Line data Source code
1 : //! Safekeeper communication endpoint to WAL proposer (compute node).
2 : //! Gets messages from the network, passes them down to consensus module and
3 : //! sends replies back.
4 :
5 : use crate::handler::SafekeeperPostgresHandler;
6 : use crate::metrics::{
7 : WAL_RECEIVERS, WAL_RECEIVER_QUEUE_DEPTH, WAL_RECEIVER_QUEUE_DEPTH_TOTAL,
8 : WAL_RECEIVER_QUEUE_SIZE_TOTAL,
9 : };
10 : use crate::safekeeper::AcceptorProposerMessage;
11 : use crate::safekeeper::ProposerAcceptorMessage;
12 : use crate::safekeeper::ServerInfo;
13 : use crate::timeline::WalResidentTimeline;
14 : use crate::wal_service::ConnectionId;
15 : use crate::GlobalTimelines;
16 : use anyhow::{anyhow, Context};
17 : use bytes::BytesMut;
18 : use parking_lot::MappedMutexGuard;
19 : use parking_lot::Mutex;
20 : use parking_lot::MutexGuard;
21 : use postgres_backend::CopyStreamHandlerEnd;
22 : use postgres_backend::PostgresBackend;
23 : use postgres_backend::PostgresBackendReader;
24 : use postgres_backend::QueryError;
25 : use pq_proto::BeMessage;
26 : use serde::Deserialize;
27 : use serde::Serialize;
28 : use std::future;
29 : use std::net::SocketAddr;
30 : use std::sync::Arc;
31 : use tokio::io::AsyncRead;
32 : use tokio::io::AsyncWrite;
33 : use tokio::sync::mpsc::error::SendTimeoutError;
34 : use tokio::sync::mpsc::{channel, Receiver, Sender};
35 : use tokio::task;
36 : use tokio::task::JoinHandle;
37 : use tokio::time::{Duration, Instant, MissedTickBehavior};
38 : use tracing::*;
39 : use utils::id::TenantTimelineId;
40 : use utils::lsn::Lsn;
41 : use utils::pageserver_feedback::PageserverFeedback;
42 :
43 : const DEFAULT_FEEDBACK_CAPACITY: usize = 8;
44 :
45 : /// Registry of WalReceivers (compute connections). Timeline holds it (wrapped
46 : /// in Arc).
47 : pub struct WalReceivers {
48 : mutex: Mutex<WalReceiversShared>,
49 : pageserver_feedback_tx: tokio::sync::broadcast::Sender<PageserverFeedback>,
50 :
51 : num_computes_tx: tokio::sync::watch::Sender<usize>,
52 : num_computes_rx: tokio::sync::watch::Receiver<usize>,
53 : }
54 :
55 : /// Id under which walreceiver is registered in shmem.
56 : type WalReceiverId = usize;
57 :
58 : impl WalReceivers {
59 0 : pub fn new() -> Arc<WalReceivers> {
60 0 : let (pageserver_feedback_tx, _) =
61 0 : tokio::sync::broadcast::channel(DEFAULT_FEEDBACK_CAPACITY);
62 0 :
63 0 : let (num_computes_tx, num_computes_rx) = tokio::sync::watch::channel(0usize);
64 0 :
65 0 : Arc::new(WalReceivers {
66 0 : mutex: Mutex::new(WalReceiversShared { slots: Vec::new() }),
67 0 : pageserver_feedback_tx,
68 0 : num_computes_tx,
69 0 : num_computes_rx,
70 0 : })
71 0 : }
72 :
73 : /// Register new walreceiver. Returned guard provides access to the slot and
74 : /// automatically deregisters in Drop.
75 0 : pub fn register(self: &Arc<WalReceivers>, conn_id: Option<ConnectionId>) -> WalReceiverGuard {
76 0 : let mut shared = self.mutex.lock();
77 0 : let slots = &mut shared.slots;
78 0 : let walreceiver = WalReceiverState {
79 0 : conn_id,
80 0 : status: WalReceiverStatus::Voting,
81 0 : };
82 : // find empty slot or create new one
83 0 : let pos = if let Some(pos) = slots.iter().position(|s| s.is_none()) {
84 0 : slots[pos] = Some(walreceiver);
85 0 : pos
86 : } else {
87 0 : let pos = slots.len();
88 0 : slots.push(Some(walreceiver));
89 0 : pos
90 : };
91 :
92 0 : self.update_num(&shared);
93 0 : WAL_RECEIVERS.inc();
94 0 :
95 0 : WalReceiverGuard {
96 0 : id: pos,
97 0 : walreceivers: self.clone(),
98 0 : }
99 0 : }
100 :
101 : /// Get reference to locked slot contents. Slot must exist (registered
102 : /// earlier).
