/** * Copyright (c) 2013, Willem-Hendrik Thiart * Use of this source code is governed by a BSD-style license that can be * found in the LICENSE file. * * @file * @brief Implementation of a Raft server * @author Willem Thiart himself@willemthiart.com * @version 0.1 */ #include #include #include #include /* for varags */ #include #include "raft.h" #include "raft_log.h" #include "raft_private.h" #define min(a, b) ((a) < (b) ? (a) : (b)) #define max(a, b) ((a) < (b) ? (b) : (a)) static void __log(raft_server_t *me_, const char *fmt, ...) { raft_server_private_t* me = (raft_server_private_t*)me_; char buf[1024]; va_list args; va_start(args, fmt); vsprintf(buf, fmt, args); if (me->cb.log) me->cb.log(me_, me->udata, buf); } raft_server_t* raft_new() { raft_server_private_t* me = (raft_server_private_t*)calloc(1, sizeof(raft_server_private_t)); if (!me) return NULL; me->current_term = 0; me->voted_for = -1; me->timeout_elapsed = 0; me->request_timeout = 200; me->election_timeout = 1000; me->log = log_new(); raft_set_state((raft_server_t*)me, RAFT_STATE_FOLLOWER); me->current_leader = -1; return (raft_server_t*)me; } void raft_set_callbacks(raft_server_t* me_, raft_cbs_t* funcs, void* udata) { raft_server_private_t* me = (raft_server_private_t*)me_; memcpy(&me->cb, funcs, sizeof(raft_cbs_t)); me->udata = udata; log_set_callbacks(me->log, &me->cb, me_); } void raft_free(raft_server_t* me_) { raft_server_private_t* me = (raft_server_private_t*)me_; log_free(me->log); free(me_); } void raft_election_start(raft_server_t* me_) { raft_server_private_t* me = (raft_server_private_t*)me_; __log(me_, "election starting: %d %d, term: %d", me->election_timeout, me->timeout_elapsed, me->current_term); raft_become_candidate(me_); } void raft_become_leader(raft_server_t* me_) { raft_server_private_t* me = (raft_server_private_t*)me_; int i; __log(me_, "becoming leader"); raft_set_state(me_, RAFT_STATE_LEADER); me->voted_for = -1; for (i = 0; i < me->num_nodes; i++) { if (me->nodeid != i) { raft_node_t* p = raft_get_node(me_, i); raft_node_set_next_idx(p, raft_get_current_idx(me_) + 1); raft_send_appendentries(me_, i); } } } void raft_become_candidate(raft_server_t* me_) { raft_server_private_t* me = (raft_server_private_t*)me_; int i; __log(me_, "becoming candidate"); memset(me->votes_for_me, 0, sizeof(int) * me->num_nodes); me->current_term += 1; raft_vote(me_, me->nodeid); me->current_leader = -1; raft_set_state(me_, RAFT_STATE_CANDIDATE); /* we need a random factor here to prevent simultaneous candidates */ me->timeout_elapsed = rand() % 500; for (i = 0; i < me->num_nodes; i++) if (me->nodeid != i) raft_send_requestvote(me_, i); } void raft_become_follower(raft_server_t* me_) { raft_server_private_t* me = (raft_server_private_t*)me_; __log(me_, "becoming follower"); raft_set_state(me_, RAFT_STATE_FOLLOWER); me->voted_for = -1; } int raft_periodic(raft_server_t* me_, int msec_since_last_period) { raft_server_private_t* me = (raft_server_private_t*)me_; switch (me->state) { case RAFT_STATE_FOLLOWER: if (me->last_applied_idx < me->commit_idx) if (-1 == raft_apply_entry(me_)) return -1; break; } me->timeout_elapsed += msec_since_last_period; if (me->state == RAFT_STATE_LEADER) { if (me->request_timeout <= me->timeout_elapsed) { raft_send_appendentries_all(me_); me->timeout_elapsed = 0; } } else if (me->election_timeout <= me->timeout_elapsed) raft_election_start(me_); return 0; } raft_entry_t* raft_get_entry_from_idx(raft_server_t* me_, int etyidx) { raft_server_private_t* me = (raft_server_private_t*)me_; return log_get_from_idx(me->log, etyidx); } int raft_recv_appendentries_response(raft_server_t* me_, int node, msg_appendentries_response_t* r) { raft_server_private_t* me = (raft_server_private_t*)me_; __log(me_, "received appendentries response node: %d %s cidx: %d 1stidx: %d", node, r->success == 1 ? "success" : "fail", r->current_idx, r->first_idx); // TODO: should force invalid leaders to stepdown raft_node_t* p = raft_get_node(me_, node); if (0 == r->success) { /* If AppendEntries fails because of log inconsistency: decrement nextIndex and retry (§5.3) */ assert(0 <= raft_node_get_next_idx(p)); // TODO can jump back to where node is different instead of iterating raft_node_set_next_idx(p, raft_node_get_next_idx(p) - 1); /* retry */ raft_send_appendentries(me_, node); return 0; } raft_node_set_next_idx(p, r->current_idx + 1); int i; for (i = r->first_idx; i <= r->current_idx; i++) log_mark_node_has_committed(me->log, i); while (1) { raft_entry_t* e = raft_get_entry_from_idx(me_, me->last_applied_idx + 1); /* majority has this */ if (e && me->num_nodes / 2 <= e->num_nodes) { if (-1 == raft_apply_entry(me_)) break; } else break; } return 0; } int raft_recv_appendentries( raft_server_t* me_, const int node, msg_appendentries_t* ae, msg_appendentries_response_t *r ) { raft_server_private_t* me = (raft_server_private_t*)me_; me->timeout_elapsed = 0; if (0 < ae->n_entries) __log(me_, "recvd appendentries from: %d, %d %d %d %d #%d", node, ae->term, ae->leader_commit, ae->prev_log_idx, ae->prev_log_term, ae->n_entries); r->term = me->current_term; /* Step down - we've found a leader who is legitimate */ if (raft_is_leader(me_) && me->current_term <= ae->term) raft_become_follower(me_); /* 1. Reply false if term < currentTerm (§5.1) */ if (ae->term < me->current_term) { __log(me_, "AE term is less than current term"); r->success = 0; return 0; } #if 0 if (-1 != ae->prev_log_idx && ae->prev_log_idx < raft_get_current_idx(me_)) { __log(me_, "AE prev_idx is less than current idx"); r->success = 0; return 0; } #endif /* Not the first appendentries we've received */ /* NOTE: the log starts at 1 */ if (0 < ae->prev_log_idx) { raft_entry_t* e = raft_get_entry_from_idx(me_, ae->prev_log_idx); if (!e) { __log(me_, "AE no log at prev_idx %d", ae->prev_log_idx); r->success = 0; return 0; } /* 2. Reply false if log doesn't contain an entry at prevLogIndex whose term matches prevLogTerm (§5.3) */ if (e->term != ae->prev_log_term) { __log(me_, "AE term doesn't match prev_idx (ie. %d vs %d)", e->term, ae->prev_log_term); r->success = 0; return 0; } } /* 3. If an existing entry conflicts with a new one (same index but different terms), delete the existing entry and all that follow it (§5.3) */ raft_entry_t* e2 = raft_get_entry_from_idx(me_, ae->prev_log_idx + 1); if (e2) log_delete(me->log, ae->prev_log_idx + 1); /* 4. If leaderCommit > commitIndex, set commitIndex = min(leaderCommit, last log index) */ if (raft_get_commit_idx(me_) < ae->leader_commit) { int last_log_idx = max(raft_get_current_idx(me_) - 1, 1); raft_set_commit_idx(me_, min(last_log_idx, ae->leader_commit)); while (0 == raft_apply_entry(me_)) ; } if (raft_is_candidate(me_)) raft_become_follower(me_); raft_set_current_term(me_, ae->term); /* update current leader because we accepted appendentries from it */ me->current_leader = node; int i; /* append all entries to log */ for (i = 0; i < ae->n_entries; i++) { msg_entry_t* cmd = &ae->entries[i]; raft_entry_t c; c.term = cmd->term; c.id = cmd->id; memcpy(&c.data, &cmd->data, sizeof(raft_entry_data_t)); int e = raft_append_entry(me_, &c); if (-1 == e) { __log(me_, "AE failure; couldn't append entry"); r->success = 0; return -1; } } r->success = 1; r->current_idx = raft_get_current_idx(me_); r->first_idx = ae->prev_log_idx + 1; return 0; } static int __should_grant_vote(raft_server_private_t* me, msg_requestvote_t* vr) { if (vr->term < raft_get_current_term((void*)me)) return 0; /* we've already voted */ if (-1 != me->voted_for) return 0; /* we have a more up-to-date log */ if (vr->last_log_idx < raft_get_current_idx((void*)me)) return 0; return 1; } int raft_recv_requestvote(raft_server_t* me_, int node, msg_requestvote_t* vr, msg_requestvote_response_t *r) { raft_server_private_t* me = (raft_server_private_t*)me_; if (raft_get_current_term(me_) < vr->term) raft_become_follower(me_); if (__should_grant_vote(me, vr)) { /* It shouldn't be possible for a leader or candidate to grant a vote * Both states would have voted for themselves */ assert(!(raft_is_leader(me_) || raft_is_candidate(me_))); raft_vote(me_, node); r->vote_granted = 1; } else r->vote_granted = 0; /* voted for someone, therefore must be in an election. */ if (0 <= me->voted_for) me->current_leader = -1; __log(me_, "node requested vote: %d replying: %s", node, r->vote_granted == 1 ? "granted" : "not granted"); r->term = raft_get_current_term(me_); return 0; } int raft_votes_is_majority(const int num_nodes, const int nvotes) { if (num_nodes < nvotes) return 0; int half = num_nodes / 2; return half + 1 <= nvotes; } int raft_recv_requestvote_response(raft_server_t* me_, int node, msg_requestvote_response_t* r) { raft_server_private_t* me = (raft_server_private_t*)me_; __log(me_, "node responded to requestvote: %d status: %s", node, r->vote_granted == 1 ? "granted" : "not granted"); if (raft_is_leader(me_)) return 0; assert(node < me->num_nodes); // TODO: if invalid leader then stepdown // if (r->term != raft_get_current_term(me_)) // return 0; if (1 == r->vote_granted) { me->votes_for_me[node] = 1; int votes = raft_get_nvotes_for_me(me_); if (raft_votes_is_majority(me->num_nodes, votes)) raft_become_leader(me_); } return 0; } int raft_recv_entry(raft_server_t* me_, int node, msg_entry_t* e, msg_entry_response_t *r) { raft_server_private_t* me = (raft_server_private_t*)me_; raft_entry_t ety; int i; if (!raft_is_leader(me_)) return -1; __log(me_, "received entry from: %d", node); ety.term = me->current_term; ety.id = e->id; memcpy(&ety.data, &e->data, e->data.len); raft_append_entry(me_, &ety); for (i = 0; i < me->num_nodes; i++) if (me->nodeid != i) raft_send_appendentries(me_, i); r->id = e->id; r->idx = raft_get_current_idx(me_); r->term = me->current_term; return 0; } int raft_send_requestvote(raft_server_t* me_, int node) { raft_server_private_t* me = (raft_server_private_t*)me_; msg_requestvote_t rv; __log(me_, "sending requestvote to: %d", node); rv.term = me->current_term; rv.last_log_idx = raft_get_current_idx(me_); rv.last_log_term = raft_get_current_term(me_); rv.candidate_id = raft_get_nodeid(me_); if (me->cb.send_requestvote) me->cb.send_requestvote(me_, me->udata, node, &rv); return 0; } int raft_append_entry(raft_server_t* me_, raft_entry_t* c) { raft_server_private_t* me = (raft_server_private_t*)me_; return log_append_entry(me->log, c); } int raft_apply_entry(raft_server_t* me_) { raft_server_private_t* me = (raft_server_private_t*)me_; raft_entry_t* e = raft_get_entry_from_idx(me_, me->last_applied_idx + 1); if (!e) return -1; __log(me_, "applying log: %d, size: %d", me->last_applied_idx, e->data.len); me->last_applied_idx++; if (me->commit_idx < me->last_applied_idx) me->commit_idx = me->last_applied_idx; if (me->cb.applylog) me->cb.applylog(me_, me->udata, e->data.buf, e->data.len); return 0; } void raft_send_appendentries(raft_server_t* me_, int node) { raft_server_private_t* me = (raft_server_private_t*)me_; if (!