const rps = m.requests.perSecond();
+  const rpm = m.requests.perMinute();
+  console.log(`Sampling test — HTTP requests: total=${totalReqs}, perSecond=${rps}, perMinute=${rpm}`);
+  expect(totalReqs).toBeGreaterThan(0);
+
+  // Stop sending
+  sending = false;
+  await sendLoop;
+
+  // After traffic stops, wait for metrics to settle
+  await tools.delayFor(500);
+  await pollMetrics(proxy);
+
+  const mAfter = proxy.getMetrics();
+  const tpAfter = mAfter.throughput.instant();
+  console.log(`Sampling test — after traffic stops: instant in=${tpAfter.in}, out=${tpAfter.out}`);
+
+  await proxy.stop();
+  await tools.delayFor(200);
+});
+
+// ════════════════════════════════════════════════════════════════════════════
+// Cleanup
+// ════════════════════════════════════════════════════════════════════════════
+tap.test('cleanup - close backend servers', async () => {
+  await new Promise<void>((resolve) => tcpEchoServer.close(() => resolve()));
+  await new Promise<void>((resolve) => httpEchoServer.close(() => resolve()));
+  await new Promise<void>((resolve) => tlsEchoServer.close(() => resolve()));
+  console.log('All backend servers closed');
+});
+
+export default tap.start();

ts/00_commitinfo_data.ts

@@ -3,6 +3,6 @@
  */
 export const commitinfo = {
   name: '@push.rocks/smartproxy',
-  version: '24.0.1',
+  version: '25.3.0',
   description: 'A powerful proxy package with unified route-based configuration for high traffic management. Features include SSL/TLS support, flexible routing patterns, WebSocket handling, advanced security options, and automatic ACME certificate management.'
 }

ts/index.ts

@@ -8,7 +8,7 @@ export { SharedRouteManager as RouteManager } from './core/routing/route-manager
 
 // Export smart-proxy models
 export type { ISmartProxyOptions, IConnectionRecord, IRouteConfig, IRouteMatch, IRouteAction, IRouteTls, IRouteContext } from './proxies/smart-proxy/models/index.js';
-export type { TSmartProxyCertProvisionObject } from './proxies/smart-proxy/models/interfaces.js';
+export type { TSmartProxyCertProvisionObject, ICertProvisionEventComms, ICertificateIssuedEvent, ICertificateFailedEvent } from './proxies/smart-proxy/models/interfaces.js';
 export * from './proxies/smart-proxy/utils/index.js';
 
 // Original: export * from './smartproxy/classes.pp.snihandler.js'

ts/proxies/smart-proxy/models/index.ts

@@ -2,6 +2,6 @@
  * SmartProxy models
  */
 // Export everything except IAcmeOptions from interfaces
-export type { ISmartProxyOptions, ISmartProxyCertStore, IConnectionRecord, TSmartProxyCertProvisionObject } from './interfaces.js';
+export type { ISmartProxyOptions, ISmartProxyCertStore, IConnectionRecord, TSmartProxyCertProvisionObject, ICertProvisionEventComms, ICertificateIssuedEvent, ICertificateFailedEvent } from './interfaces.js';
 export * from './route-types.js';
 export * from './metrics-types.js';

ts/proxies/smart-proxy/models/interfaces.ts

@@ -34,6 +34,38 @@ import type { IRouteConfig } from './route-types.js';
  */
 export type TSmartProxyCertProvisionObject = plugins.tsclass.network.ICert | 'http01';
 