103 0 : fn get_slot<'a>(
104 0 : self: &'a Arc<WalReceivers>,
105 0 : id: WalReceiverId,
106 0 : ) -> MappedMutexGuard<'a, WalReceiverState> {
107 0 : MutexGuard::map(self.mutex.lock(), |locked| {
108 0 : locked.slots[id]
109 0 : .as_mut()
110 0 : .expect("walreceiver doesn't exist")
111 0 : })
112 0 : }
113 :
114 : /// Get number of walreceivers (compute connections).
115 0 : pub fn get_num(self: &Arc<WalReceivers>) -> usize {
116 0 : self.mutex.lock().get_num()
117 0 : }
118 :
119 : /// Get channel for number of walreceivers.
120 0 : pub fn get_num_rx(self: &Arc<WalReceivers>) -> tokio::sync::watch::Receiver<usize> {
121 0 : self.num_computes_rx.clone()
122 0 : }
123 :
124 : /// Should get called after every update of slots.
125 0 : fn update_num(self: &Arc<WalReceivers>, shared: &MutexGuard<WalReceiversShared>) {
126 0 : let num = shared.get_num();
127 0 : self.num_computes_tx.send_replace(num);
128 0 : }
129 :
130 : /// Get state of all walreceivers.
131 0 : pub fn get_all(self: &Arc<WalReceivers>) -> Vec<WalReceiverState> {
132 0 : self.mutex.lock().slots.iter().flatten().cloned().collect()
133 0 : }
134 :
135 : /// Get number of streaming walreceivers (normally 0 or 1) from compute.
136 0 : pub fn get_num_streaming(self: &Arc<WalReceivers>) -> usize {
137 0 : self.mutex
138 0 : .lock()
139 0 : .slots
140 0 : .iter()
141 0 : .flatten()
142 0 : // conn_id.is_none skips recovery which also registers here
143 0 : .filter(|s| s.conn_id.is_some() && matches!(s.status, WalReceiverStatus::Streaming))
144 0 : .count()
145 0 : }
146 :
147 : /// Unregister walreceiver.
148 0 : fn unregister(self: &Arc<WalReceivers>, id: WalReceiverId) {
149 0 : let mut shared = self.mutex.lock();
150 0 : shared.slots[id] = None;
151 0 : self.update_num(&shared);
152 0 : WAL_RECEIVERS.dec();
153 0 : }
154 :
155 : /// Broadcast pageserver feedback to connected walproposers.
156 0 : pub fn broadcast_pageserver_feedback(&self, feedback: PageserverFeedback) {
157 0 : // Err means there is no subscribers, it is fine.
158 0 : let _ = self.pageserver_feedback_tx.send(feedback);
159 0 : }
160 : }
161 :
162 : /// Only a few connections are expected (normally one), so store in Vec.
163 : struct WalReceiversShared {
164 : slots: Vec<Option<WalReceiverState>>,
165 : }
166 :
167 : impl WalReceiversShared {
168 : /// Get number of walreceivers (compute connections).
169 0 : fn get_num(&self) -> usize {
170 0 : self.slots.iter().flatten().count()
171 0 : }
172 : }
173 :
174 0 : #[derive(Debug, Clone, Serialize, Deserialize)]
175 : pub struct WalReceiverState {
176 : /// None means it is recovery initiated by us (this safekeeper).
177 : pub conn_id: Option<ConnectionId>,
178 : pub status: WalReceiverStatus,
179 : }
180 :
181 : /// Walreceiver status. Currently only whether it passed voting stage and
182 : /// started receiving the stream, but it is easy to add more if needed.
