| Abhay Kumar | 40252eb | 2025-10-13 13:25:53 +0000 | [diff] [blame^] | 1 | // Copyright 2024 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 | //go:build with_tla |
| 16 | |
| 17 | package raft |
| 18 | |
| 19 | import ( |
| 20 | "strconv" |
| 21 | "time" |
| 22 | |
| 23 | "go.etcd.io/raft/v3/raftpb" |
| 24 | "go.etcd.io/raft/v3/tracker" |
| 25 | ) |
| 26 | |
| 27 | const StateTraceDeployed = true |
| 28 | |
| 29 | type stateMachineEventType int |
| 30 | |
| 31 | const ( |
| 32 | rsmInitState stateMachineEventType = iota |
| 33 | rsmBecomeCandidate |
| 34 | rsmBecomeFollower |
| 35 | rsmBecomeLeader |
| 36 | rsmCommit |
| 37 | rsmReplicate |
| 38 | rsmChangeConf |
| 39 | rsmApplyConfChange |
| 40 | rsmReady |
| 41 | rsmSendAppendEntriesRequest |
| 42 | rsmReceiveAppendEntriesRequest |
| 43 | rsmSendAppendEntriesResponse |
| 44 | rsmReceiveAppendEntriesResponse |
| 45 | rsmSendRequestVoteRequest |
| 46 | rsmReceiveRequestVoteRequest |
| 47 | rsmSendRequestVoteResponse |
| 48 | rsmReceiveRequestVoteResponse |
| 49 | rsmSendSnapshot |
| 50 | rsmReceiveSnapshot |
| 51 | ) |
| 52 | |
| 53 | func (e stateMachineEventType) String() string { |
| 54 | return []string{ |
| 55 | "InitState", |
| 56 | "BecomeCandidate", |
| 57 | "BecomeFollower", |
| 58 | "BecomeLeader", |
| 59 | "Commit", |
| 60 | "Replicate", |
| 61 | "ChangeConf", |
| 62 | "ApplyConfChange", |
| 63 | "Ready", |
| 64 | "SendAppendEntriesRequest", |
| 65 | "ReceiveAppendEntriesRequest", |
| 66 | "SendAppendEntriesResponse", |
| 67 | "ReceiveAppendEntriesResponse", |
| 68 | "SendRequestVoteRequest", |
| 69 | "ReceiveRequestVoteRequest", |
| 70 | "SendRequestVoteResponse", |
| 71 | "ReceiveRequestVoteResponse", |
| 72 | "SendSnapshot", |
| 73 | "ReceiveSnapshot", |
| 74 | }[e] |
| 75 | } |
| 76 | |
| 77 | const ( |
| 78 | ConfChangeAddNewServer string = "AddNewServer" |
| 79 | ConfChangeRemoveServer string = "RemoveServer" |
| 80 | ConfChangeAddLearner string = "AddLearner" |
| 81 | ) |
| 82 | |
| 83 | type TracingEvent struct { |
| 84 | Name string `json:"name"` |
| 85 | NodeID string `json:"nid"` |
| 86 | State TracingState `json:"state"` |
| 87 | Role string `json:"role"` |
| 88 | LogSize uint64 `json:"log"` |
| 89 | Conf [2][]string `json:"conf"` |
| 90 | Message *TracingMessage `json:"msg,omitempty"` |
| 91 | ConfChange *TracingConfChange `json:"cc,omitempty"` |
| 92 | Properties map[string]any `json:"prop,omitempty"` |
| 93 | } |
| 94 | |
| 95 | type TracingState struct { |
| 96 | Term uint64 `json:"term"` |
| 97 | Vote string `json:"vote"` |
| 98 | Commit uint64 `json:"commit"` |
| 99 | } |
| 100 | |
| 101 | type TracingMessage struct { |
| 102 | Type string `json:"type"` |
| 103 | Term uint64 `json:"term"` |
| 104 | From string `json:"from"` |
| 105 | To string `json:"to"` |
| 106 | EntryLength int `json:"entries"` |
| 107 | LogTerm uint64 `json:"logTerm"` |
| 108 | Index uint64 `json:"index"` |
| 109 | Commit uint64 `json:"commit"` |
| 110 | Vote string `json:"vote"` |
| 111 | Reject bool `json:"reject"` |
| 112 | RejectHint uint64 `json:"rejectHint"` |
| 113 | } |
| 114 | |
| 115 | type SingleConfChange struct { |
| 116 | NodeID string `json:"nid"` |
| 117 | Action string `json:"action"` |
| 118 | } |
| 119 | |
| 120 | type TracingConfChange struct { |
| 121 | Changes []SingleConfChange `json:"changes,omitempty"` |
