Report packet traffic rates in nym client (#4345)
* Report packet traffic rates * Tweak log * Check for unusual events * Log tweaks
This commit is contained in:
Generated
+7
@@ -6246,6 +6246,7 @@ dependencies = [
|
||||
"serde",
|
||||
"serde_json",
|
||||
"sha2 0.10.8",
|
||||
"si-scale",
|
||||
"sqlx",
|
||||
"tap",
|
||||
"tempfile",
|
||||
@@ -10054,6 +10055,12 @@ dependencies = [
|
||||
"dirs 4.0.0",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "si-scale"
|
||||
version = "0.2.2"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "44beb68bf488343b13ddbd74d1d5d5e6559a58b6dfaee74eb8d5ed4f7ed7666f"
|
||||
|
||||
[[package]]
|
||||
name = "signal-hook"
|
||||
version = "0.3.17"
|
||||
|
||||
@@ -46,6 +46,7 @@ nym-validator-client = { path = "../client-libs/validator-client", default-featu
|
||||
nym-task = { path = "../task" }
|
||||
nym-credential-storage = { path = "../credential-storage" }
|
||||
nym-network-defaults = { path = "../network-defaults" }
|
||||
si-scale = "0.2.2"
|
||||
|
||||
[target."cfg(not(target_arch = \"wasm32\"))".dependencies.tokio-stream]
|
||||
version = "0.1.11"
|
||||
|
||||
@@ -1,11 +1,22 @@
|
||||
use std::time::Duration;
|
||||
use std::{
|
||||
collections::VecDeque,
|
||||
time::{Duration, Instant},
|
||||
};
|
||||
|
||||
use si_scale::helpers::bibytes2;
|
||||
|
||||
use crate::spawn_future;
|
||||
|
||||
// Time interval between reporting packet statistics
|
||||
const PACKET_REPORT_INTERVAL_SECS: u64 = 2;
|
||||
// Interval for taking snapshots of the packet statistics
|
||||
const SNAPSHOT_INTERVAL_MS: u64 = 500;
|
||||
// When computing rates, we include snapshots that are up to this old. We set it to some odd number
|
||||
// a tad larger than an integer number of snapshot intervals, so that we don't have to worry about
|
||||
// threshold effects
|
||||
const RECORDING_WINDOW_MS: u64 = 1700;
|
||||
|
||||
#[derive(Default, Debug)]
|
||||
#[derive(Default, Debug, Clone)]
|
||||
struct PacketStatistics {
|
||||
// Sent
|
||||
real_packets_sent: u64,
|
||||
@@ -80,6 +91,200 @@ impl PacketStatistics {
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
fn summary(&self) -> (String, String) {
|
||||
(
|
||||
format!(
|
||||
"packets sent: {} (real: {}, cover: {}, retransmissions: {})",
|
||||
self.real_packets_sent + self.cover_packets_sent,
|
||||
self.real_packets_sent,
|
||||
self.cover_packets_sent,
|
||||
self.retransmissions_queued,
|
||||
),
|
||||
format!(
|
||||
"packets received: {}, (real: {}, cover: {}, acks: {}, acks for cover: {})",
|
||||
self.real_packets_received + self.cover_packets_received,
|
||||
self.real_packets_received,
|
||||
self.cover_packets_received,
|
||||
self.real_acks_received,
|
||||
self.cover_acks_received,
|
||||
),
|
||||
)
|
||||
}
|
||||
}
|
||||
|
||||
impl std::ops::Sub for PacketStatistics {
|
||||
type Output = Self;
|
||||
|
||||
fn sub(self, rhs: Self) -> Self::Output {
|
||||
Self {
|
||||
real_packets_sent: self.real_packets_sent - rhs.real_packets_sent,
|
||||
real_packets_sent_size: self.real_packets_sent_size - rhs.real_packets_sent_size,
|
||||
cover_packets_sent: self.cover_packets_sent - rhs.cover_packets_sent,
|
||||
cover_packets_sent_size: self.cover_packets_sent_size - rhs.cover_packets_sent_size,