+/**
+ * Communication channel passed as second argument to certProvisionFunction.
+ * Allows the callback to report metadata back to SmartProxy for event emission.
+ */
+export interface ICertProvisionEventComms {
+  /** Informational log */
+  log: (message: string) => void;
+  /** Warning (non-fatal) */
+  warn: (message: string) => void;
+  /** Error */
+  error: (message: string) => void;
+  /** Set the certificate expiry date (for the issued event) */
+  setExpiryDate: (date: Date) => void;
+  /** Set the source/method used for provisioning (e.g. 'smartacme-dns-01') */
+  setSource: (source: string) => void;
+}
+
+/** Payload for 'certificate-issued' and 'certificate-renewed' events */
+export interface ICertificateIssuedEvent {
+  domain: string;
+  expiryDate?: string;   // ISO 8601
+  source: string;        // e.g. 'certProvisionFunction', 'smartacme-dns-01'
+  isRenewal?: boolean;
+}
+
+/** Payload for 'certificate-failed' event */
+export interface ICertificateFailedEvent {
+  domain: string;
+  error: string;
+  source: string;
+}
+
 // Legacy options and type checking functions have been removed
 
 /**
@@ -140,7 +172,7 @@ export interface ISmartProxyOptions {
    * Optional certificate provider callback. Return 'http01' to use HTTP-01 challenges,
    * or a static certificate object for immediate provisioning.
    */
-  certProvisionFunction?: (domain: string) => Promise<TSmartProxyCertProvisionObject>;
+  certProvisionFunction?: (domain: string, eventComms: ICertProvisionEventComms) => Promise<TSmartProxyCertProvisionObject>;
   
   /**
    * Whether to fallback to ACME if custom certificate provision fails.
@@ -148,6 +180,21 @@ export interface ISmartProxyOptions {
    */
   certProvisionFallbackToAcme?: boolean;
 
+  /**
+   * Per-domain timeout in ms for certProvisionFunction calls.
+   * If a single domain's provisioning takes longer than this, it's aborted
+   * and a certificate-failed event is emitted.
+   * Default: 300000 (5 minutes)
+   */
+  certProvisionTimeout?: number;
+
+  /**
+   * Maximum number of domains to provision certificates for concurrently.
+   * Prevents overwhelming ACME providers when many domains provision at once.
+   * Default: 4
+   */
+  certProvisionConcurrency?: number;
+
   /**
    * Disable the default self-signed fallback certificate.
    * When false (default), a self-signed cert is generated at startup and loaded

ts/proxies/smart-proxy/rust-metrics-adapter.ts

@@ -72,12 +72,23 @@ export class RustMetricsAdapter implements IMetrics {
       return result;
     },
     byIP: (): Map<string, number> => {
-      // Per-IP tracking not yet available from Rust
-      return new Map();
+      const result = new Map<string, number>();
+      if (this.cache?.ips) {
+        for (const [ip, im] of Object.entries(this.cache.ips)) {
+          result.set(ip, (im as any).activeConnections ?? 0);
+        }
+      }
+      return result;
     },
-    topIPs: (_limit?: number): Array<{ ip: string; count: number }> => {
-      // Per-IP tracking not yet available from Rust
-      return [];
+    topIPs: (limit: number = 10): Array<{ ip: string; count: number }> => {
+      const result: Array<{ ip: string; count: number }> = [];
+      if (this.cache?.ips) {
+        for (const [ip, im] of Object.entries(this.cache.ips)) {
+          result.push({ ip, count: (im as any).activeConnections ?? 0 });
+        }
+      }
+      result.sort((a, b) => b.count - a.count);
+      return result.slice(0, limit);
     },
   };
 
