Scott Baker | eee8dd8 | 2019-09-24 12:52:34 -0700 | [diff] [blame] | 1 | // Copyright 2015 The etcd Authors |
| 2 | // |
| 3 | // Licensed under the Apache License, Version 2.0 (the "License"); |
| 4 | // you may not use this file except in compliance with the License. |
| 5 | // You may obtain a copy of the License at |
| 6 | // |
| 7 | // http://www.apache.org/licenses/LICENSE-2.0 |
| 8 | // |
| 9 | // Unless required by applicable law or agreed to in writing, software |
| 10 | // distributed under the License is distributed on an "AS IS" BASIS, |
| 11 | // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 12 | // See the License for the specific language governing permissions and |
| 13 | // limitations under the License. |
| 14 | |
| 15 | package raft |
| 16 | |
| 17 | import ( |
| 18 | "errors" |
| 19 | |
| 20 | pb "go.etcd.io/etcd/raft/raftpb" |
Scott Baker | 611f6bd | 2019-10-18 13:45:19 -0700 | [diff] [blame] | 21 | "go.etcd.io/etcd/raft/tracker" |
Scott Baker | eee8dd8 | 2019-09-24 12:52:34 -0700 | [diff] [blame] | 22 | ) |
| 23 | |
| 24 | // ErrStepLocalMsg is returned when try to step a local raft message |
| 25 | var ErrStepLocalMsg = errors.New("raft: cannot step raft local message") |
| 26 | |
| 27 | // ErrStepPeerNotFound is returned when try to step a response message |
| 28 | // but there is no peer found in raft.prs for that node. |
| 29 | var ErrStepPeerNotFound = errors.New("raft: cannot step as peer not found") |
| 30 | |
| 31 | // RawNode is a thread-unsafe Node. |
| 32 | // The methods of this struct correspond to the methods of Node and are described |
| 33 | // more fully there. |
| 34 | type RawNode struct { |
| 35 | raft *raft |
| 36 | prevSoftSt *SoftState |
| 37 | prevHardSt pb.HardState |
| 38 | } |
| 39 | |
Scott Baker | 611f6bd | 2019-10-18 13:45:19 -0700 | [diff] [blame] | 40 | // NewRawNode instantiates a RawNode from the given configuration. |
| 41 | // |
| 42 | // See Bootstrap() for bootstrapping an initial state; this replaces the former |
| 43 | // 'peers' argument to this method (with identical behavior). However, It is |
| 44 | // recommended that instead of calling Bootstrap, applications bootstrap their |
| 45 | // state manually by setting up a Storage that has a first index > 1 and which |
| 46 | // stores the desired ConfState as its InitialState. |
| 47 | func NewRawNode(config *Config) (*RawNode, error) { |
Scott Baker | eee8dd8 | 2019-09-24 12:52:34 -0700 | [diff] [blame] | 48 | r := newRaft(config) |
| 49 | rn := &RawNode{ |
| 50 | raft: r, |
| 51 | } |
Scott Baker | eee8dd8 | 2019-09-24 12:52:34 -0700 | [diff] [blame] | 52 | rn.prevSoftSt = r.softState() |
Scott Baker | 611f6bd | 2019-10-18 13:45:19 -0700 | [diff] [blame] | 53 | rn.prevHardSt = r.hardState() |
Scott Baker | eee8dd8 | 2019-09-24 12:52:34 -0700 | [diff] [blame] | 54 | return rn, nil |
| 55 | } |
| 56 | |
| 57 | // Tick advances the internal logical clock by a single tick. |
| 58 | func (rn *RawNode) Tick() { |
| 59 | rn.raft.tick() |
| 60 | } |
| 61 | |
| 62 | // TickQuiesced advances the internal logical clock by a single tick without |
| 63 | // performing any other state machine processing. It allows the caller to avoid |
| 64 | // periodic heartbeats and elections when all of the peers in a Raft group are |
| 65 | // known to be at the same state. Expected usage is to periodically invoke Tick |
| 66 | // or TickQuiesced depending on whether the group is "active" or "quiesced". |
| 67 | // |
| 68 | // WARNING: Be very careful about using this method as it subverts the Raft |
| 69 | // state machine. You should probably be using Tick instead. |
| 70 | func (rn *RawNode) TickQuiesced() { |
| 71 | rn.raft.electionElapsed++ |
| 72 | } |
| 73 | |
| 74 | // Campaign causes this RawNode to transition to candidate state. |
| 75 | func (rn *RawNode) Campaign() error { |
| 76 | return rn.raft.Step(pb.Message{ |
| 77 | Type: pb.MsgHup, |
| 78 | }) |
| 79 | } |
| 80 | |
| 81 | // Propose proposes data be appended to the raft log. |
| 82 | func (rn *RawNode) Propose(data []byte) error { |
