neon/safekeeper/src/send_wal.rs

//! This module implements the streaming side of replication protocol, starting
//! with the "START_REPLICATION" message.

use crate::handler::SafekeeperPostgresHandler;
use crate::timeline::{ReplicaState, Timeline};
use crate::wal_storage::WalReader;
use crate::GlobalTimelines;
use anyhow::Context;

use bytes::Bytes;
use postgres_ffi::get_current_timestamp;
use postgres_ffi::{TimestampTz, MAX_SEND_SIZE};
use serde::{Deserialize, Serialize};
use std::cmp::min;
use std::net::Shutdown;
use std::sync::Arc;
use std::time::Duration;
use std::{io, str, thread};
use utils::postgres_backend_async::QueryError;

use pq_proto::{BeMessage, FeMessage, ReplicationFeedback, WalSndKeepAlive, XLogDataBody};
use tokio::sync::watch::Receiver;
use tokio::time::timeout;
use tracing::*;
use utils::{bin_ser::BeSer, lsn::Lsn, postgres_backend::PostgresBackend, sock_split::ReadStream};

// See: https://www.postgresql.org/docs/13/protocol-replication.html
const HOT_STANDBY_FEEDBACK_TAG_BYTE: u8 = b'h';
const STANDBY_STATUS_UPDATE_TAG_BYTE: u8 = b'r';
// neon extension of replication protocol
const NEON_STATUS_UPDATE_TAG_BYTE: u8 = b'z';

type FullTransactionId = u64;

/// Hot standby feedback received from replica
#[derive(Debug, Clone, Copy, Serialize, Deserialize)]
pub struct HotStandbyFeedback {
    pub ts: TimestampTz,
    pub xmin: FullTransactionId,
    pub catalog_xmin: FullTransactionId,
}

impl HotStandbyFeedback {
    pub fn empty() -> HotStandbyFeedback {
        HotStandbyFeedback {
            ts: 0,
            xmin: 0,
            catalog_xmin: 0,
        }
    }
}

/// Standby status update
#[derive(Debug, Clone, Deserialize)]
pub struct StandbyReply {
    pub write_lsn: Lsn, // last lsn received by pageserver
    pub flush_lsn: Lsn, // pageserver's disk consistent lSN
    pub apply_lsn: Lsn, // pageserver's remote consistent lSN
    pub reply_ts: TimestampTz,
    pub reply_requested: bool,
}

/// A network connection that's speaking the replication protocol.
pub struct ReplicationConn {
    /// This is an `Option` because we will spawn a background thread that will
    /// `take` it from us.
    stream_in: Option<ReadStream>,
}

/// Scope guard to unregister replication connection from timeline
struct ReplicationConnGuard {
    replica: usize, // replica internal ID assigned by timeline
    timeline: Arc<Timeline>,
}

impl Drop for ReplicationConnGuard {
    fn drop(&mut self) {
        self.timeline.remove_replica(self.replica);
    }
}

impl ReplicationConn {
    /// Create a new `ReplicationConn`
    pub fn new(pgb: &mut PostgresBackend) -> Self {
        Self {
            stream_in: pgb.take_stream_in(),
        }
    }

    /// Handle incoming messages from the network.
    /// This is spawned into the background by `handle_start_replication`.
    fn background_thread(
        mut stream_in: ReadStream,
        replica_guard: Arc<ReplicationConnGuard>,
    ) -> anyhow::Result<()> {
        let replica_id = replica_guard.replica;
        let timeline = &replica_guard.timeline;

        let mut state = ReplicaState::new();
        // Wait for replica's feedback.
        while let Some(msg) = FeMessage::read(&mut stream_in)? {
            match &msg {
                FeMessage::CopyData(m) => {
                    // There's three possible data messages that the client is supposed to send here:
                    // `HotStandbyFeedback` and `StandbyStatusUpdate` and `NeonStandbyFeedback`.

                    match m.first().cloned() {
                        Some(HOT_STANDBY_FEEDBACK_TAG_BYTE) => {
                            // Note: deserializing is on m[1..] because we skip the tag byte.
                            state.hs_feedback = HotStandbyFeedback::des(&m[1..])
                                .context("failed to deserialize HotStandbyFeedback")?;
                            timeline.update_replica_state(replica_id, state);
                        }
                        Some(STANDBY_STATUS_UPDATE_TAG_BYTE) => {
                            let _reply = StandbyReply::des(&m[1..])
                                .context("failed to deserialize StandbyReply")?;
                            // This must be a regular postgres replica,
                            // because pageserver doesn't send this type of messages to safekeeper.
                            // Currently this is not implemented, so this message is ignored.

