aboutsummaryrefslogtreecommitdiff
path: root/src/peering/fullmesh.rs
blob: 068055edf0b82467036c43ab53ca129f1e8f31a3 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
use std::collections::{HashMap, VecDeque};
use std::net::SocketAddr;
use std::sync::atomic::{self, AtomicU64};
use std::sync::{Arc, RwLock};
use std::time::{Duration, Instant};

use arc_swap::ArcSwap;
use async_trait::async_trait;
use log::{debug, info, trace, warn};
use serde::{Deserialize, Serialize};

use tokio::sync::watch;

use sodiumoxide::crypto::hash;

use crate::endpoint::*;
use crate::netapp::*;
use crate::proto::*;
use crate::NodeID;

const CONN_RETRY_INTERVAL: Duration = Duration::from_secs(30);
const CONN_MAX_RETRIES: usize = 10;
const PING_INTERVAL: Duration = Duration::from_secs(10);
const LOOP_DELAY: Duration = Duration::from_secs(1);

// -- Protocol messages --

#[derive(Serialize, Deserialize)]
struct PingMessage {
	pub id: u64,
	pub peer_list_hash: hash::Digest,
}

impl Message for PingMessage {
	type Response = PingMessage;
}

#[derive(Serialize, Deserialize)]
struct PeerListMessage {
	pub list: Vec<(NodeID, SocketAddr)>,
}

impl Message for PeerListMessage {
	type Response = PeerListMessage;
}

// -- Algorithm data structures --

#[derive(Debug)]
struct PeerInfoInternal {
	addr: SocketAddr,
	state: PeerConnState,
	last_seen: Option<Instant>,
	ping: VecDeque<Duration>,
}

#[derive(Copy, Clone, Debug)]
pub struct PeerInfo {
	/// The node's identifier (its public key)
	pub id: NodeID,
	/// The node's network address
	pub addr: SocketAddr,
	/// The current status of our connection to this node
	pub state: PeerConnState,
	/// The last time at which the node was seen
	pub last_seen: Option<Instant>,
	/// The average ping to this node  on recent observations (if at least one ping value is known)
	pub avg_ping: Option<Duration>,
	/// The maximum observed ping to this node on recent observations (if at least one
	/// ping value is known)
	pub max_ping: Option<Duration>,
	/// The median ping to this node on recent observations (if at least one ping value
	/// is known)
	pub med_ping: Option<Duration>,
}

impl PeerInfo {
	/// Returns true if we can currently send requests to this peer
	pub fn is_up(&self) -> bool {
		self.state.is_up()
	}
}

/// PeerConnState: possible states for our tentative connections to given peer
/// This structure is only interested in recording connection info for outgoing
/// TCP connections
#[derive(Copy, Clone, Debug, PartialEq)]
pub enum PeerConnState {
	/// This entry represents ourself (the local node)
	Ourself,

	/// We currently have a connection to this peer
	Connected,

	/// Our next connection tentative (the nth, where n is the first value of the tuple)
	/// will be at given Instant
	Waiting(usize, Instant),

	/// A connection tentative is in progress (the nth, where n is the value stored)
	Trying(usize),

	/// We abandonned trying to connect to this peer (too many failed attempts)
	Abandonned,
}

impl PeerConnState {
	/// Returns true if we can currently send requests to this peer
	pub fn is_up(&self) -> bool {
		matches!(self, Self::Ourself | Self::Connected)
	}
}

struct KnownHosts {
	list: HashMap<NodeID, PeerInfoInternal>,
	hash: hash::Digest,
}

impl KnownHosts {
	fn new() -> Self {
		let list = HashMap::new();
		let hash = Self::calculate_hash(&list);
		Self { list, hash }
	}
	fn update_hash(&mut self) {
		self.hash = Self::calculate_hash(&self.list);
	}
	fn map_into_vec(input: &HashMap<NodeID, PeerInfoInternal>) -> Vec<(NodeID, SocketAddr)> {
		let mut list = Vec::with_capacity(input.len());
		for (id, peer) in input.iter() {
			if peer.state == PeerConnState::Connected || peer.state == PeerConnState::Ourself {
				list.push((*id, peer.addr));
			}
		}
		list
	}
	fn calculate_hash(input: &HashMap<NodeID, PeerInfoInternal>) -> hash::Digest {
		let mut list = Self::map_into_vec(input);
		list.sort();
		let mut hash_state = hash::State::new();
		for (id, addr) in list {
			hash_state.update(&id[..]);
			hash_state.update(&format!("{}\n", addr).into_bytes()[..]);
		}
		hash_state.finalize()
	}
}

