Merge pull request #96 from cita-cloud/fix_config_proc

fix proc reconfigure msg in setup

src/client.rs

@@ -19,7 +19,6 @@ use slog::Logger;
 use std::collections::HashMap;
 use std::ops::Deref;
 use std::sync::Arc;
-use tokio::sync::mpsc;
 use tokio::sync::RwLock;
 use tokio::time;
 
@@ -102,7 +101,7 @@ impl Network {
         local_id: u64,
         local_port: u16,
         network_port: u16,
-        msg_tx: mpsc::Sender<RaftMsg>,
+        msg_tx: flume::Sender<RaftMsg>,
         logger: Logger,
     ) -> Self {
         let mut inner = Inner::new(local_id, network_port, msg_tx, logger);
@@ -117,7 +116,7 @@ pub struct Inner {
     // local peer id
     local_id: u64,
     // Send msg to our local peer
-    msg_tx: mpsc::Sender<RaftMsg>,
+    msg_tx: flume::Sender<RaftMsg>,
 
     // raft_id -> origin
     // Cached origin to avoid broadcasting. These records maybe outdated,
@@ -137,7 +136,7 @@ impl Inner {
     fn new(
         local_id: u64,
         network_port: u16,
-        msg_tx: mpsc::Sender<RaftMsg>,
+        msg_tx: flume::Sender<RaftMsg>,
         logger: Logger,
     ) -> Self {
         let client_options = ClientOptions::new(
@@ -275,7 +274,7 @@ impl NetworkMsgHandlerService for Network {
                     let _ = self.client.clone().send_msg(rest_msg).await;
                 } else {
                     self.msg_tx
-                        .send(raft_msg)
+                        .send_async(raft_msg)
                         .await
                         .map_err(|_| tonic::Status::internal("consensus service is closed"))?;
                 }

