Bug 1677951 - Add telemetry about low level HTTP3 performance. r=necko-reviewers,kershaw

Differential Revision: https://phabricator.services.mozilla.com/D97422

netwerk/protocol/http/Http3Session.cpp

@@ -46,6 +46,7 @@ const uint64_t HTTP3_APP_ERROR_CONNECT_ERROR = 0x10f;
 const uint64_t HTTP3_APP_ERROR_VERSION_FALLBACK = 0x110;
 
 const uint32_t UDP_MAX_PACKET_SIZE = 4096;
+const uint32_t MAX_PTO_COUNTS = 16;
 
 NS_IMPL_ADDREF(Http3Session)
 NS_IMPL_RELEASE(Http3Session)
@@ -1641,6 +1642,39 @@ void Http3Session::CloseConnectionTelemetry(CloseError& aError, bool aClosing) {
   // connection, this will map to "closing" key and 37 in the graph.
   Telemetry::Accumulate(Telemetry::HTTP3_CONNECTTION_CLOSE_CODE,
                         aClosing ? "closing"_ns : "closed"_ns, value);
+
+  Http3Stats stats;
+  mHttp3Connection->GetStats(&stats);
+
+  if (stats.packets_tx > 0) {
+    unsigned long loss = (stats.lost * 10000) / stats.packets_tx;
+    Telemetry::Accumulate(Telemetry::HTTP3_LOSS_RATIO, loss);
+
+    Telemetry::Accumulate(Telemetry::HTTP3_LATE_ACK, "ack"_ns, stats.late_ack);
+    Telemetry::Accumulate(Telemetry::HTTP3_LATE_ACK, "pto"_ns, stats.pto_ack);
+
+    unsigned long late_ack_ratio = (stats.late_ack * 10000) / stats.packets_tx;
+    unsigned long pto_ack_ratio = (stats.pto_ack * 10000) / stats.packets_tx;
+    Telemetry::Accumulate(Telemetry::HTTP3_LATE_ACK_RATIO, "ack"_ns,
+                          late_ack_ratio);
+    Telemetry::Accumulate(Telemetry::HTTP3_LATE_ACK_RATIO, "pto"_ns,
+                          pto_ack_ratio);
+
+    for (uint32_t i = 0; i < MAX_PTO_COUNTS; i++) {
+      nsAutoCString key;
+      key.AppendInt(i);
+      Telemetry::Accumulate(Telemetry::HTTP3_COUNTS_PTO, key,
+                            stats.pto_counts[i]);
+    }
+
+    Telemetry::Accumulate(Telemetry::HTTP3_DROP_DGRAMS, stats.dropped_rx);
+    Telemetry::Accumulate(Telemetry::HTTP3_SAVED_DGRAMS, stats.saved_datagrams);
+  }
+
+  Telemetry::Accumulate(Telemetry::HTTP3_RECEIVED_SENT_DGRAMS, "received"_ns,
+                        stats.packets_rx);
+  Telemetry::Accumulate(Telemetry::HTTP3_RECEIVED_SENT_DGRAMS, "sent"_ns,
+                        stats.packets_tx);
 }
 
 void Http3Session::Finish0Rtt(bool aRestart) {

netwerk/socket/neqo_glue/NeqoHttp3Conn.h

@@ -90,6 +90,10 @@ class NeqoHttp3Conn final {
   nsrefcnt AddRef() { return neqo_http3conn_addref(this); }
   nsrefcnt Release() { return neqo_http3conn_release(this); }
 
+  void GetStats(Http3Stats* aStats) {
+    return neqo_http3conn_get_stats(this, aStats);
+  }
+
  private:
   NeqoHttp3Conn() = delete;
   ~NeqoHttp3Conn() = delete;

netwerk/socket/neqo_glue/src/lib.rs

@@ -8,8 +8,9 @@ use neqo_crypto::{init, PRErrorCode};
 use neqo_http3::Error as Http3Error;
 use neqo_http3::{Http3Client, Http3ClientEvent, Http3Parameters, Http3State};
 use neqo_qpack::QpackSettings;
-use neqo_transport::Error as TransportError;
-use neqo_transport::{ConnectionParameters, FixedConnectionIdManager, Output, QuicVersion};
+use neqo_transport::{
+    ConnectionParameters, Error as TransportError, FixedConnectionIdManager, Output, QuicVersion,
+};
 use nserror::*;
 use nsstring::*;
 use qlog::QlogStreamer;
@@ -774,3 +775,42 @@ pub extern "C" fn neqo_http3conn_set_resumption_token(
 pub extern "C" fn neqo_http3conn_is_zero_rtt(conn: &mut NeqoHttp3Conn) -> bool {
     conn.conn.state() == Http3State::ZeroRtt
 }
+
+#[repr(C)]
+#[derive(Default)]
+pub struct Http3Stats {
+    /// Total packets received, including all the bad ones.
+    pub packets_rx: usize,
+    /// Duplicate packets received.
+    pub dups_rx: usize,
+    /// Dropped packets or dropped garbage.
+    pub dropped_rx: usize,
+    /// The number of packet that were saved for later processing.
+    pub saved_datagrams: usize,
+    /// Total packets sent.
+    pub packets_tx: usize,
+    /// Total number of packets that are declared lost.
+    pub lost: usize,
+    /// Late acknowledgments, for packets that were declared lost already.
+    pub late_ack: usize,
+    /// Acknowledgments for packets that contained data that was marked
+    /// for retransmission when the PTO timer popped.
+    pub pto_ack: usize,
+    /// Count PTOs. Single PTOs, 2 PTOs in a row, 3 PTOs in row, etc. are counted
+    /// separately.
+    pub pto_counts: [usize; 16],
+}
+
+#[no_mangle]
+pub extern "C" fn neqo_http3conn_get_stats(conn: &mut NeqoHttp3Conn, stats: &mut Http3Stats) {
+    let t_stats = conn.conn.transport_stats();
+    stats.packets_rx = t_stats.packets_rx;
+    stats.dups_rx = t_stats.dups_rx;
+    stats.dropped_rx = t_stats.dropped_rx;
+    stats.saved_datagrams = t_stats.saved_datagrams;
+    stats.packets_tx = t_stats.packets_tx;
+    stats.lost = t_stats.lost;
+    stats.late_ack = t_stats.late_ack;
+    stats.pto_ack = t_stats.pto_ack;
+    stats.pto_counts = t_stats.pto_counts;
+}