@@ -89,7 +100,10 @@ export class RustMetricsAdapter implements IMetrics {
       };
     },
     recent: (): IThroughputData => {
-      return this.throughput.instant();
+      return {
+        in: this.cache?.throughputRecentInBytesPerSec ?? 0,
+        out: this.cache?.throughputRecentOutBytesPerSec ?? 0,
+      };
     },
     average: (): IThroughputData => {
       return this.throughput.instant();
@@ -97,9 +111,13 @@ export class RustMetricsAdapter implements IMetrics {
     custom: (_seconds: number): IThroughputData => {
       return this.throughput.instant();
     },
-    history: (_seconds: number): Array<IThroughputHistoryPoint> => {
-      // Throughput history not yet available from Rust
-      return [];
+    history: (seconds: number): Array<IThroughputHistoryPoint> => {
+      if (!this.cache?.throughputHistory) return [];
+      return this.cache.throughputHistory.slice(-seconds).map((p: any) => ({
+        timestamp: p.timestampMs,
+        in: p.bytesIn,
+        out: p.bytesOut,
+      }));
     },
     byRoute: (_windowSeconds?: number): Map<string, IThroughputData> => {
       const result = new Map<string, IThroughputData>();
@@ -114,21 +132,28 @@ export class RustMetricsAdapter implements IMetrics {
       return result;
     },
     byIP: (_windowSeconds?: number): Map<string, IThroughputData> => {
-      return new Map();
+      const result = new Map<string, IThroughputData>();
+      if (this.cache?.ips) {
+        for (const [ip, im] of Object.entries(this.cache.ips)) {
+          result.set(ip, {
+            in: (im as any).throughputInBytesPerSec ?? 0,
+            out: (im as any).throughputOutBytesPerSec ?? 0,
+          });
+        }
+      }
+      return result;
     },
   };
 
   public requests = {
     perSecond: (): number => {
-      // Rust tracks connections, not HTTP requests (TCP-level proxy)
-      return 0;
+      return this.cache?.httpRequestsPerSec ?? 0;
     },
     perMinute: (): number => {
-      return 0;
+      return (this.cache?.httpRequestsPerSecRecent ?? 0) * 60;
     },
     total: (): number => {
-      // Use total connections as a proxy for total requests
-      return this.cache?.totalConnections ?? 0;
+      return this.cache?.totalHttpRequests ?? this.cache?.totalConnections ?? 0;
     },
   };
 

ts/proxies/smart-proxy/smart-proxy.ts

@@ -12,9 +12,10 @@ import { SharedRouteManager as RouteManager } from '../../core/routing/route-man
 import { RouteValidator } from './utils/route-validator.js';
 import { generateDefaultCertificate } from './utils/default-cert-generator.js';
 import { Mutex } from './utils/mutex.js';
+import { ConcurrencySemaphore } from './utils/concurrency-semaphore.js';
 
 // Types
-import type { ISmartProxyOptions, TSmartProxyCertProvisionObject, IAcmeOptions } from './models/interfaces.js';
+import type { ISmartProxyOptions, TSmartProxyCertProvisionObject, IAcmeOptions, ICertProvisionEventComms, ICertificateIssuedEvent, ICertificateFailedEvent } from './models/interfaces.js';
 import type { IRouteConfig } from './models/route-types.js';
 import type { IMetrics } from './models/metrics-types.js';
 
@@ -38,6 +39,7 @@ export class SmartProxy extends plugins.EventEmitter {
   private metricsAdapter: RustMetricsAdapter;
   private routeUpdateLock: Mutex;
   private stopping = false;
+  private certProvisionPromise: Promise<void> | null = null;
 
   constructor(settingsArg: ISmartProxyOptions) {
     super();
@@ -191,13 +193,18 @@ export class SmartProxy extends plugins.EventEmitter {
       }
     }
 
-    // Handle certProvisionFunction
-    await this.provisionCertificatesViaCallback(preloadedDomains);
-
-    // Start metrics polling
+    // Start metrics polling BEFORE cert provisioning — the Rust engine is already
+    // running and accepting connections, so metrics should be available immediately.
+    // Cert provisioning can hang indefinitely (e.g. DNS-01 ACME timeouts) and must
+    // not block metrics collection.
     this.metricsAdapter.startPolling();
 
     logger.log('info', 'SmartProxy started (Rust engine)', { component: 'smart-proxy' });
+
+    // Fire-and-forget cert provisioning — Rust engine is already running and serving traffic.
+    // Events (certificate-issued / certificate-failed) fire independently per domain.
+    this.certProvisionPromise = this.provisionCertificatesViaCallback(preloadedDomains)
+      .catch((err) => logger.log('error', `Unexpected error in cert provisioning: ${err.message}`, { component: 'smart-proxy' }));
   }
 
   /**
@@ -207,6 +214,12 @@ export class SmartProxy extends plugins.EventEmitter {
     logger.log('info', 'SmartProxy shutting down...', { component: 'smart-proxy' });
     this.stopping = true;
 