|
||||
|
||||
real_packets_received: self.real_packets_received - rhs.real_packets_received,
|
||||
real_packets_received_size: self.real_packets_received_size
|
||||
- rhs.real_packets_received_size,
|
||||
cover_packets_received: self.cover_packets_received - rhs.cover_packets_received,
|
||||
cover_packets_received_size: self.cover_packets_received_size
|
||||
- rhs.cover_packets_received_size,
|
||||
|
||||
total_acks_received: self.total_acks_received - rhs.total_acks_received,
|
||||
total_acks_received_size: self.total_acks_received_size - rhs.total_acks_received_size,
|
||||
real_acks_received: self.real_acks_received - rhs.real_acks_received,
|
||||
real_acks_received_size: self.real_acks_received_size - rhs.real_acks_received_size,
|
||||
cover_acks_received: self.cover_acks_received - rhs.cover_acks_received,
|
||||
cover_acks_received_size: self.cover_acks_received_size - rhs.cover_acks_received_size,
|
||||
|
||||
real_packets_queued: self.real_packets_queued - rhs.real_packets_queued,
|
||||
retransmissions_queued: self.retransmissions_queued - rhs.retransmissions_queued,
|
||||
reply_surbs_queued: self.reply_surbs_queued - rhs.reply_surbs_queued,
|
||||
additional_reply_surbs_queued: self.additional_reply_surbs_queued
|
||||
- rhs.additional_reply_surbs_queued,
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
#[derive(Debug, Clone)]
|
||||
struct PacketRates {
|
||||
real_packets_sent: f64,
|
||||
real_packets_sent_size: f64,
|
||||
cover_packets_sent: f64,
|
||||
cover_packets_sent_size: f64,
|
||||
|
||||
real_packets_received: f64,
|
||||
real_packets_received_size: f64,
|
||||
cover_packets_received: f64,
|
||||
cover_packets_received_size: f64,
|
||||
|
||||
total_acks_received: f64,
|
||||
total_acks_received_size: f64,
|
||||
real_acks_received: f64,
|
||||
real_acks_received_size: f64,
|
||||
cover_acks_received: f64,
|
||||
cover_acks_received_size: f64,
|
||||
|
||||
real_packets_queued: f64,
|
||||
retransmissions_queued: f64,
|
||||
reply_surbs_queued: f64,
|
||||
additional_reply_surbs_queued: f64,
|
||||
}
|
||||
|
||||
impl From<PacketStatistics> for PacketRates {
|
||||
fn from(stats: PacketStatistics) -> Self {
|
||||
Self {
|
||||
real_packets_sent: stats.real_packets_sent as f64,
|
||||
real_packets_sent_size: stats.real_packets_sent_size as f64,
|
||||
cover_packets_sent: stats.cover_packets_sent as f64,
|
||||
cover_packets_sent_size: stats.cover_packets_sent_size as f64,
|
||||
|
||||
real_packets_received: stats.real_packets_received as f64,
|
||||
real_packets_received_size: stats.real_packets_received_size as f64,
|
||||
cover_packets_received: stats.cover_packets_received as f64,
|
||||
cover_packets_received_size: stats.cover_packets_received_size as f64,
|
||||
|
||||
total_acks_received: stats.total_acks_received as f64,
|
||||
total_acks_received_size: stats.total_acks_received_size as f64,
|
||||
real_acks_received: stats.real_acks_received as f64,
|
||||
real_acks_received_size: stats.real_acks_received_size as f64,
|
||||
cover_acks_received: stats.cover_acks_received as f64,
|
||||
cover_acks_received_size: stats.cover_acks_received_size as f64,
|
||||
|
||||
real_packets_queued: stats.real_packets_queued as f64,
|
||||
