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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599
//! The Raft actor's module and its associated logic. mod admin; mod append_entries; mod apply_logs; mod client; mod install_snapshot; mod replication; mod state; mod vote; use std::{ collections::BTreeMap, sync::Arc, time::{Duration, Instant}, }; use actix::prelude::*; use futures::sync::{mpsc}; use log::{error}; use crate::{ AppData, AppDataResponse, AppError, NodeId, common::{ApplyLogsTask, DependencyAddr, UpdateCurrentLeader}, config::Config, messages::{ClientPayload, MembershipConfig}, metrics::{RaftMetrics, State}, network::RaftNetwork, raft::state::{CandidateState, FollowerState, LeaderState, RaftState, ReplicationState, SnapshotState}, replication::{ReplicationStream, RSTerminate}, storage::{GetInitialState, GetLogEntries, HardState, InitialState, RaftStorage, SaveHardState}, }; const FATAL_ACTIX_MAILBOX_ERR: &str = "Fatal actix MailboxError while communicating with Raft dependency. Raft is shutting down."; const FATAL_STORAGE_ERR: &str = "Fatal storage error encountered which can not be recovered from. Stopping Raft node."; ////////////////////////////////////////////////////////////////////////////////////////////////// // Raft ////////////////////////////////////////////////////////////////////////////////////////// /// An actor which implements the Raft protocol's core business logic. /// /// For more information on the Raft protocol, see the specification here: /// https://raft.github.io/raft.pdf (**pdf warning**). /// /// The beginning of §5, the spec has a condensed summary of the Raft consensus algorithm. This /// crate, and especially this actor, attempts to follow the terminology and nomenclature used /// there as precisely as possible to aid in understanding this system. /// /// ### api /// This actor's API is broken up into 3 different layers, all based on message handling. In order /// to effectively use this actor, only these 3 layers need to be considered. /// /// #### network /// The network interface of the parent application is responsible for providing a conduit to /// exchange the various messages types defined in this system. The `RaftNetwork` trait defines /// the interface needed for being able to allow Raft cluster members to be able to communicate /// with each other. In addition to the `RaftNetwork` trait, applications are expected to provide /// and interface for their clients to be able to submit data which needs to be managed by Raft. /// /// ##### raft rpc messages /// These are Raft request PRCs coming from other nodes of the cluster. They are defined in the /// `messages` module of this crate. They are `AppendEntriesRequest`, `VoteRequest` & /// `InstallSnapshotRequest`. This actor will use the `RaftNetwork` impl of the parent application /// to send RPCs to other nodes. /// /// The application's networking layer must decode these message types and pass them over to the /// appropriate handler on this type, await the response, and then send the response back over the /// wire to the caller. /// /// ##### client request messages /// These are messages coming from an application's clients, represented by the /// `messages::ClientPayload` type. When the message type's handler is called a future will be /// returned which will resolve with the appropriate response type. Only data mutating messages /// should ever need to go through Raft. The contentsof these messages are entirely specific to /// your application. /// /// #### storage /// The storage interface is typically going to be the most involved as this is where your /// application really exists. SQL, NoSQL, mutable, immutable, KV, append only ... whatever your /// application's data model, this is where it comes to life. /// /// The storage interface is defined by the `RaftStorage` trait. Depending on the data storage /// system being used, the actor my be sync or async. It just needs to implement handlers for /// the needed actix message types. /// /// #### admin /// These are admin commands which may be issued to a Raft node in order to influence it in ways /// outside of the normal Raft lifecycle. Dynamic membership changes and cluster initialization /// are the main commands of this layer. pub struct Raft<D: AppData, R: AppDataResponse, E: AppError, N: RaftNetwork<D>, S: RaftStorage<D, R, E>> { /// This node's ID. id: NodeId, /// This node's runtime config. config: