/** * 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_, raft_node_t* node, 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_, node, 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(); me->voting_cfg_change_log_idx = -1; raft_set_state((raft_server_t*)me, RAFT_STATE_FOLLOWER); me->current_leader = NULL; 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_, NULL, "election starting: %d %d, term: %d ci: %d", me->election_timeout, me->timeout_elapsed, me->current_term, raft_get_current_idx(me_)); 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_, NULL, "becoming leader term:%d", raft_get_current_term(me_)); raft_set_state(me_, RAFT_STATE_LEADER); for (i = 0; i < me->num_nodes; i++) { if (me->node == me->nodes[i] || !raft_node_is_voting(me->nodes[i])) continue; raft_node_t* node = me->nodes[i]; raft_node_set_next_idx(node, raft_get_current_idx(me_) + 1); raft_node_set_match_idx(node, 0); raft_send_appendentries(me_, node); } } void raft_become_candidate(raft_server_t* me_) { raft_server_private_t* me = (raft_server_private_t*)me_; int i; __log(me_, NULL, "becoming candidate"); raft_set_current_term(me_, raft_get_current_term(me_) + 1); for (i = 0; i < me->num_nodes; i++) raft_node_vote_for_me(me->nodes[i], 0); raft_vote(me_, me->node); me->current_leader = NULL; raft_set_state(me_, RAFT_STATE_CANDIDATE); /* we need a random factor here to prevent simultaneous candidates */ /* TODO: this should probably be lower */ me->timeout_elapsed = rand() % me->election_timeout; for (i = 0; i < me->num_nodes; i++) if (me->node != me->nodes[i] && raft_node_is_voting(me->nodes[i])) raft_send_requestvote(me_, me->nodes[i]); } void raft_become_follower(raft_server_t* me_) { __log(me_, NULL, "becoming follower"); raft_set_state(me_, RAFT_STATE_FOLLOWER); } int raft_periodic(raft_server_t* me_, int msec_since_last_period) { raft_server_private_t* me = (raft_server_private_t*)me_; 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_); } else if (me->election_timeout <= me->timeout_elapsed) { if (1 < me->num_nodes) raft_election_start(me_); } if (me->last_applied_idx < me->commit_idx) if (-1 == raft_apply_entry(me_)) return -1; 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_, raft_node_t* node, msg_appendentries_response_t* r) { raft_server_private_t* me = (raft_server_private_t*)me_; __log(me_, node, "received appendentries response %s ci:%d rci:%d 1stidx:%d", r->success == 1 ? "SUCCESS" : "fail", raft_get_current_idx(me_), r->current_idx, r->first_idx); /* Stale response -- ignore */ if (r->current_idx != 0 && r->current_idx <= raft_node_get_match_idx(node)) return 0; if (!raft_is_leader(me_)) return -1; /* If response contains term T > currentTerm: set currentTerm = T and convert to follower (§5.3) */ if (me->current_term < r->term) { raft_set_current_term(me_, r->term); raft_become_follower(me_); return 0; } else if (me->current_term != r->term) return 0; /* stop processing, this is a node we don't have in our configuration */ if (!node) return 0; if (0 == r->success) { /* If AppendEntries fails because of log inconsistency: decrement nextIndex and retry (§5.3) */ assert(0 <= raft_node_get_next_idx(node)); int next_idx = raft_node_get_next_idx(node); assert(0 <= next_idx); if (r->current_idx < next_idx - 1) raft_node_set_next_idx(node, min(r->current_idx + 1, raft_get_current_idx(me_))); else raft_node_set_next_idx(node, next_idx - 1); /* retry */ raft_send_appendentries(me_, node); return 0; } assert(r->current_idx <= raft_get_current_idx(me_)); raft_node_set_next_idx(node, r->current_idx + 1); raft_node_set_match_idx(node, r->current_idx); if (!raft_node_is_voting(node) && -1 == me->voting_cfg_change_log_idx && raft_get_current_idx(me_) <= r->current_idx + 1 && me->cb.node_has_sufficient_logs && 0 == raft_node_has_sufficient_logs(node) ) { raft_node_set_has_sufficient_logs(node); me->cb.node_has_sufficient_logs(me_, me->udata, node); } /* Update commit idx */ int votes = 1; /* include me */ int point = r->current_idx; int i; for (i = 0; i < me->num_nodes; i++) { if (me->node == me->nodes[i] || !raft_node_is_voting(me->nodes[i])) continue; int match_idx = raft_node_get_match_idx(me->nodes[i]); if (0 < match_idx) { raft_entry_t* ety = raft_get_entry_from_idx(me_, match_idx); if (ety->term == me->current_term && point <= match_idx) votes++; } } if (me->num_nodes / 2 < votes && raft_get_commit_idx(me_) < point) raft_set_commit_idx(me_, point); /* Aggressively send remaining entries */ if (raft_get_entry_from_idx(me_, raft_node_get_next_idx(node))) raft_send_appendentries(me_, node); /* periodic applies committed entries lazily */ return 0; } int raft_recv_appendentries( raft_server_t* me_, raft_node_t* 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_, node, "recvd appendentries from: %lx, t:%d ci:%d lc:%d pli:%d plt:%d #%d", node, ae->term, raft_get_current_idx(me_), ae->leader_commit, ae->prev_log_idx, ae->prev_log_term, ae->n_entries); r->term = me->current_term; if (raft_is_candidate(me_) && me->current_term == ae->term) { me->voted_for = -1; raft_become_follower(me_); } else if (me->current_term < ae->term) { raft_set_current_term(me_, ae->term); r->term = ae->term; raft_become_follower(me_); } else if (ae->term < me->current_term) { /* 1. Reply false if term < currentTerm (§5.1) */ __log(me_, node, "AE term %d is less than current term %d", ae->term, me->current_term); goto fail_with_current_idx; } /* 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_, node, "AE no log at prev_idx %d", ae->prev_log_idx); goto fail_with_current_idx; } /* 2. Reply false if log doesn't contain an entry at prevLogIndex whose term matches prevLogTerm (§5.3) */ if (raft_get_current_idx(me_) < ae->prev_log_idx) goto fail_with_current_idx; if (e->term != ae->prev_log_term) { __log(me_, node, "AE term doesn't match prev_term (ie. %d vs %d) ci:%d pli:%d", e->term, ae->prev_log_term, raft_get_current_idx(me_), ae->prev_log_idx); assert(me->commit_idx < ae->prev_log_idx); /* Delete all the following log entries because they don't match */ log_delete(me->log, ae->prev_log_idx); r->current_idx = ae->prev_log_idx - 1; goto fail; } } /* 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) */ if (ae->n_entries == 0 && 0 < ae->prev_log_idx && ae->prev_log_idx + 1 < raft_get_current_idx(me_)) { assert(me->commit_idx < ae->prev_log_idx + 1); log_delete(me->log, ae->prev_log_idx + 1); } r->current_idx = ae->prev_log_idx; int i; for (i = 0; i < ae->n_entries; i++) { msg_entry_t* ety = &ae->entries[i]; int ety_index = ae->prev_log_idx + 1 + i; raft_entry_t* existing_ety = raft_get_entry_from_idx(me_, ety_index); r->current_idx = ety_index; if (existing_ety && existing_ety->term != ety->term) { assert(me->commit_idx < ety_index); log_delete(me->log, ety_index); break; } else if (!existing_ety) break; } /* Pick up remainder in case of mismatch or missing entry */ for (; i < ae->n_entries; i++) { int e = raft_append_entry(me_, &ae->entries[i]); if (-1 == e) goto fail_with_current_idx; r->current_idx = ae->prev_log_idx + 1 + i; } /* 4. If leaderCommit > commitIndex, set commitIndex = min(leaderCommit, index of most recent entry) */ if (raft_get_commit_idx(me_) < ae->leader_commit) { int last_log_idx = max(raft_get_current_idx(me_), 1); raft_set_commit_idx(me_, min(last_log_idx, ae->leader_commit)); } /* update current leader because we accepted appendentries from it */ me->current_leader = node; r->success = 1; r->first_idx = ae->prev_log_idx + 1; return 0; fail_with_current_idx: r->current_idx = raft_get_current_idx(me_); fail: r->success = 0; r->first_idx = 0; return -1; } 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; /* TODO: if voted for is candiate return 1 (if below checks pass) */ /* we've already voted */ if (0 < me->voted_for) return 0; int current_idx = raft_get_current_idx((void*)me); if (0 == current_idx) return 1; raft_entry_t* e = raft_get_entry_from_idx((void*)me, current_idx); if (e->term < vr->last_log_term) return 1; if (vr->last_log_term == e->term && current_idx <= vr->last_log_idx) return 1; return 0; } int raft_recv_requestvote(raft_server_t* me_, raft_node_t* 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_set_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_for_nodeid(me_, vr->candidate_id); r->vote_granted = 1; /* there must be in an election. */ me->current_leader = NULL; me->timeout_elapsed = 0; } else r->vote_granted = 0; __log(me_, node, "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_, raft_node_t* node, msg_requestvote_response_t* r) { raft_server_private_t* me = (raft_server_private_t*)me_; __log(me_, node, "node responded to requestvote status: %s", r->vote_granted == 1 ? "granted" : "not granted"); if (!raft_is_candidate(me_)) { return 0; } else if (raft_get_current_term(me_) < r->term) { raft_set_current_term(me_, r->term); raft_become_follower(me_); return 0; } else if (raft_get_current_term(me_) != r->term) { /* The node who voted for us would have obtained our term. * Therefore this is an old message we should ignore. * This happens if the network is pretty choppy. */ return 0; } __log(me_, node, "node responded to requestvote: %d status: %s ct:%d rt:%d", node, r->vote_granted == 1 ? "granted" : "not granted", me->current_term, r->term); if (1 == r->vote_granted) { if (node) raft_node_vote_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_, msg_entry_t* e, msg_entry_response_t *r) { raft_server_private_t* me = (raft_server_private_t*)me_; int i; /* Only one voting cfg change at a time */ if (raft_entry_is_voting_cfg_change(e)) if (-1 != me->voting_cfg_change_log_idx) return -1; if (!raft_is_leader(me_)) return -1; __log(me_, NULL, "received entry t:%d id: %d idx: %d", me->current_term, e->id, raft_get_current_idx(me_) + 1); raft_entry_t ety; ety.term = me->current_term; ety.id = e->id; ety.type = e->type; memcpy(&ety.data, &e->data, sizeof(raft_entry_data_t)); raft_append_entry(me_, &ety); for (i = 0; i < me->num_nodes; i++) { if (me->node == me->nodes[i] || !me->nodes[i] || !raft_node_is_voting(me->nodes[i])) continue; /* Only send new entries. * Don't send the entry to peers who are behind, to prevent them from * becoming congested. */ int next_idx = raft_node_get_next_idx(me->nodes[i]); if (next_idx == raft_get_current_idx(me_)) raft_send_appendentries(me_, me->nodes[i]); } /* if we're the only node, we can consider the entry committed */ if (1 == me->num_nodes) me->commit_idx = raft_get_current_idx(me_); r->id = e->id; r->idx = raft_get_current_idx(me_); r->term = me->current_term; if (raft_entry_is_voting_cfg_change(e)) me->voting_cfg_change_log_idx = raft_get_current_idx(me_); return 0; } int raft_send_requestvote(raft_server_t* me_, raft_node_t* node) { raft_server_private_t* me = (raft_server_private_t*)me_; msg_requestvote_t rv; assert(node); assert(node != me->node); __log(me_, node, "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_last_log_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* ety) { raft_server_private_t* me = (raft_server_private_t*)me_; if (raft_entry_is_voting_cfg_change(ety)) me->voting_cfg_change_log_idx = raft_get_current_idx(me_); return log_append_entry(me->log, ety); } int