-
Notifications
You must be signed in to change notification settings - Fork 0
/
mod.rs
455 lines (432 loc) · 17.4 KB
/
mod.rs
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
// Copyright (c) The Diem Core Contributors
// SPDX-License-Identifier: Apache-2.0
//! Protocol used to ensure peer liveness
//!
//! The HealthChecker is responsible for ensuring liveness of all peers of a node.
//! It does so by periodically selecting a random connected peer and sending a Ping probe. A
//! healthy peer is expected to respond with a corresponding Pong message.
//!
//! If a certain number of successive liveness probes for a peer fail, the HealthChecker initiates a
//! disconnect from the peer. It relies on ConnectivityManager or the remote peer to re-establish
//! the connection.
//!
//! Future Work
//! -----------
//! We can make a few other improvements to the health checker. These are:
//! - Make the policy for interpreting ping failures pluggable
//! - Use successful inbound pings as a sign of remote note being healthy
//! - Ping a peer only in periods of no application-level communication with the peer
use crate::{
application::interface::NetworkInterface,
constants::NETWORK_CHANNEL_SIZE,
counters,
error::NetworkError,
logging::NetworkSchema,
peer_manager::{ConnectionRequestSender, PeerManagerRequestSender},
protocols::{
health_checker::interface::{HealthCheckData, HealthCheckNetworkInterface},
network::{Event, NetworkEvents, NetworkSender, NewNetworkSender},
rpc::error::RpcError,
},
ProtocolId,
};
use bytes::Bytes;
use channel::message_queues::QueueStyle;
use diem_config::network_id::NetworkContext;
use diem_logger::prelude::*;
use diem_metrics::IntCounterVec;
use diem_time_service::{TimeService, TimeServiceTrait};
use diem_types::PeerId;
use futures::{
channel::oneshot,
stream::{FuturesUnordered, StreamExt},
};
use rand::{rngs::SmallRng, Rng, SeedableRng};
use serde::{Deserialize, Serialize};
use short_hex_str::AsShortHexStr;
use std::{collections::hash_map::Entry, sync::Arc, time::Duration};
pub mod builder;
mod interface;
#[cfg(test)]
mod test;
/// The interface from Network to HealthChecker layer.
///
/// `HealthCheckerNetworkEvents` is a `Stream` of `PeerManagerNotification` where the
/// raw `Bytes` rpc messages are deserialized into
/// `HealthCheckerMsg` types. `HealthCheckerNetworkEvents` is a thin wrapper
/// around an `channel::Receiver<PeerManagerNotification>`.
pub type HealthCheckerNetworkEvents = NetworkEvents<HealthCheckerMsg>;
/// The interface from HealthChecker to Networking layer.
///
/// This is a thin wrapper around a `NetworkSender<HealthCheckerMsg>`, so it is
/// easy to clone and send off to a separate task. For example, the rpc requests
/// return Futures that encapsulate the whole flow, from sending the request to
/// remote, to finally receiving the response and deserializing. It therefore
/// makes the most sense to make the rpc call on a separate async task, which
/// requires the `HealthCheckerNetworkSender` to be `Clone` and `Send`.
#[derive(Clone)]
pub struct HealthCheckerNetworkSender {
inner: NetworkSender<HealthCheckerMsg>,
}
/// Configuration for the network endpoints to support HealthChecker.
pub fn network_endpoint_config() -> (
Vec<ProtocolId>,
Vec<ProtocolId>,
QueueStyle,
usize,
Option<&'static IntCounterVec>,
) {
(
vec![ProtocolId::HealthCheckerRpc],
vec![],
QueueStyle::LIFO,
NETWORK_CHANNEL_SIZE,
Some(&counters::PENDING_HEALTH_CHECKER_NETWORK_EVENTS),
)
}
impl NewNetworkSender for HealthCheckerNetworkSender {
fn new(
peer_mgr_reqs_tx: PeerManagerRequestSender,
connection_reqs_tx: ConnectionRequestSender,
) -> Self {
Self {
inner: NetworkSender::new(peer_mgr_reqs_tx, connection_reqs_tx),
}
}
}
impl HealthCheckerNetworkSender {
/// Send a HealthChecker Ping RPC request to remote peer `recipient`. Returns
/// the remote peer's future `Pong` reply.