/// A "Full Mesh" peering strategy is a peering strategy that tries
/// to establish and maintain a direct connection with all of the
/// known nodes in the network.
pub struct FullMeshPeeringStrategy {
	netapp: Arc<NetApp>,
	known_hosts: RwLock<KnownHosts>,
	public_peer_list: ArcSwap<Vec<PeerInfo>>,

	next_ping_id: AtomicU64,
	ping_endpoint: Arc<Endpoint<PingMessage, Self>>,
	peer_list_endpoint: Arc<Endpoint<PeerListMessage, Self>>,
}

impl FullMeshPeeringStrategy {
	/// Create a new Full Mesh peering strategy.
	/// The strategy will not be run until `.run()` is called and awaited.
	/// Once that happens, the peering strategy will try to connect
	/// to all of the nodes specified in the bootstrap list.
	pub fn new(
		netapp: Arc<NetApp>,
		bootstrap_list: Vec<(NodeID, SocketAddr)>,
		our_addr: Option<SocketAddr>,
	) -> Arc<Self> {
		let mut known_hosts = KnownHosts::new();
		for (id, addr) in bootstrap_list {
			if id != netapp.id {
				known_hosts.list.insert(
					id,
					PeerInfoInternal {
						addr,
						state: PeerConnState::Waiting(0, Instant::now()),
						last_seen: None,
						ping: VecDeque::new(),
					},
				);
			}
		}

		if let Some(addr) = our_addr {
			known_hosts.list.insert(
				netapp.id,
				PeerInfoInternal {
					addr,
					state: PeerConnState::Ourself,
					last_seen: None,
					ping: VecDeque::new(),
				},
			);
		}

		let strat = Arc::new(Self {
			netapp: netapp.clone(),
			known_hosts: RwLock::new(known_hosts),
			public_peer_list: ArcSwap::new(Arc::new(Vec::new())),
			next_ping_id: AtomicU64::new(42),
			ping_endpoint: netapp.endpoint("__netapp/peering/fullmesh.rs/Ping".into()),
			peer_list_endpoint: netapp.endpoint("__netapp/peering/fullmesh.rs/PeerList".into()),
		});

		strat.update_public_peer_list(&strat.known_hosts.read().unwrap());

		strat.ping_endpoint.set_handler(strat.clone());
		strat.peer_list_endpoint.set_handler(strat.clone());

		let strat2 = strat.clone();
		netapp.on_connected(move |id: NodeID, addr: SocketAddr, is_incoming: bool| {
			let strat2 = strat2.clone();
			strat2.on_connected(id, addr, is_incoming);
		});

		let strat2 = strat.clone();
		netapp.on_disconnected(move |id: NodeID, is_incoming: bool| {
			let strat2 = strat2.clone();
			strat2.on_disconnected(id, is_incoming);
		});

		strat
	}

	/// Run the full mesh peering strategy.
	/// This future exits when the `must_exit` watch becomes true.
	pub async fn run(self: Arc<Self>, must_exit: watch::Receiver<bool>) {
		while !*must_exit.borrow() {
			// 1. Read current state: get list of connected peers (ping them)
			let (to_ping, to_retry) = {
				let known_hosts = self.known_hosts.read().unwrap();
				trace!("known_hosts: {} peers", known_hosts.list.len());

				let mut to_ping = vec![];
				let mut to_retry = vec![];
				for (id, info) in known_hosts.list.iter() {
					trace!("{}, {:?}", hex::encode(id), info);
					match info.state {
						PeerConnState::Connected => {
							let must_ping = match info.last_seen {
								None => true,
								Some(t) => Instant::now() - t > PING_INTERVAL,
							};
							if must_ping {
								to_ping.push(*id);
							}
						}
						PeerConnState::Waiting(_, t) => {
							if Instant::now() >= t {
								to_retry.push(*id);
							}
						}
						_ => (),
					}
				}
				(to_ping, to_retry)
			};