183 0 : #[derive(Debug, Clone, Serialize, Deserialize)]
184 : pub enum WalReceiverStatus {
185 : Voting,
186 : Streaming,
187 : }
188 :
189 : /// Scope guard to access slot in WalReceivers registry and unregister from
190 : /// it in Drop.
191 : pub struct WalReceiverGuard {
192 : id: WalReceiverId,
193 : walreceivers: Arc<WalReceivers>,
194 : }
195 :
196 : impl WalReceiverGuard {
197 : /// Get reference to locked shared state contents.
198 0 : fn get(&self) -> MappedMutexGuard<WalReceiverState> {
199 0 : self.walreceivers.get_slot(self.id)
200 0 : }
201 : }
202 :
203 : impl Drop for WalReceiverGuard {
204 0 : fn drop(&mut self) {
205 0 : self.walreceivers.unregister(self.id);
206 0 : }
207 : }
208 :
209 : pub const MSG_QUEUE_SIZE: usize = 256;
210 : pub const REPLY_QUEUE_SIZE: usize = 16;
211 :
212 : impl SafekeeperPostgresHandler {
213 : /// Wrapper around handle_start_wal_push_guts handling result. Error is
214 : /// handled here while we're still in walreceiver ttid span; with API
215 : /// extension, this can probably be moved into postgres_backend.
216 0 : pub async fn handle_start_wal_push<IO: AsyncRead + AsyncWrite + Unpin>(
217 0 : &mut self,
218 0 : pgb: &mut PostgresBackend<IO>,
219 0 : ) -> Result<(), QueryError> {
220 0 : let mut tli: Option<WalResidentTimeline> = None;
221 0 : if let Err(end) = self.handle_start_wal_push_guts(pgb, &mut tli).await {
222 : // Log the result and probably send it to the client, closing the stream.
223 0 : let handle_end_fut = pgb.handle_copy_stream_end(end);
224 : // If we managed to create the timeline, augment logging with current LSNs etc.
225 0 : if let Some(tli) = tli {
226 0 : let info = tli.get_safekeeper_info(&self.conf).await;
227 0 : handle_end_fut
228 0 : .instrument(info_span!("", term=%info.term, last_log_term=%info.last_log_term, flush_lsn=%Lsn(info.flush_lsn), commit_lsn=%Lsn(info.commit_lsn)))
229 0 : .await;
230 : } else {
231 0 : handle_end_fut.await;
232 : }
233 0 : }
234 0 : Ok(())
235 0 : }
236 :
237 0 : pub async fn handle_start_wal_push_guts<IO: AsyncRead + AsyncWrite + Unpin>(
238 0 : &mut self,
239 0 : pgb: &mut PostgresBackend<IO>,
240 0 : tli: &mut Option<WalResidentTimeline>,
241 0 : ) -> Result<(), CopyStreamHandlerEnd> {
242 0 : // Notify the libpq client that it's allowed to send `CopyData` messages
243 0 : pgb.write_message(&BeMessage::CopyBothResponse).await?;
244 :
245 : // Experiments [1] confirm that doing network IO in one (this) thread and
246 : // processing with disc IO in another significantly improves
247 : // performance; we spawn off WalAcceptor thread for message processing
248 : // to this end.
249 : //
250 : // [1] https://github.com/neondatabase/neon/pull/1318
251 0 : let (msg_tx, msg_rx) = channel(MSG_QUEUE_SIZE);
252 0 : let (reply_tx, reply_rx) = channel(REPLY_QUEUE_SIZE);
253 0 : let mut acceptor_handle: Option<JoinHandle<anyhow::Result<()>>> = None;
254 :
255 : // Concurrently receive and send data; replies are not synchronized with
256 : // sends, so this avoids deadlocks.
257 0 : let mut pgb_reader = pgb.split().context("START_WAL_PUSH split")?;
258 0 : let peer_addr = *pgb.get_peer_addr();
259 0 : let mut network_reader = NetworkReader {
260 0 : ttid: self.ttid,
261 0 : conn_id: self.conn_id,
262 0 : pgb_reader: &mut pgb_reader,
263 0 : peer_addr,
264 0 : acceptor_handle: &mut acceptor_handle,
265 0 : };
266 :
267 : // Read first message and create timeline if needed.