(me->cb.send_appendentries)) return; raft_node_t* p = raft_get_node(me_, node); msg_appendentries_t ae; ae.term = me->current_term; ae.leader_commit = raft_get_commit_idx(me_); ae.prev_log_idx = 0; ae.prev_log_term = 0; ae.n_entries = 0; ae.entries = NULL; int next_idx = raft_node_get_next_idx(p); msg_entry_t mety; raft_entry_t* ety = raft_get_entry_from_idx(me_, next_idx); if (ety) { mety.term = ety->term; mety.id = ety->id; mety.data.len = ety->data.len; mety.data.buf = ety->data.buf; ae.entries = &mety; // TODO: we want to send more than 1 at a time ae.n_entries = 1; } /* previous log is the log just before the new logs */ if (1 < next_idx) { raft_entry_t* prev_ety = raft_get_entry_from_idx(me_, next_idx - 1); ae.prev_log_idx = next_idx - 1; if (prev_ety) ae.prev_log_term = prev_ety->term; } __log(me_, "sending appendentries node: %d, %d %d %d %d", node, ae.term, ae.leader_commit, ae.prev_log_idx, ae.prev_log_term); me->cb.send_appendentries(me_, me->udata, node, &ae); } void raft_send_appendentries_all(raft_server_t* me_) { raft_server_private_t* me = (raft_server_private_t*)me_; int i; for (i = 0; i < me->num_nodes; i++) if (me->nodeid != i) raft_send_appendentries(me_, i); } void raft_set_configuration(raft_server_t* me_, raft_node_configuration_t* nodes, int my_idx) { raft_server_private_t* me = (raft_server_private_t*)me_; int num_nodes; /* TODO: one memory allocation only please */ for (num_nodes = 0; nodes->udata_address; nodes++) { num_nodes++; me->nodes = (raft_node_t*)realloc(me->nodes, sizeof(raft_node_t*) * num_nodes); me->num_nodes = num_nodes; me->nodes[num_nodes - 1] = raft_node_new(nodes->udata_address); } me->votes_for_me = (int*)calloc(num_nodes, sizeof(int)); me->nodeid = my_idx; } int raft_add_node(raft_server_t* me_, void* udata, int is_self) { raft_server_private_t* me = (raft_server_private_t*)me_; /* TODO: does not yet support dynamic membership changes */ if (me->current_term != 0 && me->timeout_elapsed != 0 && me->election_timeout != 0) return -1; me->num_nodes++; me->nodes = (raft_node_t*)realloc(me->nodes, sizeof(raft_node_t*) * me->num_nodes); me->nodes[me->num_nodes - 1] = raft_node_new(udata); me->votes_for_me = (int*)realloc(me->votes_for_me, me->num_nodes * sizeof(int)); if (is_self) me->nodeid = me->num_nodes - 1; return 0; } int raft_get_nvotes_for_me(raft_server_t* me_) { raft_server_private_t* me = (raft_server_private_t*)me_; int i, votes; for (i = 0, votes = 0; i < me->num_nodes; i++) if (me->nodeid != i) if (1 == me->votes_for_me[i]) votes += 1; if (me->voted_for == me->nodeid) votes += 1; return votes; } void raft_vote(raft_server_t* me_, const int node) { raft_server_private_t* me = (raft_server_private_t*)me_; me->voted_for = node; if (me->cb.persist_vote) me->cb.persist_vote(me_, me->udata, node); } int raft_msg_entry_response_committed(raft_server_t* me_, const msg_entry_response_t* r) { raft_entry_t* ety = raft_get_entry_from_idx(me_, r->idx); if (!ety) return 0; /* entry from another leader has invalidated this entry message */ if (r->term != ety->term) return -1; return r->idx <= raft_get_commit_idx(me_); }