Arc<Config>, /// The cluster's current membership configuration. membership: MembershipConfig, /// The current state of this Raft node. state: RaftState<D, R, E, N, S>, /// The address of the actor responsible for implementing the `RaftNetwork` interface. network: Addr<N>, /// The address of the actor responsible for implementing the `RaftStorage` interface. storage: Addr<S::Actor>, /// The address of the actor responsible for recieving metrics output from this Node. metrics: Recipient<RaftMetrics>, /// The index of the highest log entry known to be committed cluster-wide. /// /// The definition of a committed log is that the leader which has created the log has /// successfully replicated the log to a majority of the cluster. This value is updated via /// AppendEntries RPC from the leader, or if a node is the leader, it will update this value /// as new entries have been successfully replicated to a majority of the cluster. /// /// Is initialized to 0, and increases monotonically. This is always based on the leader's /// commit index which is communicated to other members via the AppendEntries protocol. commit_index: u64, /// The index of the highest log entry which has been applied to the local state machine. /// /// Is initialized to 0, increases monotonically following the `commit_index` as logs are /// applied to the state machine (via the storage interface). last_applied: u64, /// The current term. /// /// Is initialized to 0 on first boot, and increases monotonically. This is normally based on /// the leader's term which is communicated to other members via the AppendEntries protocol, /// but this may also be incremented when a follower becomes a candidate. current_term: u64, /// The ID of the current leader of the Raft cluster. /// /// This value is kept up-to-date based on a very simple algorithm, which is the only way to /// do so reasonably using only the canonical Raft RPCs described in the spec. When a new /// leader comes to power, it will send AppendEntries RPCs to establish its leadership. When /// such an RPC is observed with a newer term, this value will be updated. This value will be /// set to `None` when a newer term is observed in any other way. current_leader: Option<NodeId>, /// The ID of the candidate which received this node's vote for the current term. /// /// Each server will vote for at most one candidate in a given term, on a /// first-come-first-served basis. See §5.4.1 for additional restriction on votes. voted_for: Option<NodeId>, /// The index of the last log to be appended. last_log_index: u64, /// The term of the last log to be appended. last_log_term: u64, /// A flag to indicate if this system is currently appending logs. is_appending_logs: bool, /// The entrypoint to the pipeline of logs which need to be applied to the state machine. apply_logs_pipeline: mpsc::UnboundedSender<ApplyLogsTask<D, R, E>>, /// The receiving end of the pipeline for applying logs. This is moved out and spawned when Raft starts. _apply_logs_pipeline_receiver: Option<mpsc::UnboundedReceiver<ApplyLogsTask<D, R, E>>>, /// A handle to the election timeout callback. election_timeout: Option<actix::SpawnHandle>, /// The currently scheduled election timeout. election_timeout_stamp: Option<Instant>, } impl<D: AppData, R: AppDataResponse, E: AppError, N: RaftNetwork<D>, S: RaftStorage<D, R, E>> Raft<D, R, E, N, S> { /// Create a new Raft instance. /// /// This actor will need to be started after instantiation, which must be done within a /// running actix system. pub fn new(id: NodeId, config: Config, network: Addr<N>, storage: Addr<S::Actor>, metrics: Recipient<RaftMetrics>) -> Self { let state = RaftState::Initializing; let config = Arc::new(config); let (tx, rx) = mpsc::unbounded(); let membership = MembershipConfig{is_in_joint_consensus: false, members: vec![id], non_voters: vec![], removing: vec![]}; Self{ id, config, membership, state, network, storage, metrics, commit_index: 0, last_applied: 0, current_term: 0, current_leader: None, voted_for: None, last_log_index: 0, last_log_term: 0, is_appending_logs: false, apply_logs_pipeline: tx, _apply_logs_pipeline_receiver: Some(rx), election_timeout: None, election_timeout_stamp: None, } } /// Transition to the Raft non-voter state. fn become_non_voter(&mut self, ctx: &mut Context<Self>) { // Cleanup previous state. self.cleanup_state(ctx); // Ensure there is no election timeout. self.cancel_election_timeout(ctx); // Perform