268 0 : let res = network_reader.read_first_message().await;
269 :
270 0 : let network_res = if let Ok((timeline, next_msg)) = res {
271 0 : let pageserver_feedback_rx: tokio::sync::broadcast::Receiver<PageserverFeedback> =
272 0 : timeline
273 0 : .get_walreceivers()
274 0 : .pageserver_feedback_tx
275 0 : .subscribe();
276 0 : *tli = Some(timeline.wal_residence_guard().await?);
277 :
278 0 : tokio::select! {
279 : // todo: add read|write .context to these errors
280 0 : r = network_reader.run(msg_tx, msg_rx, reply_tx, timeline, next_msg) => r,
281 0 : r = network_write(pgb, reply_rx, pageserver_feedback_rx) => r,
282 : }
283 : } else {
284 0 : res.map(|_| ())
285 : };
286 :
287 : // Join pg backend back.
288 0 : pgb.unsplit(pgb_reader)?;
289 :
290 : // Join the spawned WalAcceptor. At this point chans to/from it passed
291 : // to network routines are dropped, so it will exit as soon as it
292 : // touches them.
293 0 : match acceptor_handle {
294 : None => {
295 : // failed even before spawning; read_network should have error
296 0 : Err(network_res.expect_err("no error with WalAcceptor not spawn"))
297 : }
298 0 : Some(handle) => {
299 0 : let wal_acceptor_res = handle.await;
300 :
301 : // If there was any network error, return it.
302 0 : network_res?;
303 :
304 : // Otherwise, WalAcceptor thread must have errored.
305 0 : match wal_acceptor_res {
306 0 : Ok(Ok(_)) => Ok(()), // can't happen currently; would be if we add graceful termination
307 0 : Ok(Err(e)) => Err(CopyStreamHandlerEnd::Other(e.context("WAL acceptor"))),
308 0 : Err(_) => Err(CopyStreamHandlerEnd::Other(anyhow!(
309 0 : "WalAcceptor task panicked",
310 0 : ))),
311 : }
312 : }
313 : }
314 0 : }
315 : }
316 :
317 : struct NetworkReader<'a, IO> {
318 : ttid: TenantTimelineId,
319 : conn_id: ConnectionId,
320 : pgb_reader: &'a mut PostgresBackendReader<IO>,
321 : peer_addr: SocketAddr,
322 : // WalAcceptor is spawned when we learn server info from walproposer and
323 : // create timeline; handle is put here.
324 : acceptor_handle: &'a mut Option<JoinHandle<anyhow::Result<()>>>,
325 : }
326 :
327 : impl<'a, IO: AsyncRead + AsyncWrite + Unpin> NetworkReader<'a, IO> {
328 0 : async fn read_first_message(
329 0 : &mut self,
330 0 : ) -> Result<(WalResidentTimeline, ProposerAcceptorMessage), CopyStreamHandlerEnd> {
331 : // Receive information about server to create timeline, if not yet.
332 0 : let next_msg = read_message(self.pgb_reader).await?;
333 0 : let tli = match next_msg {
334 0 : ProposerAcceptorMessage::Greeting(ref greeting) => {
335 0 : info!(
336 0 : "start handshake with walproposer {} sysid {} timeline {}",
337 : self.peer_addr, greeting.system_id, greeting.tli,
338 : );
339 0 : let server_info = ServerInfo {
340 0 : pg_version: greeting.pg_version,
341 0 : system_id: greeting.system_id,
342 0 : wal_seg_size: greeting.wal_seg_size,
343 0 : };
344 0 : let tli =
345 0 : GlobalTimelines::create(self.ttid, server_info, Lsn::INVALID, Lsn::INVALID)
346 0 : .await
347 0 : .context("create timeline")?;
348 0 : tli.wal_residence_guard().await?