+    // Wait for in-flight cert provisioning to bail out (it checks this.stopping)
+    if (this.certProvisionPromise) {
+      await this.certProvisionPromise;
+      this.certProvisionPromise = null;
+    }
+
     // Stop metrics polling
     this.metricsAdapter.stopPolling();
 
@@ -234,7 +247,7 @@ export class SmartProxy extends plugins.EventEmitter {
    * Update routes atomically.
    */
   public async updateRoutes(newRoutes: IRouteConfig[]): Promise<void> {
-    return this.routeUpdateLock.runExclusive(async () => {
+    await this.routeUpdateLock.runExclusive(async () => {
       // Validate
       const validation = RouteValidator.validateRoutes(newRoutes);
       if (!validation.valid) {
@@ -270,11 +283,13 @@ export class SmartProxy extends plugins.EventEmitter {
       // Update stored routes
       this.settings.routes = newRoutes;
 
-      // Handle cert provisioning for new routes
-      await this.provisionCertificatesViaCallback();
-
       logger.log('info', `Routes updated (${newRoutes.length} routes)`, { component: 'smart-proxy' });
     });
+
+    // Fire-and-forget cert provisioning outside the mutex — routes are already updated,
+    // cert provisioning doesn't need the route update lock and may be slow.
+    this.certProvisionPromise = this.provisionCertificatesViaCallback()
+      .catch((err) => logger.log('error', `Unexpected error in cert provisioning after route update: ${err.message}`, { component: 'smart-proxy' }));
   }
 
   /**
@@ -396,6 +411,7 @@ export class SmartProxy extends plugins.EventEmitter {
       extendedKeepAliveLifetime: this.settings.extendedKeepAliveLifetime,
       acceptProxyProtocol: this.settings.acceptProxyProtocol,
       sendProxyProtocol: this.settings.sendProxyProtocol,
+      metrics: this.settings.metrics,
     };
   }
 
@@ -408,7 +424,9 @@ export class SmartProxy extends plugins.EventEmitter {
     const provisionFn = this.settings.certProvisionFunction;
     if (!provisionFn) return;
 
-    const provisionedDomains = new Set<string>(skipDomains);
+    // Phase 1: Collect all unique (domain, route) pairs that need provisioning
+    const seen = new Set<string>(skipDomains);
+    const tasks: Array<{ domain: string; route: IRouteConfig }> = [];
 
     for (const route of this.settings.routes) {
       if (route.action.tls?.certificate !== 'auto') continue;
@@ -418,64 +436,139 @@ export class SmartProxy extends plugins.EventEmitter {
       const certDomains = this.normalizeDomainsForCertProvisioning(rawDomains);
 