///
/// The rpc request can be canceled at any point by dropping the returned
/// future.
pub async fn send_rpc(
&mut self,
recipient: PeerId,
req_msg: HealthCheckerMsg,
timeout: Duration,
) -> Result<HealthCheckerMsg, RpcError> {
let protocol = ProtocolId::HealthCheckerRpc;
self.inner
.send_rpc(recipient, protocol, req_msg, timeout)
.await
}
pub async fn disconnect_peer(&mut self, peer_id: PeerId) -> Result<(), NetworkError> {
self.inner.disconnect_peer(peer_id).await
}
}
#[derive(Clone, Debug, Deserialize, Serialize)]
pub enum HealthCheckerMsg {
Ping(Ping),
Pong(Pong),
}
#[derive(Clone, Debug, Deserialize, Serialize)]
pub struct Ping(u32);
#[derive(Clone, Debug, Deserialize, Serialize)]
pub struct Pong(u32);
/// The actor performing health checks by running the Ping protocol
pub struct HealthChecker {
network_context: Arc<NetworkContext>,
/// A handle to a time service for easily mocking time-related operations.
time_service: TimeService,
/// Network interface to send requests to the Network Layer
network_interface: HealthCheckNetworkInterface,
/// Random-number generator.
rng: SmallRng,
/// Time we wait between each set of pings.
ping_interval: Duration,
/// Ping timeout duration.
ping_timeout: Duration,
/// Number of successive ping failures we tolerate before declaring a node as unhealthy and
/// disconnecting from it. In the future, this can be replaced with a more general failure
/// detection policy.
ping_failures_tolerated: u64,
/// Counter incremented in each round of health checks
round: u64,
}
impl HealthChecker {
/// Create new instance of the [`HealthChecker`] actor.
pub fn new(
network_context: Arc<NetworkContext>,
time_service: TimeService,
network_interface: HealthCheckNetworkInterface,
ping_interval: Duration,
ping_timeout: Duration,
ping_failures_tolerated: u64,
) -> Self {
HealthChecker {
network_context,
time_service,
network_interface,
rng: SmallRng::from_entropy(),
ping_interval,
ping_timeout,
ping_failures_tolerated,
round: 0,
}
}
pub async fn start(mut self) {
let mut tick_handlers = FuturesUnordered::new();
info!(
NetworkSchema::new(&self.network_context),
"{} Health checker actor started", self.network_context
);
let ticker = self.time_service.interval(self.ping_interval);
tokio::pin!(ticker);
loop {
futures::select! {
maybe_event = self.network_interface.next() => {
// Shutdown the HealthChecker when this network instance shuts
// down. This happens when the `PeerManager` drops.
let event = match maybe_event {
Some(event) => event,
None => break,
};
match event {
Event::NewPeer(metadata) => {
self.network_interface.insert_app_data(
metadata.remote_peer_id,
HealthCheckData::new(self.round)
);
}
Event::LostPeer(metadata) => {
self.network_interface.remove_app_data(
&metadata.remote_peer_id
);
}
Event::RpcRequest(peer_id, msg, res_tx) => {
match msg {
HealthCheckerMsg::Ping(ping) => self.handle_ping_request(peer_id, ping, res_tx),
_ => {
warn!(
SecurityEvent::InvalidHealthCheckerMsg,
NetworkSchema::new(&self.network_context).remote_peer(&peer_id),
rpc_message = msg,
"{} Unexpected RPC message from {}",
self.network_context,
peer_id
);
debug_assert!(false, "Unexpected rpc request");
}
};
}
Event::Message(peer_id, msg) => {
error!(
SecurityEvent::InvalidNetworkEventHC,
NetworkSchema::new(&self.network_context).remote_peer(&peer_id),
"{} Unexpected direct send from {} msg {:?}",
self.network_context,
peer_id,
msg,
);
debug_assert!(false, "Unexpected network event");
}
}
}
_ = ticker.select_next_some() => {
self.round += 1;
let connected = self.network_interface.connected_peers();
if connected.is_empty() {
trace!(
NetworkSchema::new(&self.network_context),
round = self.round,
"{} No connected peer to ping round: {}",
self.network_context,
self.round
);
continue
}
for peer_id in connected {
let nonce = self.rng.gen::<u32>();
trace!(
NetworkSchema::new(&self.network_context),
round = self.round,
"{} Will ping: {} for round: {} nonce: {}",
self.network_context,
peer_id.short_str(),
self.round,
nonce
);
tick_handlers.push(Self::ping_peer(
self.network_context.clone(),
self.network_interface.sender(),
peer_id,
self.round,
nonce,
self.ping_timeout,
));
}
}
res = tick_handlers.select_next_some() => {
let (peer_id, round, nonce, ping_result) = res;
self.handle_ping_response(peer_id, round, nonce, ping_result).await;
}
}
}
warn!(
NetworkSchema::new(&self.network_context),
"{} Health checker actor terminated", self.network_context
);
}
fn handle_ping_request(
&mut self,
peer_id: PeerId,
ping: Ping,
res_tx: oneshot::Sender<Result<Bytes, RpcError>>,
) {
let message = match bcs::to_bytes(&HealthCheckerMsg::Pong(Pong(ping.0))) {
Ok(msg) => msg,
Err(e) => {
warn!(
NetworkSchema::new(&self.network_context),
error = ?e,
"{} Unable to serialize pong response: {}", self.network_context, e
);
return;
}
};
trace!(
NetworkSchema::new(&self.network_context).remote_peer(&peer_id),
"{} Sending Pong response to peer: {} with nonce: {}",
self.network_context,
peer_id.short_str(),
ping.0,
);
let _ = res_tx.send(Ok(message.into()));
}
async fn handle_ping_response(
&mut self,
peer_id: PeerId,
round: u64,
req_nonce: u32,
ping_result: Result<Pong, RpcError>,
) {
match ping_result {
Ok(pong) => {
if pong.0 == req_nonce {
trace!(
NetworkSchema::new(&self.network_context).remote_peer(&peer_id),
rount = round,
"{} Ping successful for peer: {} round: {}",
self.network_context,
peer_id.short_str(),
round
);
// Update last successful ping to current round.