the transition. self.state = RaftState::NonVoter; self.report_metrics(ctx); } /// Transition to the Raft follower state. fn become_follower(&mut self, ctx: &mut Context<Self>) { // Don't transition to follower state if the cluster has this node configured as a non-voter. if !self.membership.contains(&self.id) || self.membership.non_voters.contains(&self.id) { return; } // Cleanup previous state. self.cleanup_state(ctx); // Ensure we have an election timeout loop running. self.update_election_timeout(ctx); // Cleans-up any old timeout task, and spawns a new one. // Perform the transition. self.state = RaftState::Follower(FollowerState::default()); self.report_metrics(ctx); } /// Transition to the Raft candidate state and start a new election campaign, per §5.2. /// /// As part of an election campaign, a follower increments its current term and transitions to /// candidate state, it then votes for itself (will then save its hard state) and issues /// RequestVote RPCs in parallel to each of the other nodes in the cluster. /// /// A candidate remains in the candidate state until one of three things happens: /// /// 1. It wins the election. /// 2. Another server establishes itself as leader. /// 3. A period of time goes by with no winner. /// /// (1) a candidate wins an election if it receives votes from a majority of the servers /// in the full cluster for the same term. Each server will vote for at most one candidate in /// a given term, on a first-come-first-served basis (§5.4 adds an additional restriction on /// votes). The majority rule ensures that at most one candidate can win the election for a /// particular term. Once a candidate wins an election, it becomes leader. It then sends /// heartbeat messages to all of the other servers to establish its authority and prevent new /// elections. /// /// (2) While waiting for votes, a candidate may receive an AppendEntries RPC from another /// server claiming to be leader. If the leader’s term in the RPC is at least as large as the /// candidate’s current term, then the candidate recognizes the leader as legitimate and /// returns to follower state. If the term in the RPC is smaller than the candidate’s current /// term, then the candidate rejects the RPC and continues in candidate state. /// /// (3) The third possible outcome is that a candidate neither wins nor loses the election: if /// many followers become candidates at the same time, votes could be split so that no /// candidate obtains a majority. When this happens, each candidate will time out and start a /// new election by incrementing its term and initiating another round of RequestVote RPCs. /// The randomization of election timeouts per node helps to avoid this issue. fn become_candidate(&mut self, ctx: &mut Context<Self>) { // Cleanup previous state. self.cleanup_state(ctx); // Setup new term. self.current_term += 1; self.voted_for = Some(self.id); self.update_current_leader(ctx, UpdateCurrentLeader::Unknown); self.save_hard_state(ctx); // Send RPCs to all members in parallel. let mut requests = BTreeMap::new(); let peers = self.membership.members.iter().filter(|member| *member != &self.id).map(|e| *e).collect::<Vec<_>>(); for member in peers { let f = self.request_vote(ctx, member); let handle = ctx.spawn(f); requests.insert(member, handle); } // Update the election timeout. self.update_election_timeout(ctx); // Update Raft state as candidate. let votes_granted = 1; // We must vote for ourselves per the Raft spec. let votes_needed = ((self.membership.members.len() / 2) + 1) as u64; // Just need a majority. self.state = RaftState::Candidate(CandidateState{requests, votes_granted, votes_needed}); self.report_metrics(ctx); } /// Transition to the Raft leader state. /// /// Once a node becomes the Raft cluster leader, its behavior will be a bit different. Upon /// election: /// /// - Each cluster member gets a `ReplicationStream` actor spawned. Addr is retained. /// - Initial AppendEntries RPCs (heartbeats) are sent to each cluster member, and is repeated /// during idle periods to prevent election timeouts, per §5.2. This is handled by the /// `ReplicationStream` actors. /// - A new blank log entry is generated and committed to the cluster in order to ensure that /// there are no unapplied entries from the last term, per the end of §8. This blank entry is /// used to ensure that a joint consensus config, if present, has been properly