                            warn!("unexpected StandbyReply. Read-only postgres replicas are not supported in safekeepers yet.");
                            // timeline.update_replica_state(replica_id, Some(state));
                        }
                        Some(NEON_STATUS_UPDATE_TAG_BYTE) => {
                            // Note: deserializing is on m[9..] because we skip the tag byte and len bytes.
                            let buf = Bytes::copy_from_slice(&m[9..]);
                            let reply = ReplicationFeedback::parse(buf);

                            trace!("ReplicationFeedback is {:?}", reply);
                            // Only pageserver sends ReplicationFeedback, so set the flag.
                            // This replica is the source of information to resend to compute.
                            state.pageserver_feedback = Some(reply);

                            timeline.update_replica_state(replica_id, state);
                        }
                        _ => warn!("unexpected message {:?}", msg),
                    }
                }
                FeMessage::Sync => {}
                FeMessage::CopyFail => {
                    // Shutdown the connection, because rust-postgres client cannot be dropped
                    // when connection is alive.
                    let _ = stream_in.shutdown(Shutdown::Both);
                    anyhow::bail!("Copy failed");
                }
                _ => {
                    // We only handle `CopyData`, 'Sync', 'CopyFail' messages. Anything else is ignored.
                    info!("unexpected message {:?}", msg);
                }
            }
        }

        Ok(())
    }

    ///
    /// Handle START_REPLICATION replication command
    ///
    pub fn run(
        &mut self,
        spg: &mut SafekeeperPostgresHandler,
        pgb: &mut PostgresBackend,
        mut start_pos: Lsn,
    ) -> Result<(), QueryError> {
        let _enter = info_span!("WAL sender", ttid = %spg.ttid).entered();

        let tli = GlobalTimelines::get(spg.ttid)?;

        // spawn the background thread which receives HotStandbyFeedback messages.
        let bg_timeline = Arc::clone(&tli);
        let bg_stream_in = self.stream_in.take().unwrap();
        let bg_timeline_id = spg.timeline_id.unwrap();

        let state = ReplicaState::new();
        // This replica_id is used below to check if it's time to stop replication.
        let replica_id = bg_timeline.add_replica(state);

        // Use a guard object to remove our entry from the timeline, when the background
        // thread and us have both finished using it.
        let replica_guard = Arc::new(ReplicationConnGuard {
            replica: replica_id,
            timeline: bg_timeline,
        });
        let bg_replica_guard = Arc::clone(&replica_guard);

        // TODO: here we got two threads, one for writing WAL and one for receiving
        // feedback. If one of them fails, we should shutdown the other one too.
        let _ = thread::Builder::new()
            .name("HotStandbyFeedback thread".into())
            .spawn(move || {
                let _enter =
                    info_span!("HotStandbyFeedback thread", timeline = %bg_timeline_id).entered();
                if let Err(err) = Self::background_thread(bg_stream_in, bg_replica_guard) {
                    error!("Replication background thread failed: {}", err);
                }
            })?;

        let runtime = tokio::runtime::Builder::new_current_thread()
            .enable_all()
            .build()?;

        runtime.block_on(async move {
            let (inmem_state, persisted_state) = tli.get_state();
            // add persisted_state.timeline_start_lsn == Lsn(0) check