| 83 | return rn.raft.Step(pb.Message{ |
| 84 | Type: pb.MsgProp, |
| 85 | From: rn.raft.id, |
| 86 | Entries: []pb.Entry{ |
| 87 | {Data: data}, |
| 88 | }}) |
| 89 | } |
| 90 | |
Scott Baker | 611f6bd | 2019-10-18 13:45:19 -0700 | [diff] [blame] | 91 | // ProposeConfChange proposes a config change. See (Node).ProposeConfChange for |
| 92 | // details. |
| 93 | func (rn *RawNode) ProposeConfChange(cc pb.ConfChangeI) error { |
| 94 | m, err := confChangeToMsg(cc) |
Scott Baker | eee8dd8 | 2019-09-24 12:52:34 -0700 | [diff] [blame] | 95 | if err != nil { |
| 96 | return err |
| 97 | } |
Scott Baker | 611f6bd | 2019-10-18 13:45:19 -0700 | [diff] [blame] | 98 | return rn.raft.Step(m) |
Scott Baker | eee8dd8 | 2019-09-24 12:52:34 -0700 | [diff] [blame] | 99 | } |
| 100 | |
| 101 | // ApplyConfChange applies a config change to the local node. |
Scott Baker | 611f6bd | 2019-10-18 13:45:19 -0700 | [diff] [blame] | 102 | func (rn *RawNode) ApplyConfChange(cc pb.ConfChangeI) *pb.ConfState { |
| 103 | cs := rn.raft.applyConfChange(cc.AsV2()) |
| 104 | return &cs |
Scott Baker | eee8dd8 | 2019-09-24 12:52:34 -0700 | [diff] [blame] | 105 | } |
| 106 | |
| 107 | // Step advances the state machine using the given message. |
| 108 | func (rn *RawNode) Step(m pb.Message) error { |
| 109 | // ignore unexpected local messages receiving over network |
| 110 | if IsLocalMsg(m.Type) { |
| 111 | return ErrStepLocalMsg |
| 112 | } |
Scott Baker | 611f6bd | 2019-10-18 13:45:19 -0700 | [diff] [blame] | 113 | if pr := rn.raft.prs.Progress[m.From]; pr != nil || !IsResponseMsg(m.Type) { |
Scott Baker | eee8dd8 | 2019-09-24 12:52:34 -0700 | [diff] [blame] | 114 | return rn.raft.Step(m) |
| 115 | } |
| 116 | return ErrStepPeerNotFound |
| 117 | } |
| 118 | |
Scott Baker | 611f6bd | 2019-10-18 13:45:19 -0700 | [diff] [blame] | 119 | // Ready returns the outstanding work that the application needs to handle. This |
| 120 | // includes appending and applying entries or a snapshot, updating the HardState, |
| 121 | // and sending messages. The returned Ready() *must* be handled and subsequently |
| 122 | // passed back via Advance(). |
Scott Baker | eee8dd8 | 2019-09-24 12:52:34 -0700 | [diff] [blame] | 123 | func (rn *RawNode) Ready() Ready { |
Scott Baker | 611f6bd | 2019-10-18 13:45:19 -0700 | [diff] [blame] | 124 | rd := rn.readyWithoutAccept() |
| 125 | rn.acceptReady(rd) |
Scott Baker | eee8dd8 | 2019-09-24 12:52:34 -0700 | [diff] [blame] | 126 | return rd |
| 127 | } |
| 128 | |
Scott Baker | 611f6bd | 2019-10-18 13:45:19 -0700 | [diff] [blame] | 129 | // readyWithoutAccept returns a Ready. This is a read-only operation, i.e. there |
| 130 | // is no obligation that the Ready must be handled. |
| 131 | func (rn *RawNode) readyWithoutAccept() Ready { |
| 132 | return newReady(rn.raft, rn.prevSoftSt, rn.prevHardSt) |
| 133 | } |
| 134 | |
| 135 | // acceptReady is called when the consumer of the RawNode has decided to go |
| 136 | // ahead and handle a Ready. Nothing must alter the state of the RawNode between |
| 137 | // this call and the prior call to Ready(). |
| 138 | func (rn *RawNode) acceptReady(rd Ready) { |
| 139 | if rd.SoftState != nil { |
| 140 | rn.prevSoftSt = rd.SoftState |
| 141 | } |
| 142 | if len(rd.ReadStates) != 0 { |
| 143 | rn.raft.readStates = nil |
| 144 | } |
| 145 | rn.raft.msgs = nil |
| 146 | } |
| 147 | |
Scott Baker | eee8dd8 | 2019-09-24 12:52:34 -0700 | [diff] [blame] | 148 | // HasReady called when RawNode user need to check if any Ready pending. |
| 149 | // Checking logic in this method should be consistent with Ready.containsUpdates(). |
| 150 | func (rn *RawNode) HasReady() bool { |
| 151 | r := rn.raft |
| 152 | if !r.softState().equal(rn.prevSoftSt) { |
| 153 | return true |
| 154 | } |
| 155 | if hardSt := r.hardState(); !IsEmptyHardState(hardSt) && !isHardStateEqual(hardSt, rn.prevHardSt) { |
| 156 | return true |
| 157 | } |
| 158 | if r.raftLog.unstable.snapshot != nil && !IsEmptySnap(*r.raftLog.unstable.snapshot) { |
| 159 | return true |
| 160 | } |