committed to /// the cluster. /// /// See the `ClientRpcIn` handler for more details on the write path for client requests. fn become_leader(&mut self, ctx: &mut Context<Self>) { // Cleanup previous state & ensure we've cancelled the election timeout system. self.cleanup_state(ctx); self.cancel_election_timeout(ctx); // Prep new leader state. let (client_request_queue, client_request_receiver) = mpsc::unbounded(); let mut new_state = LeaderState::new(client_request_queue, &self.membership); // Spawn stream which consumes client RPCs. ctx.spawn(fut::wrap_stream(client_request_receiver) .and_then(|msg, act: &mut Self, ctx| act.process_client_rpc(ctx, msg)) .then(|_, _, _| fut::ok(())) // Ensure errors don't cause the stream to close. .finish()); // Spawn new replication stream actors. let targets = self.membership.members.iter().filter(|elem| *elem != &self.id) .chain(self.membership.non_voters.iter()); for target in targets { // Build the replication stream for the target member. let rs = ReplicationStream::new( self.id, *target, self.current_term, self.config.clone(), self.last_log_index, self.last_log_term, self.commit_index, ctx.address(), self.network.clone(), self.storage.clone().recipient::<GetLogEntries<D, E>>(), ); let addr = rs.start(); // Start the actor on the same thread. // Retain the addr of the replication stream. let state = ReplicationState{match_index: self.last_log_index, is_at_line_rate: true, addr, remove_after_commit: None}; new_state.nodes.insert(*target, state); } // Initialize new state as leader. self.state = RaftState::Leader(new_state); self.update_current_leader(ctx, UpdateCurrentLeader::ThisNode); self.report_metrics(ctx); // Commit a new blank entry to the cluster to guard against stale-reads, per §8. // If the cluster has just formed, and the current index is 0, then commit the current config. let payload = if self.last_log_index == 0 { ClientPayload::new_config(self.membership.clone()) } else { ClientPayload::new_blank_payload() }; ctx.spawn(fut::wrap_future(ctx.address().send(payload)) .map_err(|_, _, _| ()) .and_then(|res, _, _| fut::result(res.map_err(|_| ()))) // In the case that there was a stale record and it was a joint consensus // finalization, ensure it is handled properly. .and_then(|res, act: &mut Self, ctx| act.handle_joint_consensus_finalization(ctx, res)) ); } /// Clean up the current Raft state. /// /// This will typically be called before a state transition takes place. fn cleanup_state(&mut self, ctx: &mut Context<Self>) { match &mut self.state { RaftState::Follower(inner) => { inner.snapshot_state = SnapshotState::Idle; } RaftState::Candidate(inner) => { for handle in inner.cleanup() { ctx.cancel_future(handle); } } RaftState::Leader(inner) => { inner.nodes.values().for_each(|rsstate| { let _ = rsstate.addr.do_send(RSTerminate); }); } _ => (), } } /// Perform the initialization routine for the Raft node. /// /// If this node has configuration present from being online previously, then this node will /// begin a standard lifecycle as a follower. If this node is pristine, then it will wait in /// standby mode. /// /// ### previous state | follower /// If the node has previous state, then there are a few cases to account for. /// /// If the node has been offline for some time and was removed from the cluster, no problem. /// Any RPCs sent from this node will be rejected until it is added to the cluster. Once it is /// added to the cluster again, the standard Raft protocol will resume as normal. /// /// If the node went down only very briefly, then it should immediately start receiving /// heartbeats and resume as normal, else it will start an election if it doesn't receive any /// heartbeats from the leader per normal Raft protocol. /// /// If the node was running standalone, it will win the election and resume as a standalone. /// /// ### pristine state | standby /// While in standby mode, the Raft leader of the current cluster may discover this node and /// add it to the cluster. In such a case, it will begin receiving heartbeats from the leader /// and business proceeds as usual. /// /// If there is no current cluster, while in standby mode, the node may receive an admin /// command instructing it to campaign with a specific config, or to begin operating as the /// leader of a standalone cluster. fn initialize(&mut self, ctx: &mut Context<Self>, state: InitialState) { self.last_log_index = state.last_log_index; self.last_log_term = state.last_log_term; self.current_term = state.hard_state.current_term; self.voted_for = state.hard_state.voted_for; self.membership = state.hard_state.membership; self.last_applied = state.last_applied_log; // NOTE: this is repeated here for clarity. It is unsafe to initialize the node's commit // index to any other value. The commit index must be determined by a leader after // successfully committing a new log to the cluster. self.commit_index = 0; // Spawn the stream for applying logs to the state machine. This will always be `Some` here, never after. if let Some(rx) = self._apply_logs_pipeline_receiver.take() { ctx.spawn(fut::wrap_stream(rx) .and_then(|msg, act: &mut Self, ctx| act.process_apply_logs_task(ctx, msg)) .finish()); } // Start the metrics reporter. ctx.run_interval(self.config.metrics_rate.clone(), |act, ctx| act.report_metrics(ctx)); // Set initial state based on state recovered from disk. let is_only_configured_member = self.membership.len() == 1 && self.membership.contains(&self.id); // If this is the only configured member and there is live state, then this is // a single-node cluster currently. Become leader. if is_only_configured_member && &self.last_log_index != &u64::min_value() { self.become_leader(ctx); } // Else if there are other members, that can only mean that state was recovered. Become follower. else if !is_only_configured_member { self.state = RaftState::Follower(FollowerState::default()); self.update_election_timeout(ctx); } // Else, for any other condition, stay non-voter. else { self.state = RaftState::NonVoter; } } /// Transform and log an actix MailboxError. /// /// This method treats the error as being fatal, as Raft can not function properly if the /// `RaftNetowrk` & `RaftStorage` interfaces are returning mailbox errors. This method will /// shutdown the Raft actor. fn map_fatal_actix_messaging_error(&mut self, ctx: &mut Context<Self>, err: actix::MailboxError, dep: DependencyAddr) { error!("{} {:?} {:?}", FATAL_ACTIX_MAILBOX_ERR, dep, err); ctx.terminate(); } /// Transform an log the result of a `RaftStorage` interaction. /// /// This method assumes that a storage error observed here is non-recoverable. As such, the /// Raft node will be instructed to stop. If such behavior is not needed, then don't use this /// interface. fn map_fatal_storage_result<T>(&mut self, ctx: &mut Context<Self>, res: Result<T, E>) -> impl ActorFuture<Actor=Self, Item=T, Error=()> { let res = res.map_err(|err| { error!("{} {:?}", FATAL_STORAGE_ERR, err); ctx.terminate(); }); fut::result(res) } /// Report a metrics payload on the current state of the Raft node. fn report_metrics(&mut self, _: &mut Context<Self>) { let state = match &self.state { RaftState::NonVoter => State::NonVoter, RaftState::Follower(_) => State::Follower, RaftState::Candidate(_) => State::Candidate, RaftState::Leader(_) => State::Leader, _ => return, }; let _ = self.metrics.do_send(RaftMetrics{ id: self.id, state, current_term: self.current_term, last_log_index: self.last_log_index, last_applied: self.last_applied, current_leader: self.current_leader, membership_config: self.membership.clone(), }).map_err(|err| { error!("Error reporting metrics. {}", err); }); } /// Save the Raft node's current hard state to disk. /// /// DEPRECATED: use `save_hard_state_async`. fn save_hard_state(&mut self, ctx: &mut Context<Self>) { let hs = HardState{current_term: self.current_term, voted_for: self.voted_for, membership: self.membership.clone()}; let f = fut::wrap_future(self.storage.send::<SaveHardState<E>>(SaveHardState::new(hs))) .map_err(|err, act: &mut Self, ctx| act.map_fatal_actix_messaging_error(ctx, err, DependencyAddr::RaftStorage)) .and_then(|res, act, ctx| act.map_fatal_storage_result(ctx, res)); ctx.spawn(f); } /// Save the Raft node's current hard state to disk. fn save_hard_state_async(&mut self, _: &mut Context<Self>) -> impl ActorFuture<Actor=Self, Item=(), Error=()> { let hs = HardState{current_term: self.current_term, voted_for: self.voted_for, membership: self.membership.clone()}; fut::wrap_future(self.storage.send::<SaveHardState<E>>(SaveHardState::new(hs))) .map_err(|err, act: &mut Self, ctx| act.map_fatal_actix_messaging_error(ctx, err, DependencyAddr::RaftStorage)) .and_then(|res, act, ctx| act.map_fatal_storage_result(ctx, res)) } /// Update the value of the `current_leader` property. /// /// NOTE WELL: there was previously a bit of log encapsulated