       for (const domain of certDomains) {
-        if (provisionedDomains.has(domain)) continue;
-        provisionedDomains.add(domain);
+        if (seen.has(domain)) continue;
+        seen.add(domain);
+        tasks.push({ domain, route });
+      }
+    }
+
+    if (tasks.length === 0) return;
+
+    // Phase 2: Process all domains in parallel with concurrency limit
+    const concurrency = this.settings.certProvisionConcurrency ?? 4;
+    const semaphore = new ConcurrencySemaphore(concurrency);
+
+    const promises = tasks.map(async ({ domain, route }) => {
+      await semaphore.acquire();
+      try {
+        await this.provisionSingleDomain(domain, route, provisionFn);
+      } finally {
+        semaphore.release();
+      }
+    });
+
+    await Promise.allSettled(promises);
+  }
+
+  /**
+   * Provision a single domain's certificate via the callback.
+   * Includes per-domain timeout and shutdown checks.
+   */
+  private async provisionSingleDomain(
+    domain: string,
+    route: IRouteConfig,
+    provisionFn: (domain: string, eventComms: ICertProvisionEventComms) => Promise<TSmartProxyCertProvisionObject>,
+  ): Promise<void> {
+    if (this.stopping) return;
+
+    let expiryDate: string | undefined;
+    let source = 'certProvisionFunction';
+
+    const eventComms: ICertProvisionEventComms = {
+      log: (msg) => logger.log('info', `[certProvision ${domain}] ${msg}`, { component: 'smart-proxy' }),
+      warn: (msg) => logger.log('warn', `[certProvision ${domain}] ${msg}`, { component: 'smart-proxy' }),
+      error: (msg) => logger.log('error', `[certProvision ${domain}] ${msg}`, { component: 'smart-proxy' }),
+      setExpiryDate: (date) => { expiryDate = date.toISOString(); },
+      setSource: (s) => { source = s; },
+    };
+
+    const timeoutMs = this.settings.certProvisionTimeout ?? 300_000; // 5 min default
+
+    try {
+      const result: TSmartProxyCertProvisionObject = await this.withTimeout(
+        provisionFn(domain, eventComms),
+        timeoutMs,
+        `Certificate provisioning timed out for ${domain} after ${timeoutMs}ms`,
+      );
+
+      if (this.stopping) return;
+
+      if (result === 'http01') {
+        if (route.name) {
+          try {
+            await this.bridge.provisionCertificate(route.name);
+            logger.log('info', `Triggered Rust ACME for ${domain} (route: ${route.name})`, { component: 'smart-proxy' });
+          } catch (provisionErr: any) {
+            logger.log('warn', `Cannot provision cert for ${domain} — callback returned 'http01' but Rust ACME failed: ${provisionErr.message}. ` +
+              'Note: Rust ACME is disabled when certProvisionFunction is set.', { component: 'smart-proxy' });
+          }
+        }
+        return;
+      }
+
+      if (result && typeof result === 'object') {
+        if (this.stopping) return;
+
+        const certObj = result as plugins.tsclass.network.ICert;
+        await this.bridge.loadCertificate(
+          domain,
+          certObj.publicKey,
+          certObj.privateKey,
+        );
+        logger.log('info', `Certificate loaded via provision function for ${domain}`, { component: 'smart-proxy' });
+
+        // Persist to consumer store
+        if (this.settings.certStore?.save) {
+          try {
+            await this.settings.certStore.save(domain, certObj.publicKey, certObj.privateKey);
+          } catch (storeErr: any) {
+            logger.log('warn', `certStore.save() failed for ${domain}: ${storeErr.message}`, { component: 'smart-proxy' });
+          }
+        }
+
+        this.emit('certificate-issued', {
+          domain,
+          expiryDate: expiryDate || (certObj.validUntil ? new Date(certObj.validUntil).toISOString() : undefined),
+          source,
+        } satisfies ICertificateIssuedEvent);
+      }
+    } catch (err: any) {
+      logger.log('warn', `certProvisionFunction failed for ${domain}: ${err.message}`, { component: 'smart-proxy' });
+
+      this.emit('certificate-failed', {
+        domain,
+        error: err.message,
+        source,
+      } satisfies ICertificateFailedEvent);
+
+      // Fallback to ACME if enabled and route has a name
+      if (this.settings.certProvisionFallbackToAcme !== false && route.name) {
         try {
-          const result: TSmartProxyCertProvisionObject = await provisionFn(domain);
-
-          if (result === 'http01') {
-            // Callback wants HTTP-01 for this domain — trigger Rust ACME explicitly
-            if (route.name) {
-              try {
-                await this.bridge.provisionCertificate(route.name);
-                logger.log('info', `Triggered Rust ACME for ${domain} (route: ${route.name})`, { component: 'smart-proxy' });
-              } catch (provisionErr: any) {
-                logger.log('warn', `Cannot provision cert for ${domain} — callback returned 'http01' but Rust ACME failed: ${provisionErr.message}. ` +
-                  'Note: Rust ACME is disabled when certProvisionFunction is set.', { component: 'smart-proxy' });
-              }
-            }
-            continue;
-          }
-
-          // Got a static cert object - load it into Rust
-          if (result && typeof result === 'object') {
-            const certObj = result as plugins.tsclass.network.ICert;
-            await this.bridge.loadCertificate(
-              domain,
-              certObj.publicKey,
-              certObj.privateKey,
-            );
-            logger.log('info', `Certificate loaded via provision function for ${domain}`, { component: 'smart-proxy' });
-
-            // Persist to consumer store
-            if (this.settings.certStore?.save) {
-              try {
-                await this.settings.certStore.save(domain, certObj.publicKey, certObj.privateKey);
-              } catch (storeErr: any) {
-                logger.log('warn', `certStore.save() failed for ${domain}: ${storeErr.message}`, { component: 'smart-proxy' });
-              }
-            }
-          }
-        } catch (err: any) {
-          logger.log('warn', `certProvisionFunction failed for ${domain}: ${err.message}`, { component: 'smart-proxy' });
-
-          // Fallback to ACME if enabled and route has a name
-          if (this.settings.certProvisionFallbackToAcme !== false && route.name) {
-            try {
-              await this.bridge.provisionCertificate(route.name);
-              logger.log('info', `Falling back to Rust ACME for ${domain} (route: ${route.name})`, { component: 'smart-proxy' });
-            } catch (acmeErr: any) {
-              logger.log('warn', `ACME fallback also failed for ${domain}: ${acmeErr.message}` +
-                (this.settings.disableDefaultCert
-                  ? ' — TLS will fail for this domain (disableDefaultCert is true)'
-                  : ' — default self-signed fallback cert will be used'), { component: 'smart-proxy' });
-            }
-          }
+          await this.bridge.provisionCertificate(route.name);
+          logger.log('info', `Falling back to Rust ACME for ${domain} (route: ${route.name})`, { component: 'smart-proxy' });
+        } catch (acmeErr: any) {
+          logger.log('warn', `ACME fallback also failed for ${domain}: ${acmeErr.message}` +
+            (this.settings.disableDefaultCert
+              ? ' — TLS will fail for this domain (disableDefaultCert is true)'
+              : ' — default self-signed fallback cert will be used'), { component: 'smart-proxy' });
         }
       }
     }
   }
 