retransmissions_queued: stats.retransmissions_queued as f64,
|
||||
reply_surbs_queued: stats.reply_surbs_queued as f64,
|
||||
additional_reply_surbs_queued: stats.additional_reply_surbs_queued as f64,
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
impl std::ops::Sub for PacketRates {
|
||||
type Output = Self;
|
||||
|
||||
fn sub(self, rhs: Self) -> Self::Output {
|
||||
Self {
|
||||
real_packets_sent: self.real_packets_sent - rhs.real_packets_sent,
|
||||
real_packets_sent_size: self.real_packets_sent_size - rhs.real_packets_sent_size,
|
||||
cover_packets_sent: self.cover_packets_sent - rhs.cover_packets_sent,
|
||||
cover_packets_sent_size: self.cover_packets_sent_size - rhs.cover_packets_sent_size,
|
||||
|
||||
real_packets_received: self.real_packets_received - rhs.real_packets_received,
|
||||
real_packets_received_size: self.real_packets_received_size
|
||||
- rhs.real_packets_received_size,
|
||||
cover_packets_received: self.cover_packets_received - rhs.cover_packets_received,
|
||||
cover_packets_received_size: self.cover_packets_received_size
|
||||
- rhs.cover_packets_received_size,
|
||||
|
||||
total_acks_received: self.total_acks_received - rhs.total_acks_received,
|
||||
total_acks_received_size: self.total_acks_received_size - rhs.total_acks_received_size,
|
||||
real_acks_received: self.real_acks_received - rhs.real_acks_received,
|
||||
real_acks_received_size: self.real_acks_received_size - rhs.real_acks_received_size,
|
||||
cover_acks_received: self.cover_acks_received - rhs.cover_acks_received,
|
||||
cover_acks_received_size: self.cover_acks_received_size - rhs.cover_acks_received_size,
|
||||
|
||||
real_packets_queued: self.real_packets_queued - rhs.real_packets_queued,
|
||||
retransmissions_queued: self.retransmissions_queued - rhs.retransmissions_queued,
|
||||
reply_surbs_queued: self.reply_surbs_queued - rhs.reply_surbs_queued,
|
||||
additional_reply_surbs_queued: self.additional_reply_surbs_queued
|
||||
- rhs.additional_reply_surbs_queued,
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
impl std::ops::Div<f64> for PacketRates {
|
||||
type Output = Self;
|
||||
|
||||
fn div(self, rhs: f64) -> Self::Output {
|
||||
Self {
|
||||
real_packets_sent: self.real_packets_sent / rhs,
|
||||
real_packets_sent_size: self.real_packets_sent_size / rhs,
|
||||
cover_packets_sent: self.cover_packets_sent / rhs,
|
||||
cover_packets_sent_size: self.cover_packets_sent_size / rhs,
|
||||
|
||||
real_packets_received: self.real_packets_received / rhs,
|
||||
real_packets_received_size: self.real_packets_received_size / rhs,
|
||||
cover_packets_received: self.cover_packets_received / rhs,
|
||||
cover_packets_received_size: self.cover_packets_received_size / rhs,
|
||||
|
||||
total_acks_received: self.total_acks_received / rhs,
|
||||
total_acks_received_size: self.total_acks_received_size / rhs,
|
||||
real_acks_received: self.real_acks_received / rhs,
|
||||
real_acks_received_size: self.real_acks_received_size / rhs,
|
||||
cover_acks_received: self.cover_acks_received / rhs,
|
||||
cover_acks_received_size: self.cover_acks_received_size / rhs,
|
||||
|
||||
real_packets_queued: self.real_packets_queued / rhs,
|
||||
retransmissions_queued: self.retransmissions_queued / rhs,
|
||||