349 : }
350 : _ => {
351 0 : return Err(CopyStreamHandlerEnd::Other(anyhow::anyhow!(
352 0 : "unexpected message {next_msg:?} instead of greeting"
353 0 : )))
354 : }
355 : };
356 0 : Ok((tli, next_msg))
357 0 : }
358 :
359 0 : async fn run(
360 0 : self,
361 0 : msg_tx: Sender<ProposerAcceptorMessage>,
362 0 : msg_rx: Receiver<ProposerAcceptorMessage>,
363 0 : reply_tx: Sender<AcceptorProposerMessage>,
364 0 : tli: WalResidentTimeline,
365 0 : next_msg: ProposerAcceptorMessage,
366 0 : ) -> Result<(), CopyStreamHandlerEnd> {
367 0 : *self.acceptor_handle = Some(WalAcceptor::spawn(
368 0 : tli,
369 0 : msg_rx,
370 0 : reply_tx,
371 0 : Some(self.conn_id),
372 0 : ));
373 0 :
374 0 : // Forward all messages to WalAcceptor
375 0 : read_network_loop(self.pgb_reader, msg_tx, next_msg).await
376 0 : }
377 : }
378 :
379 : /// Read next message from walproposer.
380 : /// TODO: Return Ok(None) on graceful termination.
381 0 : async fn read_message<IO: AsyncRead + AsyncWrite + Unpin>(
382 0 : pgb_reader: &mut PostgresBackendReader<IO>,
383 0 : ) -> Result<ProposerAcceptorMessage, CopyStreamHandlerEnd> {
384 0 : let copy_data = pgb_reader.read_copy_message().await?;
385 0 : let msg = ProposerAcceptorMessage::parse(copy_data)?;
386 0 : Ok(msg)
387 0 : }
388 :
389 0 : async fn read_network_loop<IO: AsyncRead + AsyncWrite + Unpin>(
390 0 : pgb_reader: &mut PostgresBackendReader<IO>,
391 0 : msg_tx: Sender<ProposerAcceptorMessage>,
392 0 : mut next_msg: ProposerAcceptorMessage,
393 0 : ) -> Result<(), CopyStreamHandlerEnd> {
394 : /// Threshold for logging slow WalAcceptor sends.
395 : const SLOW_THRESHOLD: Duration = Duration::from_secs(5);
396 :
397 : loop {
398 0 : let started = Instant::now();
399 0 : let size = next_msg.size();
400 0 : match msg_tx.send_timeout(next_msg, SLOW_THRESHOLD).await {
401 0 : Ok(()) => {}
402 : // Slow send, log a message and keep trying. Log context has timeline ID.
403 0 : Err(SendTimeoutError::Timeout(next_msg)) => {
404 0 : warn!(
405 0 : "slow WalAcceptor send blocked for {:.3}s",
406 0 : Instant::now().duration_since(started).as_secs_f64()
407 : );
408 0 : if msg_tx.send(next_msg).await.is_err() {
409 0 : return Ok(()); // WalAcceptor terminated
410 0 : }
411 0 : warn!(
412 0 : "slow WalAcceptor send completed after {:.3}s",
413 0 : Instant::now().duration_since(started).as_secs_f64()
414 : )
415 : }
416 : // WalAcceptor terminated.
417 0 : Err(SendTimeoutError::Closed(_)) => return Ok(()),
418 : }
419 :
420 : // Update metrics. Will be decremented in WalAcceptor.
421 0 : WAL_RECEIVER_QUEUE_DEPTH_TOTAL.inc();
422 0 : WAL_RECEIVER_QUEUE_SIZE_TOTAL.add(size as i64);
423 :
424 0 : next_msg = read_message(pgb_reader).await?;
425 : }
426 0 : }
427 :
428 : /// Read replies from WalAcceptor and pass them back to socket. Returns Ok(())
429 : /// if reply_rx closed; it must mean WalAcceptor terminated, joining it should
430 : /// tell the error.