			// 2. Dispatch ping to hosts
			trace!("to_ping: {} peers", to_retry.len());
			for id in to_ping {
				tokio::spawn(self.clone().ping(id));
			}

			// 3. Try reconnects
			trace!("to_retry: {} peers", to_retry.len());
			if !to_retry.is_empty() {
				let mut known_hosts = self.known_hosts.write().unwrap();
				for id in to_retry {
					if let Some(h) = known_hosts.list.get_mut(&id) {
						if let PeerConnState::Waiting(i, _) = h.state {
							info!(
								"Retrying connection to {} at {} ({})",
								hex::encode(&id),
								h.addr,
								i + 1
							);
							h.state = PeerConnState::Trying(i);
							tokio::spawn(self.clone().try_connect(id, h.addr));
						}
					}
				}
				self.update_public_peer_list(&known_hosts);
			}

			// 4. Sleep before next loop iteration
			tokio::time::sleep(LOOP_DELAY).await;
		}
	}

	/// Returns a list of currently known peers in the network.
	pub fn get_peer_list(&self) -> Arc<Vec<PeerInfo>> {
		self.public_peer_list.load_full()
	}

	// -- internal stuff --

	fn update_public_peer_list(&self, known_hosts: &KnownHosts) {
		let mut pub_peer_list = Vec::with_capacity(known_hosts.list.len());
		for (id, info) in known_hosts.list.iter() {
			let mut pings = info.ping.iter().cloned().collect::<Vec<_>>();
			pings.sort();
			if !pings.is_empty() {
				pub_peer_list.push(PeerInfo {
					id: *id,
					addr: info.addr,
					state: info.state,
					last_seen: info.last_seen,
					avg_ping: Some(
						pings
							.iter()
							.fold(Duration::from_secs(0), |x, y| x + *y)
							.div_f64(pings.len() as f64),
					),
					max_ping: pings.last().cloned(),
					med_ping: Some(pings[pings.len() / 2]),
				});
			} else {
				pub_peer_list.push(PeerInfo {
					id: *id,
					addr: info.addr,
					state: info.state,
					last_seen: info.last_seen,
					avg_ping: None,
					max_ping: None,
					med_ping: None,
				});
			}
		}
		self.public_peer_list.store(Arc::new(pub_peer_list));
	}

	async fn ping(self: Arc<Self>, id: NodeID) {
		let peer_list_hash = self.known_hosts.read().unwrap().hash;
		let ping_id = self.next_ping_id.fetch_add(1u64, atomic::Ordering::Relaxed);
		let ping_time = Instant::now();
		let ping_msg = PingMessage {
			id: ping_id,
			peer_list_hash,
		};

		debug!(
			"Sending ping {} to {} at {:?}",
			ping_id,
			hex::encode(id),
			ping_time
		);
		match self.ping_endpoint.call(&id, &ping_msg, PRIO_HIGH).await {
			Err(e) => warn!("Error pinging {}: {}", hex::encode(id), e),
			Ok(ping_resp) => {
				let resp_time = Instant::now();
				debug!(
					"Got ping response from {} at {:?}",
					hex::encode(id),
					resp_time
				);
				{
					let mut known_hosts = self.known_hosts.write().unwrap();
					if let Some(host) = known_hosts.list.get_mut(&id) {
						host.last_seen = Some(resp_time);
						host.ping.push_back(resp_time - ping_time);
						while host.ping.len() > 10 {
							host.ping.pop_front();
						}
						self.update_public_peer_list(&known_hosts);
					}
				}
				if ping_resp.peer_list_hash != peer_list_hash {
					self.exchange_peers(&id).await;
				}
			}
		}
	}

	async fn exchange_peers(self: Arc<Self>, id: &NodeID) {
		let peer_list = KnownHosts::map_into_vec(&self.known_hosts.read().unwrap().list);
		let pex_message = PeerListMessage { list: peer_list };
		match self
			.peer_list_endpoint
			.call(id, &pex_message, PRIO_BACKGROUND)
			.await
		{
			Err(e) => warn!("Error doing peer exchange: {}", e),
			Ok(resp) => {
				self.handle_peer_list(&resp.list[..]);
			}
		}
	}