reply_surbs_queued: self.reply_surbs_queued / rhs,
|
||||
additional_reply_surbs_queued: self.additional_reply_surbs_queued / rhs,
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
impl PacketRates {
|
||||
fn summary(&self) -> String {
|
||||
format!(
|
||||
"rx: {}/s (real: {}/s), tx: {}/s (real: {}/s)",
|
||||
bibytes2(self.real_packets_received_size + self.cover_packets_received_size),
|
||||
bibytes2(self.real_packets_received_size),
|
||||
bibytes2(self.real_packets_sent_size + self.cover_packets_sent_size),
|
||||
bibytes2(self.real_packets_sent_size),
|
||||
)
|
||||
}
|
||||
|
||||
fn detailed_summary(&self) -> String {
|
||||
format!(
|
||||
"RX: {:.1} mixpkt/s, {}/s (real: {}/s, acks: {}/s), TX: {:.1} mixpkt/s, {}/s (real: {}/s)",
|
||||
self.real_packets_received + self.cover_packets_received,
|
||||
bibytes2(self.real_packets_received_size + self.cover_packets_received_size),
|
||||
bibytes2(self.real_packets_received_size),
|
||||
bibytes2(self.total_acks_received_size),
|
||||
self.real_packets_sent + self.cover_packets_sent,
|
||||
bibytes2(self.real_packets_sent_size + self.cover_packets_sent_size),
|
||||
bibytes2(self.real_packets_sent_size),
|
||||
)
|
||||
}
|
||||
}
|
||||
|
||||
pub(crate) enum PacketStatisticsEvent {
|
||||
@@ -130,45 +335,121 @@ impl PacketStatisticsReporter {
|
||||
pub(crate) struct PacketStatisticsControl {
|
||||
// Incoming packet stats events from other tasks
|
||||
stats_rx: PacketStatisticsReceiver,
|
||||
|
||||
// Keep track of packet statistics over time
|
||||
stats: PacketStatistics,
|
||||
|
||||
// We keep snapshots of the statistics over time so we can compute rates, and also keeping the
|
||||
// full history allows for some more fancy averaging if we want to do that.
|
||||
history: VecDeque<(Instant, PacketStatistics)>,
|
||||
|
||||
// Keep previous rates so that we can detect notable events
|
||||
rates: VecDeque<(Instant, PacketRates)>,
|
||||
}
|
||||
|
||||
impl PacketStatisticsControl {
|
||||
pub(crate) fn new() -> (Self, PacketStatisticsReporter) {
|
||||
let (stats_tx, stats_rx) = tokio::sync::mpsc::unbounded_channel();
|
||||
|
||||
(
|
||||
Self {
|
||||
stats_rx,
|
||||
stats: PacketStatistics::default(),
|
||||
history: VecDeque::new(),
|
||||
rates: VecDeque::new(),
|
||||
},
|
||||
PacketStatisticsReporter::new(stats_tx),
|
||||
)
|
||||
}
|
||||
|
||||
fn report_statistics(&self) {
|
||||
// Add the current stats to the history, and remove old ones.
|
||||
fn update_history(&mut self) {
|
||||
// Update latest
|
||||
self.history.push_back((Instant::now(), self.stats.clone()));
|
||||
|
||||
// Filter out old ones
|
||||
let recording_window = Instant::now() - Duration::from_millis(RECORDING_WINDOW_MS);
|
||||
while self
|
||||
.history
|
||||
.front()
|
||||
.map_or(false, |&(t, _)| t < recording_window)
|
||||
{
|
||||
self.history.pop_front();
|
||||
}
|
||||
}
|
||||
|
||||
fn compute_rates(&self) -> Option<PacketRates> {
|
||||
// NOTE: consider changing this to compute rates over the history instead of using current
|
||||
// stats. Currently it should not make much of a difference since we call this just after
|
||||
// updating the history, but it seems like it could be more internally consistent to do it
|
||||
// that way.