+  /**
+   * Race a promise against a timeout. Rejects with the given message if the timeout fires first.
+   */
+  private withTimeout<T>(promise: Promise<T>, ms: number, message: string): Promise<T> {
+    return new Promise<T>((resolve, reject) => {
+      const timer = setTimeout(() => reject(new Error(message)), ms);
+      promise.then(
+        (val) => { clearTimeout(timer); resolve(val); },
+        (err) => { clearTimeout(timer); reject(err); },
+      );
+    });
+  }
+
   /**
    * Normalize routing glob patterns into valid domain identifiers for cert provisioning.
    * - `*nevermind.cloud` → `['nevermind.cloud', '*.nevermind.cloud']`

ts/proxies/smart-proxy/utils/concurrency-semaphore.ts

@@ -0,0 +1,28 @@
+/**
+ * Async concurrency semaphore — limits the number of concurrent async operations.
+ */
+export class ConcurrencySemaphore {
+  private running = 0;
+  private waitQueue: Array<() => void> = [];
+
+  constructor(private readonly maxConcurrency: number) {}
+
+  async acquire(): Promise<void> {
+    if (this.running < this.maxConcurrency) {
+      this.running++;
+      return;
+    }
+    return new Promise<void>((resolve) => {
+      this.waitQueue.push(() => {
+        this.running++;
+        resolve();
+      });
+    });
+  }
+
+  release(): void {
+    this.running--;
+    const next = this.waitQueue.shift();
+    if (next) next();
+  }
+}

ts/proxies/smart-proxy/utils/index.ts

@@ -17,6 +17,9 @@ export * from './route-utils.js';
 // Export default certificate generator
 export { generateDefaultCertificate } from './default-cert-generator.js';
 
+// Export concurrency semaphore
+export { ConcurrencySemaphore } from './concurrency-semaphore.js';
+
 // Export additional functions from route-helpers that weren't already exported
 export {
   createApiGatewayRoute,