431 0 : async fn network_write<IO: AsyncRead + AsyncWrite + Unpin>(
432 0 : pgb_writer: &mut PostgresBackend<IO>,
433 0 : mut reply_rx: Receiver<AcceptorProposerMessage>,
434 0 : mut pageserver_feedback_rx: tokio::sync::broadcast::Receiver<PageserverFeedback>,
435 0 : ) -> Result<(), CopyStreamHandlerEnd> {
436 0 : let mut buf = BytesMut::with_capacity(128);
437 0 :
438 0 : // storing append_response to inject PageserverFeedback into it
439 0 : let mut last_append_response = None;
440 :
441 : loop {
442 : // trying to read either AcceptorProposerMessage or PageserverFeedback
443 0 : let msg = tokio::select! {
444 0 : reply = reply_rx.recv() => {
445 0 : if let Some(msg) = reply {
446 0 : if let AcceptorProposerMessage::AppendResponse(append_response) = &msg {
447 0 : last_append_response = Some(append_response.clone());
448 0 : }
449 0 : Some(msg)
450 : } else {
451 0 : return Ok(()); // chan closed, WalAcceptor terminated
452 : }
453 : }
454 :
455 0 : feedback = pageserver_feedback_rx.recv() =>
456 0 : match (feedback, &last_append_response) {
457 0 : (Ok(feedback), Some(append_response)) => {
458 0 : // clone AppendResponse and inject PageserverFeedback into it
459 0 : let mut append_response = append_response.clone();
460 0 : append_response.pageserver_feedback = Some(feedback);
461 0 : Some(AcceptorProposerMessage::AppendResponse(append_response))
462 : }
463 0 : _ => None,
464 : }
465 : };
466 :
467 0 : let Some(msg) = msg else {
468 0 : continue;
469 : };
470 :
471 0 : buf.clear();
472 0 : msg.serialize(&mut buf)?;
473 0 : pgb_writer.write_message(&BeMessage::CopyData(&buf)).await?;
474 : }
475 0 : }
476 :
477 : /// The WAL flush interval. This ensures we periodically flush the WAL and send AppendResponses to
478 : /// walproposer, even when it's writing a steady stream of messages.
479 : const FLUSH_INTERVAL: Duration = Duration::from_secs(1);
480 :
481 : /// The metrics computation interval.
482 : ///
483 : /// The Prometheus poll interval is 60 seconds at the time of writing. We sample the queue depth
484 : /// every 5 seconds, for 12 samples per poll. This will give a count of up to 12x active timelines.
485 : const METRICS_INTERVAL: Duration = Duration::from_secs(5);
486 :
487 : /// Encapsulates a task which takes messages from msg_rx, processes and pushes
488 : /// replies to reply_tx.
489 : ///
490 : /// Reading from socket and writing to disk in parallel is beneficial for
491 : /// performance, this struct provides the writing to disk part.
492 : pub struct WalAcceptor {
493 : tli: WalResidentTimeline,
494 : msg_rx: Receiver<ProposerAcceptorMessage>,
495 : reply_tx: Sender<AcceptorProposerMessage>,
496 : conn_id: Option<ConnectionId>,
497 : }
498 :
499 : impl WalAcceptor {
500 : /// Spawn task with WalAcceptor running, return handle to it. Task returns
501 : /// Ok(()) if either of channels has closed, and Err if any error during
502 : /// message processing is encountered.
503 : ///
504 : /// conn_id None means WalAcceptor is used by recovery initiated at this safekeeper.
505 0 : pub fn spawn(
506 0 : tli: WalResidentTimeline,
507 0 : msg_rx: Receiver<ProposerAcceptorMessage>,
508 0 : reply_tx: Sender<AcceptorProposerMessage>,
509 0 : conn_id: Option<ConnectionId>,
510 0 : ) -> JoinHandle<anyhow::Result<()>> {
511 0 : task::spawn(async move {
512 0 : let mut wa = WalAcceptor {
513 0 : tli,
514 0 : msg_rx,
515 0 : reply_tx,
516 0 : conn_id,
517 0 : };
518 0 :
519 0 : let span_ttid = wa.tli.ttid; // satisfy borrow checker
520 0 : wa.run()
521 0 : .instrument(
522 0 : info_span!("WAL acceptor", cid = %conn_id.unwrap_or(0), ttid = %span_ttid),
523 : )
524 0 : .await
525 0 : })
526 0 : }
527 :
528 : /// The main loop. Returns Ok(()) if either msg_rx or reply_tx got closed;
529 : /// it must mean that network thread terminated.
530 0 : async fn run(&mut self) -> anyhow::Result<()> {
531 0 : let walreceiver_guard = self.tli.get_walreceivers().register(self.conn_id);
532 0 :
533 0 : // Periodically flush the WAL and compute metrics.