|
||||
|
||||
// Do basic averaging over the entire history, which just uses the first and last
|
||||
if let Some((start, start_stats)) = self.history.front() {
|
||||
let duration_secs = Instant::now().duration_since(*start).as_secs_f64();
|
||||
let delta = self.stats.clone() - start_stats.clone();
|
||||
let rates = PacketRates::from(delta) / duration_secs;
|
||||
Some(rates)
|
||||
} else {
|
||||
None
|
||||
}
|
||||
}
|
||||
|
||||
fn update_rates(&mut self) {
|
||||
// Update latest
|
||||
if let Some(rates) = self.compute_rates() {
|
||||
self.rates.push_back((Instant::now(), rates));
|
||||
}
|
||||
|
||||
// Filter out old ones
|
||||
let recording_window = Instant::now() - Duration::from_millis(RECORDING_WINDOW_MS);
|
||||
while self
|
||||
.rates
|
||||
.front()
|
||||
.map_or(false, |&(t, _)| t < recording_window)
|
||||
{
|
||||
self.rates.pop_front();
|
||||
}
|
||||
}
|
||||
|
||||
fn report_rates(&self) {
|
||||
if let Some((_, rates)) = self.rates.back() {
|
||||
log::info!("{}", rates.summary());
|
||||
log::debug!("{}", rates.detailed_summary());
|
||||
}
|
||||
}
|
||||
|
||||
fn report_counters(&self) {
|
||||
log::trace!("packet statistics: {:?}", &self.stats);
|
||||
log::info!(
|
||||
"packets sent: {} (real: {}, cover: {}, retransmissions: {})",
|
||||
self.stats.real_packets_sent + self.stats.cover_packets_sent,
|
||||
self.stats.real_packets_sent,
|
||||
self.stats.cover_packets_sent,
|
||||
self.stats.retransmissions_queued,
|
||||
);
|
||||
log::info!(
|
||||
"packets received: {}, (real: {}, cover: {}, acks: {}, acks for cover: {})",
|
||||
self.stats.real_packets_received + self.stats.cover_packets_received,
|
||||
self.stats.real_packets_received,
|
||||
self.stats.cover_packets_received,
|
||||
self.stats.real_acks_received,
|
||||
self.stats.cover_acks_received,
|
||||
);
|
||||
let (summary_sent, summary_recv) = self.stats.summary();
|
||||
log::debug!("{}", summary_sent);
|
||||
log::debug!("{}", summary_recv);
|
||||
}
|
||||
|
||||
fn check_for_notable_events(&self) {
|
||||
let Some((_, latest_rates)) = self.rates.back() else {
|
||||
return;
|
||||
};
|
||||
|
||||
// If we get a burst of retransmissions
|
||||
// TODO: consider making this the number of retransmissions since last report instead.
|
||||
if latest_rates.retransmissions_queued > 0.0 {
|
||||
log::info!(
|
||||
"retransmissions: {} pkt/s",
|
||||
latest_rates.retransmissions_queued
|
||||
);
|
||||
}
|
||||
|
||||
// IDEA: if there is a burst of acks, that could indicate tokio task starvation.
|
||||
}
|
||||
|
||||
pub(crate) async fn run_with_shutdown(&mut self, mut shutdown: nym_task::TaskClient) {
|
||||
log::debug!("Started PacketStatisticsControl with graceful shutdown support");
|
||||
|
||||
let interval = Duration::from_secs(PACKET_REPORT_INTERVAL_SECS);
|
||||
let mut interval = tokio::time::interval(interval);
|
||||
let report_interval = Duration::from_secs(PACKET_REPORT_INTERVAL_SECS);
|
||||
let mut report_interval = tokio::time::interval(report_interval);
|
||||
let snapshot_interval = Duration::from_millis(SNAPSHOT_INTERVAL_MS);
|
||||
let mut snapshot_interval = tokio::time::interval(snapshot_interval);
|
||||
|
||||
loop {
|
||||
tokio::select! {
|
||||
@@ -182,8 +463,14 @@ impl PacketStatisticsControl {
|
||||
break;
|
||||
}
|
||||
},
|
||||
_ = interval.tick() => {
|
||||
self.report_statistics();
|
||||
_ = snapshot_interval.tick() => {
|
||||
self.update_history();
|
||||
self.update_rates();
|
||||
}
|
||||
_ = report_interval.tick() => {
|
||||
self.report_rates();
|
||||
self.check_for_notable_events();
|
||||
self.report_counters();
|
||||
}
|
||||
_ = shutdown.recv_with_delay() => {
|
||||
log::trace!("PacketStatisticsControl: Received shutdown");
|
||||
|
||||
Reference in New Issue
Block a user