| 161 | if len(r.msgs) > 0 || len(r.raftLog.unstableEntries()) > 0 || r.raftLog.hasNextEnts() { |
| 162 | return true |
| 163 | } |
| 164 | if len(r.readStates) != 0 { |
| 165 | return true |
| 166 | } |
| 167 | return false |
| 168 | } |
| 169 | |
| 170 | // Advance notifies the RawNode that the application has applied and saved progress in the |
| 171 | // last Ready results. |
| 172 | func (rn *RawNode) Advance(rd Ready) { |
Scott Baker | 611f6bd | 2019-10-18 13:45:19 -0700 | [diff] [blame] | 173 | if !IsEmptyHardState(rd.HardState) { |
| 174 | rn.prevHardSt = rd.HardState |
| 175 | } |
| 176 | rn.raft.advance(rd) |
Scott Baker | eee8dd8 | 2019-09-24 12:52:34 -0700 | [diff] [blame] | 177 | } |
| 178 | |
Scott Baker | 611f6bd | 2019-10-18 13:45:19 -0700 | [diff] [blame] | 179 | // Status returns the current status of the given group. This allocates, see |
| 180 | // BasicStatus and WithProgress for allocation-friendlier choices. |
| 181 | func (rn *RawNode) Status() Status { |
Scott Baker | eee8dd8 | 2019-09-24 12:52:34 -0700 | [diff] [blame] | 182 | status := getStatus(rn.raft) |
Scott Baker | 611f6bd | 2019-10-18 13:45:19 -0700 | [diff] [blame] | 183 | return status |
Scott Baker | eee8dd8 | 2019-09-24 12:52:34 -0700 | [diff] [blame] | 184 | } |
| 185 | |
Scott Baker | 611f6bd | 2019-10-18 13:45:19 -0700 | [diff] [blame] | 186 | // BasicStatus returns a BasicStatus. Notably this does not contain the |
| 187 | // Progress map; see WithProgress for an allocation-free way to inspect it. |
| 188 | func (rn *RawNode) BasicStatus() BasicStatus { |
| 189 | return getBasicStatus(rn.raft) |
Scott Baker | eee8dd8 | 2019-09-24 12:52:34 -0700 | [diff] [blame] | 190 | } |
| 191 | |
| 192 | // ProgressType indicates the type of replica a Progress corresponds to. |
| 193 | type ProgressType byte |
| 194 | |
| 195 | const ( |
| 196 | // ProgressTypePeer accompanies a Progress for a regular peer replica. |
| 197 | ProgressTypePeer ProgressType = iota |
| 198 | // ProgressTypeLearner accompanies a Progress for a learner replica. |
| 199 | ProgressTypeLearner |
| 200 | ) |
| 201 | |
| 202 | // WithProgress is a helper to introspect the Progress for this node and its |
| 203 | // peers. |
Scott Baker | 611f6bd | 2019-10-18 13:45:19 -0700 | [diff] [blame] | 204 | func (rn *RawNode) WithProgress(visitor func(id uint64, typ ProgressType, pr tracker.Progress)) { |
| 205 | rn.raft.prs.Visit(func(id uint64, pr *tracker.Progress) { |
| 206 | typ := ProgressTypePeer |
| 207 | if pr.IsLearner { |
| 208 | typ = ProgressTypeLearner |
| 209 | } |
| 210 | p := *pr |
| 211 | p.Inflights = nil |
| 212 | visitor(id, typ, p) |
| 213 | }) |
Scott Baker | eee8dd8 | 2019-09-24 12:52:34 -0700 | [diff] [blame] | 214 | } |
| 215 | |
| 216 | // ReportUnreachable reports the given node is not reachable for the last send. |
| 217 | func (rn *RawNode) ReportUnreachable(id uint64) { |
| 218 | _ = rn.raft.Step(pb.Message{Type: pb.MsgUnreachable, From: id}) |
| 219 | } |
| 220 | |
| 221 | // ReportSnapshot reports the status of the sent snapshot. |
| 222 | func (rn *RawNode) ReportSnapshot(id uint64, status SnapshotStatus) { |
| 223 | rej := status == SnapshotFailure |
| 224 | |
| 225 | _ = rn.raft.Step(pb.Message{Type: pb.MsgSnapStatus, From: id, Reject: rej}) |
| 226 | } |
| 227 | |
| 228 | // TransferLeader tries to transfer leadership to the given transferee. |
| 229 | func (rn *RawNode) TransferLeader(transferee uint64) { |
| 230 | _ = rn.raft.Step(pb.Message{Type: pb.MsgTransferLeader, From: transferee}) |
| 231 | } |
| 232 | |
| 233 | // ReadIndex requests a read state. The read state will be set in ready. |
| 234 | // Read State has a read index. Once the application advances further than the read |
| 235 | // index, any linearizable read requests issued before the read request can be |
| 236 | // processed safely. The read state will have the same rctx attached. |
| 237 | func (rn *RawNode) ReadIndex(rctx []byte) { |
| 238 | _ = rn.raft.Step(pb.Message{Type: pb.MsgReadIndex, Entries: []pb.Entry{{Data: rctx}}}) |
| 239 | } |