| 122 | NewConf []string `json:"newconf,omitempty"` |
| 123 | } |
| 124 | |
| 125 | func makeTracingState(r *raft) TracingState { |
| 126 | hs := r.hardState() |
| 127 | return TracingState{ |
| 128 | Term: hs.Term, |
| 129 | Vote: strconv.FormatUint(hs.Vote, 10), |
| 130 | Commit: hs.Commit, |
| 131 | } |
| 132 | } |
| 133 | |
| 134 | func makeTracingMessage(m *raftpb.Message) *TracingMessage { |
| 135 | if m == nil { |
| 136 | return nil |
| 137 | } |
| 138 | |
| 139 | logTerm := m.LogTerm |
| 140 | entries := len(m.Entries) |
| 141 | index := m.Index |
| 142 | if m.Type == raftpb.MsgSnap { |
| 143 | index = 0 |
| 144 | logTerm = 0 |
| 145 | entries = int(m.Snapshot.Metadata.Index) |
| 146 | } |
| 147 | return &TracingMessage{ |
| 148 | Type: m.Type.String(), |
| 149 | Term: m.Term, |
| 150 | From: strconv.FormatUint(m.From, 10), |
| 151 | To: strconv.FormatUint(m.To, 10), |
| 152 | EntryLength: entries, |
| 153 | LogTerm: logTerm, |
| 154 | Index: index, |
| 155 | Commit: m.Commit, |
| 156 | Vote: strconv.FormatUint(m.Vote, 10), |
| 157 | Reject: m.Reject, |
| 158 | RejectHint: m.RejectHint, |
| 159 | } |
| 160 | } |
| 161 | |
| 162 | type TraceLogger interface { |
| 163 | TraceEvent(*TracingEvent) |
| 164 | } |
| 165 | |
| 166 | func traceEvent(evt stateMachineEventType, r *raft, m *raftpb.Message, prop map[string]any) { |
| 167 | if r.traceLogger == nil { |
| 168 | return |
| 169 | } |
| 170 | |
| 171 | r.traceLogger.TraceEvent(&TracingEvent{ |
| 172 | Name: evt.String(), |
| 173 | NodeID: strconv.FormatUint(r.id, 10), |
| 174 | State: makeTracingState(r), |
| 175 | LogSize: r.raftLog.lastIndex(), |
| 176 | Conf: [2][]string{formatConf(r.trk.Voters[0].Slice()), formatConf(r.trk.Voters[1].Slice())}, |
| 177 | Role: r.state.String(), |
| 178 | Message: makeTracingMessage(m), |
| 179 | Properties: prop, |
| 180 | }) |
| 181 | } |
| 182 | |
| 183 | func traceNodeEvent(evt stateMachineEventType, r *raft) { |
| 184 | traceEvent(evt, r, nil, nil) |
| 185 | } |
| 186 | |
| 187 | func formatConf(s []uint64) []string { |
| 188 | if s == nil { |
| 189 | return []string{} |
| 190 | } |
| 191 | |
| 192 | r := make([]string, len(s)) |
| 193 | for i, v := range s { |
| 194 | r[i] = strconv.FormatUint(v, 10) |
| 195 | } |
| 196 | return r |
| 197 | } |
| 198 | |
| 199 | // Use following helper functions to trace specific state and/or |
| 200 | // transition at corresponding code lines |
| 201 | func traceInitState(r *raft) { |
| 202 | if r.traceLogger == nil { |
| 203 | return |
| 204 | } |
| 205 | |
| 206 | traceNodeEvent(rsmInitState, r) |
| 207 | } |
| 208 | |
| 209 | func traceReady(r *raft) { |
| 210 | traceNodeEvent(rsmReady, r) |
| 211 | } |
| 212 | |
| 213 | func traceCommit(r *raft) { |
| 214 | traceNodeEvent(rsmCommit, r) |
| 215 | } |
| 216 | |
| 217 | func traceReplicate(r *raft, es ...raftpb.Entry) { |
| 218 | for i := range es { |
| 219 | if es[i].Type == raftpb.EntryNormal { |
| 220 | traceNodeEvent(rsmReplicate, r) |
| 221 | } |
| 222 | } |
| 223 | } |
| 224 | |
| 225 | func traceBecomeFollower(r *raft) { |
| 226 | traceNodeEvent(rsmBecomeFollower, r) |
| 227 | } |
| 228 | |
| 229 | func traceBecomeCandidate(r *raft) { |
| 230 | traceNodeEvent(rsmBecomeCandidate, r) |
| 231 | } |
| 232 | |
| 233 | func traceBecomeLeader(r *raft) { |
| 234 | traceNodeEvent(rsmBecomeLeader, r) |