raft_apply_entry(raft_server_t* me_) { raft_server_private_t* me = (raft_server_private_t*)me_; /* Don't apply after the commit_idx */ if (me->last_applied_idx == me->commit_idx) return -1; int log_idx = me->last_applied_idx + 1; raft_entry_t* e = raft_get_entry_from_idx(me_, log_idx); if (!e) return -1; __log(me_, NULL, "applying log: %d, id: %d size: %d", me->last_applied_idx, e->id, e->data.len); me->last_applied_idx++; if (me->cb.applylog) me->cb.applylog(me_, me->udata, e); /* voting cfg change is now complete */ if (log_idx == me->voting_cfg_change_log_idx) me->voting_cfg_change_log_idx = -1; return 0; } int raft_send_appendentries(raft_server_t* me_, raft_node_t* node) { raft_server_private_t* me = (raft_server_private_t*)me_; assert(node); assert(node != me->node); if (!(me->cb.send_appendentries)) return -1; 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(node); 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.type = ety->type; 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_, node, "sending appendentries node: ci:%d t:%d lc:%d pli:%d plt:%d", raft_get_current_idx(me_), ae.term, ae.leader_commit, ae.prev_log_idx, ae.prev_log_term); me->cb.send_appendentries(me_, me->udata, node, &ae); return 0; } void raft_send_appendentries_all(raft_server_t* me_) { raft_server_private_t* me = (raft_server_private_t*)me_; int i; me->timeout_elapsed = 0; for (i = 0; i < me->num_nodes; i++) if (me->node != me->nodes[i]) raft_send_appendentries(me_, me->nodes[i]); } raft_node_t* raft_add_node(raft_server_t* me_, void* udata, int id, int is_self) { raft_server_private_t* me = (raft_server_private_t*)me_; /* set to voting if node already exists */ raft_node_t* node = raft_get_node(me_, id); if (node) { raft_node_set_voting(node, 1); return node; } 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, id); assert(me->nodes[me->num_nodes - 1]); if (is_self) me->node = me->nodes[me->num_nodes - 1]; return me->nodes[me->num_nodes - 1]; } raft_node_t* raft_add_non_voting_node(raft_server_t* me_, void* udata, int id, int is_self) { raft_node_t* node = raft_add_node(me_, udata, id, is_self); raft_node_set_voting(node, 0); return node; } void raft_remove_node(raft_server_t* me_, raft_node_t* node) { raft_server_private_t* me = (raft_server_private_t*)me_; raft_node_t* new_array, *new_node; new_array = (raft_node_t*)calloc((me->num_nodes - 1), sizeof(raft_node_t*)); new_node = new_array; for (int i = 0; inum_nodes; i++) { if (me->nodes[i] == node) continue; *new_node = me->nodes[i]; new_node++; } me->num_nodes--; free(me->nodes); me->nodes = new_array; free(node); } 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->node != me->nodes[i] && raft_node_is_voting(me->nodes[i])) if (raft_node_has_vote_for_me(me->nodes[i])) votes += 1; if (me->voted_for == raft_get_nodeid(me_)) votes += 1; return votes; } void raft_vote(raft_server_t* me_, raft_node_t* node) { raft_vote_for_nodeid(me_, node ? raft_node_get_id(node) : -1); } void raft_vote_for_nodeid(raft_server_t* me_, const int nodeid) { raft_server_private_t* me = (raft_server_private_t*)me_; me->voted_for = nodeid; if (me->cb.persist_vote) me->cb.persist_vote(me_, me->udata, nodeid); } 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_); } void raft_apply_all(raft_server_t* me_) { while (raft_get_last_applied_idx(me_) < raft_get_commit_idx(me_)) raft_apply_entry(me_); } int raft_entry_is_voting_cfg_change(raft_entry_t* ety) { return RAFT_LOGTYPE_ADD_NODE == ety->type || RAFT_LOGTYPE_REMOVE_NODE == ety->type; } int raft_entry_is_cfg_change(raft_entry_t* ety) { return ( RAFT_LOGTYPE_ADD_NODE == ety->type || RAFT_LOGTYPE_ADD_NONVOTING_NODE == ety->type || RAFT_LOGTYPE_REMOVE_NODE == ety->type); }