src/grpc_server.rs

@@ -0,0 +1,133 @@
+// Copyright Rivtower Technologies LLC.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use crate::client::Network;
+use crate::config::ConsensusServiceConfig;
+use crate::health_check::HealthCheckServer;
+use cita_cloud_proto::common::ConsensusConfiguration;
+use cita_cloud_proto::common::ProposalWithProof;
+use cita_cloud_proto::common::StatusCode;
+use cita_cloud_proto::consensus::consensus_service_server::ConsensusService;
+use cita_cloud_proto::consensus::consensus_service_server::ConsensusServiceServer;
+use cita_cloud_proto::health_check::health_server::HealthServer;
+use cita_cloud_proto::network::network_msg_handler_service_server::NetworkMsgHandlerServiceServer;
+use cloud_util::metrics::{run_metrics_exporter, MiddlewareLayer};
+use slog::info;
+use slog::Logger;
+use tokio::sync::mpsc;
+use tonic::transport::Server;
+
+#[derive(Debug)]
+pub struct RaftConsensusService(mpsc::Sender<ConsensusConfiguration>);
+
+#[tonic::async_trait]
+impl ConsensusService for RaftConsensusService {
+    async fn reconfigure(
+        &self,
+        request: tonic::Request<ConsensusConfiguration>,
+    ) -> std::result::Result<tonic::Response<StatusCode>, tonic::Status> {
+        let config = request.into_inner();
+        let tx = self.0.clone();
+        // FIXME: it's not safe; but if we wait for it, it may cause a deadlock
+        tokio::spawn(async move {
+            let _ = tx.send(config).await;
+        });
+        // horrible
+        Ok(tonic::Response::new(StatusCode { code: 0 }))
+    }
+
+    async fn check_block(
+        &self,
+        _request: tonic::Request<ProposalWithProof>,
+    ) -> Result<tonic::Response<StatusCode>, tonic::Status> {
+        // Reply ok since we assume no byzantine faults.
+        // horrible
+        Ok(tonic::Response::new(StatusCode { code: 0 }))
+    }
+}
+
+pub async fn start_grpc_server(
+    config: ConsensusServiceConfig,
+    logger: Logger,
+    rx_signal: flume::Receiver<()>,
+    controller_tx: mpsc::Sender<ConsensusConfiguration>,
+    network: Network,
+) {
+    let raft_svc = RaftConsensusService(controller_tx);
+
+    let logger_cloned = logger.clone();
+    let grpc_listen_port = config.grpc_listen_port;
+    info!(
+        logger_cloned,
+        "grpc port of consensus_raft: {}", grpc_listen_port
+    );
+
+    let layer = if config.enable_metrics {
+        tokio::spawn(async move {
+            run_metrics_exporter(config.metrics_port).await.unwrap();
+        });
+
+        Some(
+            tower::ServiceBuilder::new()
+                .layer(MiddlewareLayer::new(config.metrics_buckets))
+                .into_inner(),
+        )
+    } else {
+        None
+    };
+
+    info!(logger_cloned, "start consensus_raft grpc server");
+    if let Some(layer) = layer {
+        tokio::spawn(async move {
+            info!(logger_cloned, "metrics on");
+            let addr = format!("127.0.0.1:{grpc_listen_port}").parse().unwrap();
+            let res = Server::builder()
+                .layer(layer)
+                .add_service(ConsensusServiceServer::new(raft_svc))
+                .add_service(NetworkMsgHandlerServiceServer::new(network))
+                .add_service(HealthServer::new(HealthCheckServer {}))
+                .serve_with_shutdown(
+                    addr,
+                    cloud_util::graceful_shutdown::grpc_serve_listen_term(rx_signal),
+                )
+                .await;
+
+            if let Err(e) = res {
+                info!(logger_cloned, "grpc service exit with error: `{:?}`", e);
+            } else {
+                info!(logger_cloned, "grpc service exit");
+            }
+        });
+    } else {
+        tokio::spawn(async move {
+            info!(logger_cloned, "metrics off");
+            let addr = format!("127.0.0.1:{grpc_listen_port}").parse().unwrap();
+            let res = Server::builder()
+                .add_service(ConsensusServiceServer::new(raft_svc))
+                .add_service(NetworkMsgHandlerServiceServer::new(network))
+                .add_service(HealthServer::new(HealthCheckServer {}))
+                .serve_with_shutdown(
+                    addr,
+                    cloud_util::graceful_shutdown::grpc_serve_listen_term(rx_signal),
+                )
+                .await;
+
+            if let Err(e) = res {
+                info!(logger_cloned, "grpc service exit with error: `{:?}`", e);
+            } else {
+                info!(logger_cloned, "grpc service exit");
+            }
+        });
+    }
+}

src/main.rs

@@ -14,21 +14,29 @@
 
 mod client;
 mod config;
+mod grpc_server;
 mod health_check;
 mod peer;
 mod storage;
 mod utils;
 
+use crate::client::{Controller, Network};
+use crate::grpc_server::start_grpc_server;
+use cita_cloud_proto::common::ProposalWithProof;
+use cita_cloud_proto::common::{ConsensusConfiguration, Proposal};
 use clap::{crate_authors, crate_version, value_parser, Arg, ArgAction, Command};
 use config::ConsensusServiceConfig;
 use peer::Peer;
-use slog::info;
+use slog::{debug, info, o, warn};
 use sloggers::file::FileLoggerBuilder;
 use sloggers::terminal::TerminalLoggerBuilder;
 use sloggers::Build as _;
 use std::path::Path;
 use std::path::PathBuf;
-use utils::clap_about;
+use tokio::sync::mpsc;
+use tokio::task::JoinHandle;
+use tokio::time::Duration;
+use utils::{addr_to_peer_id, clap_about};
 
 fn main() {
     let run_cmd = Command::new("run")
@@ -112,10 +120,141 @@ fn main() {
             let rt = tokio::runtime::Runtime::new().unwrap();
 