	fn handle_peer_list(&self, list: &[(NodeID, SocketAddr)]) {
		let mut known_hosts = self.known_hosts.write().unwrap();

		let mut changed = false;
		for (id, addr) in list.iter() {
			if !known_hosts.list.contains_key(id) {
				known_hosts.list.insert(*id, self.new_peer(id, *addr));
				changed = true;
			}
		}

		if changed {
			known_hosts.update_hash();
			self.update_public_peer_list(&known_hosts);
		}
	}

	async fn try_connect(self: Arc<Self>, id: NodeID, addr: SocketAddr) {
		let conn_result = self.netapp.clone().try_connect(addr, id).await;
		if let Err(e) = conn_result {
			warn!("Error connecting to {}: {}", hex::encode(id), e);
			let mut known_hosts = self.known_hosts.write().unwrap();
			if let Some(host) = known_hosts.list.get_mut(&id) {
				host.state = match host.state {
					PeerConnState::Trying(i) => {
						if i >= CONN_MAX_RETRIES {
							PeerConnState::Abandonned
						} else {
							PeerConnState::Waiting(i + 1, Instant::now() + CONN_RETRY_INTERVAL)
						}
					}
					_ => PeerConnState::Waiting(0, Instant::now() + CONN_RETRY_INTERVAL),
				};
				self.update_public_peer_list(&known_hosts);
			}
		} else {
			self.on_connected(id, addr, false);
		}
	}

	fn on_connected(self: Arc<Self>, id: NodeID, addr: SocketAddr, is_incoming: bool) {
		if is_incoming {
			let mut known_hosts = self.known_hosts.write().unwrap();
			if !known_hosts.list.contains_key(&id) {
				known_hosts
					.list
					.insert(id, self.new_peer(&id, addr));
				known_hosts.update_hash();
				self.update_public_peer_list(&known_hosts);
			}
		} else {
			info!("Successfully connected to {} at {}", hex::encode(&id), addr);
			let mut known_hosts = self.known_hosts.write().unwrap();
			if let Some(host) = known_hosts.list.get_mut(&id) {
				host.state = PeerConnState::Connected;
				host.addr = addr;
				known_hosts.update_hash();
				self.update_public_peer_list(&known_hosts);
			}
		}
	}

	fn on_disconnected(self: Arc<Self>, id: NodeID, is_incoming: bool) {
		if !is_incoming {
			info!("Connection to {} was closed", hex::encode(id));
			let mut known_hosts = self.known_hosts.write().unwrap();
			if let Some(host) = known_hosts.list.get_mut(&id) {
				host.state = PeerConnState::Waiting(0, Instant::now());
				known_hosts.update_hash();
				self.update_public_peer_list(&known_hosts);
			}
		}
	}

	fn new_peer(&self, id: &NodeID, addr: SocketAddr) -> PeerInfoInternal {
		let state = if *id == self.netapp.id {
			PeerConnState::Ourself
		} else {
			PeerConnState::Waiting(0, Instant::now())
		};
		PeerInfoInternal {
			addr,
			state,
			last_seen: None,
			ping: VecDeque::new(),
		}
	}
}

#[async_trait]
impl EndpointHandler<PingMessage> for FullMeshPeeringStrategy {
	async fn handle(self: &Arc<Self>, ping: &PingMessage, from: NodeID) -> PingMessage {
		let ping_resp = PingMessage {
			id: ping.id,
			peer_list_hash: self.known_hosts.read().unwrap().hash,
		};
		debug!("Ping from {}", hex::encode(&from));
		ping_resp
	}
}

#[async_trait]
impl EndpointHandler<PeerListMessage> for FullMeshPeeringStrategy {
	async fn handle(
		self: &Arc<Self>,
		peer_list: &PeerListMessage,
		_from: NodeID,
	) -> PeerListMessage {
		self.handle_peer_list(&peer_list.list[..]);
		let peer_list = KnownHosts::map_into_vec(&self.known_hosts.read().unwrap().list);
		PeerListMessage { list: peer_list }
	}
}