534 0 : let mut flush_ticker = tokio::time::interval(FLUSH_INTERVAL);
535 0 : flush_ticker.set_missed_tick_behavior(MissedTickBehavior::Delay);
536 0 : flush_ticker.tick().await; // skip the initial, immediate tick
537 :
538 0 : let mut metrics_ticker = tokio::time::interval(METRICS_INTERVAL);
539 0 : metrics_ticker.set_missed_tick_behavior(MissedTickBehavior::Skip);
540 0 :
541 0 : // Tracks whether we have unflushed appends.
542 0 : let mut dirty = false;
543 :
544 : loop {
545 0 : let reply = tokio::select! {
546 : // Process inbound message.
547 0 : msg = self.msg_rx.recv() => {
548 : // If disconnected, break to flush WAL and return.
549 0 : let Some(mut msg) = msg else {
550 0 : break;
551 : };
552 :
553 : // Update gauge metrics.
554 0 : WAL_RECEIVER_QUEUE_DEPTH_TOTAL.dec();
555 0 : WAL_RECEIVER_QUEUE_SIZE_TOTAL.sub(msg.size() as i64);
556 0 :
557 0 : // Update walreceiver state in shmem for reporting.
558 0 : if let ProposerAcceptorMessage::Elected(_) = &msg {
559 0 : walreceiver_guard.get().status = WalReceiverStatus::Streaming;
560 0 : }
561 :
562 : // Don't flush the WAL on every append, only periodically via flush_ticker.
563 : // This batches multiple appends per fsync. If the channel is empty after
564 : // sending the reply, we'll schedule an immediate flush.
565 : //
566 : // Note that a flush can still happen on segment bounds, which will result
567 : // in an AppendResponse.
568 0 : if let ProposerAcceptorMessage::AppendRequest(append_request) = msg {
569 0 : msg = ProposerAcceptorMessage::NoFlushAppendRequest(append_request);
570 0 : dirty = true;
571 0 : }
572 :
573 0 : self.tli.process_msg(&msg).await?
574 : }
575 :
576 : // While receiving AppendRequests, flush the WAL periodically and respond with an
577 : // AppendResponse to let walproposer know we're still alive.
578 0 : _ = flush_ticker.tick(), if dirty => {
579 0 : dirty = false;
580 0 : self.tli
581 0 : .process_msg(&ProposerAcceptorMessage::FlushWAL)
582 0 : .await?
583 : }
584 :
585 : // If there are no pending messages, flush the WAL immediately.
586 : //
587 : // TODO: this should be done via flush_ticker.reset_immediately(), but that's always
588 : // delayed by 1ms due to this bug: https://github.com/tokio-rs/tokio/issues/6866.
589 0 : _ = future::ready(()), if dirty && self.msg_rx.is_empty() => {
590 0 : dirty = false;
591 0 : flush_ticker.reset();
592 0 : self.tli
593 0 : .process_msg(&ProposerAcceptorMessage::FlushWAL)
594 0 : .await?
595 : }
596 :
597 : // Update histogram metrics periodically.
598 0 : _ = metrics_ticker.tick() => {
599 0 : WAL_RECEIVER_QUEUE_DEPTH.observe(self.msg_rx.len() as f64);
600 0 : None // no reply
601 : }
602 : };
603 :
604 : // Send reply, if any.
605 0 : if let Some(reply) = reply {
606 0 : if self.reply_tx.send(reply).await.is_err() {
607 0 : break; // disconnected, break to flush WAL and return
608 0 : }
609 0 : }
610 : }
611 :
612 : // Flush WAL on disconnect, see https://github.com/neondatabase/neon/issues/9259.
613 0 : if dirty {
614 0 : self.tli
615 0 : .process_msg(&ProposerAcceptorMessage::FlushWAL)
616 0 : .await?;
617 0 : }
618 :
619 0 : Ok(())
620 0 : }
621 : }
622 :
623 : /// On drop, drain msg_rx and update metrics to avoid leaks.
624 : impl Drop for WalAcceptor {
625 0 : fn drop(&mut self) {
626 0 : self.msg_rx.close(); // prevent further sends
627 0 : while let Ok(msg) = self.msg_rx.try_recv() {
628 0 : WAL_RECEIVER_QUEUE_DEPTH_TOTAL.dec();
629 0 : WAL_RECEIVER_QUEUE_SIZE_TOTAL.sub(msg.size() as i64);
630 0 : }
631 0 : }
632 : }
|