             rt.block_on(async move {
+                let node_addr = {
+                    let s = &config.node_addr;
+                    hex::decode(s.strip_prefix("0x").unwrap_or(s)).expect("decode node_addr failed")
+                };
+
+                let local_id = addr_to_peer_id(&node_addr);
+
+                //start grpc server
                 let rx_signal = cloud_util::graceful_shutdown::graceful_shutdown();
-                let mut peer = Peer::setup(config, logger.clone(), rx_signal.clone()).await;
-                peer.run(rx_signal).await;
-                info!(logger, "raft service exit");
+                // raft Communicate with controller
+                let (controller_tx, mut controller_rx) =
+                    mpsc::channel::<ConsensusConfiguration>(1000);
+
+                // raft Communicate with network
+                let (peer_tx, peer_rx) = flume::bounded(64);
+
+                // Controller grpc client
+                let controller = {
+                    let logger = logger.new(o!("tag" => "controller"));
+                    Controller::new(config.controller_port, logger)
+                };
+
+                // Network grpc client
+                let network = {
+                    let logger = logger.new(o!("tag" => "network"));
+                    Network::setup(
+                        local_id,
+                        config.grpc_listen_port,
+                        config.network_port,
+                        peer_tx,
+                        logger,
+                    )
+                    .await
+                };
+
+                // stop raft when I'm not in the validators list
+                let (stop_tx, stop_rx) = flume::bounded(1);
+
+                // start consensu grpc server
+                start_grpc_server(
+                    config.clone(),
+                    logger.clone(),
+                    rx_signal.clone(),
+                    controller_tx,
+                    network.clone(),
+                )
+                .await;
+
+                // wait for exit signal
+                tokio::spawn(async move {
+                    let _ = rx_signal.recv_async().await;
+                    // stop task before exit
+                    drop(stop_tx);
+                });
+
+                // raft task handle
+                let mut opt_raft_task: Option<JoinHandle<()>> = None;
+
+                loop {
+                    tokio::time::sleep(Duration::from_secs(3)).await;
+
+                    // check exit signal
+                    let ret = stop_rx.try_recv();
+                    if ret == Err(flume::TryRecvError::Disconnected) {
+                        break;
+                    }
+
+                    // check reconfigure message from controller
+                    let mut opt_trigger_config = None;
+                    // read all reconfigure messages in channel
+                    // when start a new node, sync blocks quickly, there will be many reconfigure messages in channel
+                    // we only need proc latest one
+                    while let Ok(config) = controller_rx.try_recv() {
+                        opt_trigger_config = Some(config);
+                    }
+                    if opt_trigger_config.is_none() {
+                        let pwp = ProposalWithProof {
+                            proposal: Some(Proposal {
+                                height: u64::MAX,
+                                data: vec![],
+                            }),
+                            proof: vec![],
+                        };
+                        debug!(logger, "ping_controller..");
+                        match controller.commit_block(pwp).await {
+                            Ok(config) => {
+                                opt_trigger_config = Some(config);
+                            }
+                            Err(e) => {
+                                warn!(logger, "commit block failed: {}", e);
+                            }
+                        }
+                    }
+                    if let Some(trigger_config) = opt_trigger_config {
+                        info!(
+                            logger,
+                            "get reconfigure from controller, height is {}", trigger_config.height
+                        );
+                        if trigger_config.validators.contains(&node_addr) {
+                            if opt_raft_task.is_none()
+                                || opt_raft_task.as_ref().unwrap().is_finished()
+                            {
+                                let controller_cloned = controller.clone();
+                                let network_cloned = network.clone();
+                                let logger_cloned = logger.clone();
+                                let config_cloned = config.clone();
+                                let stop_rx_cloned = stop_rx.clone();
+                                let peer_rx_cloned = peer_rx.clone();
+                                let handle = tokio::spawn(async move {
+                                    info!(logger_cloned, "raft start");
+                                    let mut peer = Peer::setup(
+                                        config_cloned,
+                                        logger_cloned.clone(),
+                                        trigger_config,
+                                        network_cloned,
+                                        controller_cloned,
+                                        peer_rx_cloned,
+                                    )
+                                    .await;
+                                    peer.run(stop_rx_cloned).await;
+                                    info!(logger_cloned, "raft exit");
+                                });
+                                opt_raft_task = Some(handle);
+                            }
+                        } else {
+                            info!(logger, "I'm not in the validators list");
+                            if let Some(ref handle) = opt_raft_task {
+                                if !handle.is_finished() {
+                                    info!(logger, "abort raft");
+                                    handle.abort();
+                                }
+                            }
+                        }
+                    }
+                }
             });
         }
         None => {