toolkit/components/telemetry/Histograms.json

@@ -3450,6 +3450,87 @@
     "bug_numbers": [1655566],
     "description": "HTTP3: number of times when a stream is blocked by the flow control while sendnig data."
   },
+  "HTTP3_LOSS_RATIO": {
+    "record_in_processes": ["main", "content"],
+    "products": ["firefox"],
+    "alert_emails": ["necko@mozilla.com", "ddamjanovic@mozilla.com"],
+    "expires_in_version": "never",
+    "kind": "exponential",
+    "high": 2000,
+    "n_buckets": 100,
+    "bug_numbers": [1677951],
+    "description": "HTTP3: packet loss ratio (multiply by 10000)."
+  },
+  "HTTP3_LATE_ACK_RATIO": {
+    "record_in_processes": ["main", "content"],
+    "products": ["firefox"],
+    "alert_emails": ["necko@mozilla.com", "ddamjanovic@mozilla.com"],
+    "expires_in_version": "never",
+    "kind": "exponential",
+    "high": 2000,
+    "n_buckets": 100,
+    "keyed": true,
+    "bug_numbers": [1677951],
+    "description": "HTTP3: spurious retransmissions ratio (spurios_retransmission / packet sent * 10000)."
+  },
+  "HTTP3_LATE_ACK": {
+    "record_in_processes": ["main", "content"],
+    "products": ["firefox"],
+    "alert_emails": ["necko@mozilla.com", "ddamjanovic@mozilla.com"],
+    "expires_in_version": "never",
+    "kind": "exponential",
+    "high": 10000,
+    "n_buckets": 100,
+    "keyed": true,
+    "bug_numbers": [1677951],
+    "description": "HTTP3: the number of spurious retransmissions."
+  },
+  "HTTP3_COUNTS_PTO": {
+    "record_in_processes": ["main", "content"],
+    "products": ["firefox"],
+    "alert_emails": ["necko@mozilla.com", "ddamjanovic@mozilla.com"],
+    "expires_in_version": "never",
+    "kind": "exponential",
+    "high": 5000,
+    "n_buckets": 100,
+    "keyed": true,
+    "bug_numbers": [1677951],
+    "description": "HTTP3: the number of PTOs."
+  },
+  "HTTP3_DROP_DGRAMS": {
+    "record_in_processes": ["main", "content"],
+    "products": ["firefox"],
+    "alert_emails": ["necko@mozilla.com", "ddamjanovic@mozilla.com"],
+    "expires_in_version": "never",
+    "kind": "exponential",
+    "high": 5000,
+    "n_buckets": 100,
+    "bug_numbers": [1677951],
+    "description": "HTTP3: the number of dropped datagrams."
+  },
+  "HTTP3_SAVED_DGRAMS": {
+    "record_in_processes": ["main", "content"],
+    "products": ["firefox"],
+    "alert_emails": ["necko@mozilla.com", "ddamjanovic@mozilla.com"],
+    "expires_in_version": "never",
+    "kind": "exponential",
+    "high": 5000,
+    "n_buckets": 100,
+    "bug_numbers": [1677951],
+    "description": "HTTP3: the number of saved datagrams that are waiting for keys to be available."
+  },
+  "HTTP3_RECEIVED_SENT_DGRAMS": {
+    "record_in_processes": ["main", "content"],
+    "products": ["firefox"],
+    "alert_emails": ["necko@mozilla.com", "ddamjanovic@mozilla.com"],
+    "expires_in_version": "never",
+    "kind": "exponential",
+    "high": 100000,
+    "n_buckets": 100,
+    "keyed": true,
+    "bug_numbers": [1677951],
+    "description": "HTTP3: the number of received/sent packets."
+  },
   "HTTP_CONTENT_ENCODING": {
     "record_in_processes": ["main", "content"],
     "products": ["firefox", "fennec"],