            // Walproposer gets special handling: safekeeper must give proposer all
            // local WAL till the end, whether committed or not (walproposer will
            // hang otherwise). That's because walproposer runs the consensus and
            // synchronizes safekeepers on the most advanced one.
            //
            // There is a small risk of this WAL getting concurrently garbaged if
            // another compute rises which collects majority and starts fixing log
            // on this safekeeper itself. That's ok as (old) proposer will never be
            // able to commit such WAL.
            let stop_pos: Option<Lsn> = if spg.is_walproposer_recovery() {
                let wal_end = tli.get_flush_lsn();
                Some(wal_end)
            } else {
                None
            };

            info!("Start replication from {:?} till {:?}", start_pos, stop_pos);

            // switch to copy
            pgb.write_message(&BeMessage::CopyBothResponse)?;

            let mut end_pos = stop_pos.unwrap_or(inmem_state.commit_lsn);

            let mut wal_reader = WalReader::new(
                spg.conf.workdir.clone(),
                spg.conf.timeline_dir(&tli.ttid),
                &persisted_state,
                start_pos,
                spg.conf.wal_backup_enabled,
            )?;

            // buffer for wal sending, limited by MAX_SEND_SIZE
            let mut send_buf = vec![0u8; MAX_SEND_SIZE];

            // watcher for commit_lsn updates
            let mut commit_lsn_watch_rx = tli.get_commit_lsn_watch_rx();

            loop {
                if let Some(stop_pos) = stop_pos {
                    if start_pos >= stop_pos {
                        break; /* recovery finished */
                    }
                    end_pos = stop_pos;
                } else {
                    /* Wait until we have some data to stream */
                    let lsn = wait_for_lsn(&mut commit_lsn_watch_rx, start_pos).await?;

                    if let Some(lsn) = lsn {
                        end_pos = lsn;
                    } else {
                        // TODO: also check once in a while whether we are walsender
                        // to right pageserver.
                        if tli.should_walsender_stop(replica_id) {
                            // Shut down, timeline is suspended.
                            return Err(QueryError::from(io::Error::new(
                                io::ErrorKind::ConnectionAborted,
                                format!("end streaming to {:?}", spg.appname),
                            )));
                        }

                        // timeout expired: request pageserver status
                        pgb.write_message(&BeMessage::KeepAlive(WalSndKeepAlive {
                            sent_ptr: end_pos.0,
                            timestamp: get_current_timestamp(),
                            request_reply: true,
                        }))?;
                        continue;
                    }
                }

                let send_size = end_pos.checked_sub(start_pos).unwrap().0 as usize;
                let send_size = min(send_size, send_buf.len());

                let send_buf = &mut send_buf[..send_size];

                // read wal into buffer
                let send_size = wal_reader.read(send_buf).await?;
                let send_buf = &send_buf[..send_size];

                // Write some data to the network socket.
                pgb.write_message(&BeMessage::XLogData(XLogDataBody {
                    wal_start: start_pos.0,
                    wal_end: end_pos.0,
                    timestamp: get_current_timestamp(),
                    data: send_buf,
                }))
                .context("Failed to send XLogData")?;

                start_pos += send_size as u64;
                trace!("sent WAL up to {}", start_pos);
            }

            Ok(())
        })
    }
}

const POLL_STATE_TIMEOUT: Duration = Duration::from_secs(1);

// Wait until we have commit_lsn > lsn or timeout expires. Returns latest commit_lsn.
async fn wait_for_lsn(rx: &mut Receiver<Lsn>, lsn: Lsn) -> anyhow::Result<Option<Lsn>> {
    let commit_lsn: Lsn = *rx.borrow();
    if commit_lsn > lsn {
        return Ok(Some(commit_lsn));
    }

    let res = timeout(POLL_STATE_TIMEOUT, async move {
        let mut commit_lsn;
        loop {
            rx.changed().await?;
            commit_lsn = *rx.borrow();
            if commit_lsn > lsn {
                break;
            }
        }

        Ok(commit_lsn)
    })
    .await;

    match res {
        // success
        Ok(Ok(commit_lsn)) => Ok(Some(commit_lsn)),
        // error inside closure
        Ok(Err(err)) => Err(err),
        // timeout
        Err(_) => Ok(None),
    }
}