| 235 | } |
| 236 | |
| 237 | func traceChangeConfEvent(cci raftpb.ConfChangeI, r *raft) { |
| 238 | cc2 := cci.AsV2() |
| 239 | cc := &TracingConfChange{ |
| 240 | Changes: []SingleConfChange{}, |
| 241 | NewConf: []string{}, |
| 242 | } |
| 243 | for _, c := range cc2.Changes { |
| 244 | switch c.Type { |
| 245 | case raftpb.ConfChangeAddNode: |
| 246 | cc.Changes = append(cc.Changes, SingleConfChange{ |
| 247 | NodeID: strconv.FormatUint(c.NodeID, 10), |
| 248 | Action: ConfChangeAddNewServer, |
| 249 | }) |
| 250 | case raftpb.ConfChangeRemoveNode: |
| 251 | cc.Changes = append(cc.Changes, SingleConfChange{ |
| 252 | NodeID: strconv.FormatUint(c.NodeID, 10), |
| 253 | Action: ConfChangeRemoveServer, |
| 254 | }) |
| 255 | case raftpb.ConfChangeAddLearnerNode: |
| 256 | cc.Changes = append(cc.Changes, SingleConfChange{ |
| 257 | NodeID: strconv.FormatUint(c.NodeID, 10), |
| 258 | Action: ConfChangeAddLearner, |
| 259 | }) |
| 260 | } |
| 261 | } |
| 262 | |
| 263 | if len(cc.Changes) == 0 { |
| 264 | return |
| 265 | } |
| 266 | |
| 267 | p := map[string]any{} |
| 268 | p["cc"] = cc |
| 269 | traceEvent(rsmChangeConf, r, nil, p) |
| 270 | } |
| 271 | |
| 272 | func traceConfChangeEvent(cfg tracker.Config, r *raft) { |
| 273 | if r.traceLogger == nil { |
| 274 | return |
| 275 | } |
| 276 | |
| 277 | cc := &TracingConfChange{ |
| 278 | Changes: []SingleConfChange{}, |
| 279 | NewConf: formatConf(cfg.Voters[0].Slice()), |
| 280 | } |
| 281 | |
| 282 | p := map[string]any{} |
| 283 | p["cc"] = cc |
| 284 | traceEvent(rsmApplyConfChange, r, nil, p) |
| 285 | } |
| 286 | |
| 287 | func traceSendMessage(r *raft, m *raftpb.Message) { |
| 288 | if r.traceLogger == nil { |
| 289 | return |
| 290 | } |
| 291 | |
| 292 | prop := map[string]any{} |
| 293 | |
| 294 | var evt stateMachineEventType |
| 295 | switch m.Type { |
| 296 | case raftpb.MsgApp: |
| 297 | evt = rsmSendAppendEntriesRequest |
| 298 | if p, exist := r.trk.Progress[m.From]; exist { |
| 299 | prop["match"] = p.Match |
| 300 | prop["next"] = p.Next |
| 301 | } |
| 302 | |
| 303 | case raftpb.MsgHeartbeat, raftpb.MsgSnap: |
| 304 | evt = rsmSendAppendEntriesRequest |
| 305 | case raftpb.MsgAppResp, raftpb.MsgHeartbeatResp: |
| 306 | evt = rsmSendAppendEntriesResponse |
| 307 | case raftpb.MsgVote: |
| 308 | evt = rsmSendRequestVoteRequest |
| 309 | case raftpb.MsgVoteResp: |
| 310 | evt = rsmSendRequestVoteResponse |
| 311 | default: |
| 312 | return |
| 313 | } |
| 314 | |
| 315 | traceEvent(evt, r, m, prop) |
| 316 | } |
| 317 | |
| 318 | func traceReceiveMessage(r *raft, m *raftpb.Message) { |
| 319 | if r.traceLogger == nil { |
| 320 | return |
| 321 | } |
| 322 | |
| 323 | var evt stateMachineEventType |
| 324 | switch m.Type { |
| 325 | case raftpb.MsgApp, raftpb.MsgHeartbeat, raftpb.MsgSnap: |
| 326 | evt = rsmReceiveAppendEntriesRequest |
| 327 | case raftpb.MsgAppResp, raftpb.MsgHeartbeatResp: |
| 328 | evt = rsmReceiveAppendEntriesResponse |
| 329 | case raftpb.MsgVote: |
| 330 | evt = rsmReceiveRequestVoteRequest |
| 331 | case raftpb.MsgVoteResp: |
| 332 | evt = rsmReceiveRequestVoteResponse |
| 333 | default: |
| 334 | return |
| 335 | } |
| 336 | |
| 337 | time.Sleep(time.Millisecond) // sleep 1ms to reduce time shift impact accross node |
| 338 | traceEvent(evt, r, m, nil) |
| 339 | } |