// If it's not in storage, don't bother updating it
let _ = self.network_interface.write_app_data(peer_id, |entry| {
match entry {
Entry::Vacant(..) => {
// Don't do anything if there isn't an entry
}
Entry::Occupied(inner) => {
let data = inner.get_mut();
// Update state if it's a newer round
if round > data.round {
data.round = round;
data.failures = 0;
}
}
};
Ok(())
});
} else {
warn!(
SecurityEvent::InvalidHealthCheckerMsg,
NetworkSchema::new(&self.network_context).remote_peer(&peer_id),
"{} Pong nonce doesn't match Ping nonce. Round: {}, Pong: {}, Ping: {}",
self.network_context,
round,
pong.0,
req_nonce
);
debug_assert!(false, "Pong nonce doesn't match our challenge Ping nonce");
}
}
Err(err) => {
warn!(
NetworkSchema::new(&self.network_context)
.remote_peer(&peer_id),
error = ?err,
round = round,
"{} Ping failed for peer: {} round: {} with error: {:?}",
self.network_context,
peer_id.short_str(),
round,
err
);
let _ = self.network_interface.write_app_data(peer_id, |entry| {
// Don't add in a default in case it's already disconnected
match entry {
Entry::Vacant(..) => {
// Don't do anything if there isn't an entry
}
Entry::Occupied(inner) => {
// If this is the result of an older ping, we ignore it.
// Increment num of failures.
let data = inner.get_mut();
if data.round <= round {
data.failures += 1;
}
}
}
Ok(())
});
// If the ping failures are now more than
// `self.ping_failures_tolerated`, we disconnect from the node.
// The HealthChecker only performs the disconnect. It relies on
// ConnectivityManager or the remote peer to re-establish the connection.
let failures = self
.network_interface
.read_app_data(&peer_id)
.map(|data| data.failures)
.unwrap_or(0);
if failures > self.ping_failures_tolerated {
info!(
NetworkSchema::new(&self.network_context).remote_peer(&peer_id),
"{} Disconnecting from peer: {}",
self.network_context,
peer_id.short_str()
);
if let Err(err) = self.network_interface.disconnect_peer(peer_id).await {
warn!(
NetworkSchema::new(&self.network_context)
.remote_peer(&peer_id),
error = ?err,
"{} Failed to disconnect from peer: {} with error: {:?}",
self.network_context,
peer_id.short_str(),
err
);
}
}
}
}
}
async fn ping_peer(
network_context: Arc<NetworkContext>,
mut network_tx: HealthCheckerNetworkSender,
peer_id: PeerId,
round: u64,
nonce: u32,
ping_timeout: Duration,
) -> (PeerId, u64, u32, Result<Pong, RpcError>) {
trace!(
NetworkSchema::new(&network_context).remote_peer(&peer_id),
round = round,
"{} Sending Ping request to peer: {} for round: {} nonce: {}",
network_context,
peer_id.short_str(),
round,
nonce
);
let res_pong_msg = network_tx
.send_rpc(peer_id, HealthCheckerMsg::Ping(Ping(nonce)), ping_timeout)
.await
.and_then(|msg| match msg {
HealthCheckerMsg::Pong(res) => Ok(res),
_ => Err(RpcError::InvalidRpcResponse),
});
(peer_id, round, nonce, res_pong_msg)
}
}