here related to forwarding /// requests to leaders and such. In order to more closely mirror the Raft spec and allow apps /// to determine how they want to handle forwarding client requests to leaders, that logic was /// removed and this handler has thus been greatly simplified. We are keeping it as is in case /// we need to add some additional logic here. fn update_current_leader(&mut self, _: &mut Context<Self>, update: UpdateCurrentLeader) { match update { UpdateCurrentLeader::ThisNode => { self.current_leader = Some(self.id); } UpdateCurrentLeader::OtherNode(target) => { self.current_leader = Some(target); } UpdateCurrentLeader::Unknown => { self.current_leader = None; }, } } /// Encapsulate the process of updating the current term, as updating the `voted_for` state must also be updated. fn update_current_term(&mut self, new_term: u64, voted_for: Option<NodeId>) { if new_term > self.current_term { self.current_term = new_term; self.voted_for = voted_for; } } /// Update the election timeout process. /// /// This will schedule a new interval job based on the configured election timeout. The /// interval job will check to see if a campaign should be started based on when the last /// heartbeat was received from the Raft leader or a candidate. /// /// The election timeout stamp will be updated everytime this node receives an RPC from the /// leader as well as any time a candidate node sends a RequestVote RPC if it is a /// valid vote request. fn update_election_timeout(&mut self, ctx: &mut Context<Self>) { // Don't update if the cluster has this node configured as a non-voter. if !self.membership.contains(&self.id) || self.membership.non_voters.contains(&self.id) { return; } // Cancel any current election timeout before spawning a new one. self.cancel_election_timeout(ctx); let timeout = Duration::from_millis(self.config.election_timeout_millis); self.election_timeout_stamp = Some(Instant::now() + timeout.clone()); self.election_timeout = Some(ctx.run_interval(timeout, |act, ctx| { if let Some(stamp) = &act.election_timeout_stamp { if &Instant::now() >= stamp { act.become_candidate(ctx) } } })); } /// Update the election timeout stamp, typically due to receiving a heartbeat from the Raft leader. fn update_election_timeout_stamp(&mut self) { self.election_timeout_stamp = Some(Instant::now() + Duration::from_millis(self.config.election_timeout_millis)); } /// Cancel the current election timeout task if present & clean-up the election timeout stamp. fn cancel_election_timeout(&mut self, ctx: &mut Context<Self>) { self.election_timeout_stamp = None; if let Some(handle) = self.election_timeout.take() { ctx.cancel_future(handle); } } /// Update the node's current membership config. /// /// NOTE WELL: if a leader is stepping down, it should not call this method, as it will cause /// the node to transition out of leader state before it can commit the config entry. fn update_membership(&mut self, ctx: &mut Context<Self>, cfg: MembershipConfig) -> impl ActorFuture<Actor=Self, Item=(), Error=()> { self.membership = cfg; // If the given config does not contain this node's ID, it means one of the following: // - the node is currently a non-voter and is replicating an old config to which it has // not yet been added. // - the node has been removed from the cluster. The parent application can observe the // transition to the non-voter state as a signal for when it is safe to shutdown a node // being removed. if !self.membership.contains(&self.id) { self.become_non_voter(ctx); } else if self.state.is_non_voter() && self.membership.members.contains(&self.id) { // The node is a NonVoter and the new config has it configured as a normal member. // Transition to follower. self.become_follower(ctx); } self.save_hard_state_async(ctx) } } impl<D: AppData, R: AppDataResponse, E: AppError, N: RaftNetwork<D>, S: RaftStorage<D, R, E>> Actor for Raft<D, R, E, N, S> { type Context = Context<Self>; /// The initialization routine for this actor. fn started(&mut self, ctx: &mut Self::Context) { // Fetch the node's initial state from the storage actor & initialize. let f = fut::wrap_future(self.storage.send::<GetInitialState<E>>(GetInitialState::new())) .map_err(|err, act: &mut Self, ctx| act.map_fatal_actix_messaging_error(ctx, err, DependencyAddr::RaftStorage)) .and_then(|res, act, ctx| act.map_fatal_storage_result(ctx, res)) .map(|state, act, ctx| act.initialize(ctx, state)); ctx.spawn(f); } }