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
|
// Copyright (C) MongoDB, Inc. 2017-present.
//
// Licensed under the Apache License, Version 2.0 (the "License"); you may
// not use this file except in compliance with the License. You may obtain
// a copy of the License at http://www.apache.org/licenses/LICENSE-2.0
package description
import (
"encoding/json"
"fmt"
"math"
"time"
"go.mongodb.org/mongo-driver/mongo/readpref"
"go.mongodb.org/mongo-driver/tag"
)
// ServerSelector is an interface implemented by types that can perform server selection given a topology description
// and list of candidate servers. The selector should filter the provided candidates list and return a subset that
// matches some criteria.
type ServerSelector interface {
SelectServer(Topology, []Server) ([]Server, error)
}
// ServerSelectorFunc is a function that can be used as a ServerSelector.
type ServerSelectorFunc func(Topology, []Server) ([]Server, error)
// SelectServer implements the ServerSelector interface.
func (ssf ServerSelectorFunc) SelectServer(t Topology, s []Server) ([]Server, error) {
return ssf(t, s)
}
// serverSelectorInfo contains metadata concerning the server selector for the
// purpose of publication.
type serverSelectorInfo struct {
Type string
Data string `json:",omitempty"`
Selectors []serverSelectorInfo `json:",omitempty"`
}
// String returns the JSON string representation of the serverSelectorInfo.
func (sss serverSelectorInfo) String() string {
bytes, _ := json.Marshal(sss)
return string(bytes)
}
// serverSelectorInfoGetter is an interface that defines an info() method to
// get the serverSelectorInfo.
type serverSelectorInfoGetter interface {
info() serverSelectorInfo
}
type compositeSelector struct {
selectors []ServerSelector
}
func (cs *compositeSelector) info() serverSelectorInfo {
csInfo := serverSelectorInfo{Type: "compositeSelector"}
for _, sel := range cs.selectors {
if getter, ok := sel.(serverSelectorInfoGetter); ok {
csInfo.Selectors = append(csInfo.Selectors, getter.info())
}
}
return csInfo
}
// String returns the JSON string representation of the compositeSelector.
func (cs *compositeSelector) String() string {
return cs.info().String()
}
// CompositeSelector combines multiple selectors into a single selector by applying them in order to the candidates
// list.
//
// For example, if the initial candidates list is [s0, s1, s2, s3] and two selectors are provided where the first
// matches s0 and s1 and the second matches s1 and s2, the following would occur during server selection:
//
// 1. firstSelector([s0, s1, s2, s3]) -> [s0, s1]
// 2. secondSelector([s0, s1]) -> [s1]
//
// The final list of candidates returned by the composite selector would be [s1].
func CompositeSelector(selectors []ServerSelector) ServerSelector {
return &compositeSelector{selectors: selectors}
}
func (cs *compositeSelector) SelectServer(t Topology, candidates []Server) ([]Server, error) {
var err error
for _, sel := range cs.selectors {
candidates, err = sel.SelectServer(t, candidates)
if err != nil {
return nil, err
}
}
return candidates, nil
}
type latencySelector struct {
latency time.Duration
}
// LatencySelector creates a ServerSelector which selects servers based on their average RTT values.
func LatencySelector(latency time.Duration) ServerSelector {
return &latencySelector{latency: latency}
}
func (latencySelector) info() serverSelectorInfo {
return serverSelectorInfo{Type: "latencySelector"}
}
func (selector latencySelector) String() string {
return selector.info().String()
}
func (selector *latencySelector) SelectServer(t Topology, candidates []Server) ([]Server, error) {
if selector.latency < 0 {
return candidates, nil
}
if t.Kind == LoadBalanced {
// In LoadBalanced mode, there should only be one server in the topology and it must be selected.
return candidates, nil
}
switch len(candidates) {
case 0, 1:
return candidates, nil
default:
min := time.Duration(math.MaxInt64)
for _, candidate := range candidates {
if candidate.AverageRTTSet {
if candidate.AverageRTT < min {
min = candidate.AverageRTT
}
}
}
if min == math.MaxInt64 {
return candidates, nil
}
max := min + selector.latency
viableIndexes := make([]int, 0, len(candidates))
for i, candidate := range candidates {
if candidate.AverageRTTSet {
if candidate.AverageRTT <= max {
viableIndexes = append(viableIndexes, i)
}
}
}
if len(viableIndexes) == len(candidates) {
return candidates, nil
}
result := make([]Server, len(viableIndexes))
for i, idx := range viableIndexes {
result[i] = candidates[idx]
}
return result, nil
}
}
type writeServerSelector struct{}
// WriteSelector selects all the writable servers.
func WriteSelector() ServerSelector {
return writeServerSelector{}
}
func (writeServerSelector) info() serverSelectorInfo {
return serverSelectorInfo{Type: "writeSelector"}
}
func (selector writeServerSelector) String() string {
return selector.info().String()
}
func (writeServerSelector) SelectServer(t Topology, candidates []Server) ([]Server, error) {
switch t.Kind {
case Single, LoadBalanced:
return candidates, nil
default:
// Determine the capacity of the results slice.
selected := 0
for _, candidate := range candidates {
switch candidate.Kind {
case Mongos, RSPrimary, Standalone:
selected++
}
}
// Append candidates to the results slice.
result := make([]Server, 0, selected)
for _, candidate := range candidates {
switch candidate.Kind {
case Mongos, RSPrimary, Standalone:
result = append(result, candidate)
}
}
return result, nil
}
}
type readPrefServerSelector struct {
rp *readpref.ReadPref
isOutputAggregate bool
}
// ReadPrefSelector selects servers based on the provided read preference.
func ReadPrefSelector(rp *readpref.ReadPref) ServerSelector {
return readPrefServerSelector{
rp: rp,
isOutputAggregate: false,
}
}
func (selector readPrefServerSelector) info() serverSelectorInfo {
return serverSelectorInfo{
Type: "readPrefSelector",
Data: selector.rp.String(),
}
}
func (selector readPrefServerSelector) String() string {
return selector.info().String()
}
func (selector readPrefServerSelector) SelectServer(t Topology, candidates []Server) ([]Server, error) {
if t.Kind == LoadBalanced {
// In LoadBalanced mode, there should only be one server in the topology and it must be selected. We check
// this before checking MaxStaleness support because there's no monitoring in this mode, so the candidate
// server wouldn't have a wire version set, which would result in an error.
return candidates, nil
}
switch t.Kind {
case Single:
return candidates, nil
case ReplicaSetNoPrimary, ReplicaSetWithPrimary:
return selectForReplicaSet(selector.rp, selector.isOutputAggregate, t, candidates)
case Sharded:
return selectByKind(candidates, Mongos), nil
}
return nil, nil
}
// OutputAggregateSelector selects servers based on the provided read preference
// given that the underlying operation is aggregate with an output stage.
func OutputAggregateSelector(rp *readpref.ReadPref) ServerSelector {
return readPrefServerSelector{
rp: rp,
isOutputAggregate: true,
}
}
func selectForReplicaSet(rp *readpref.ReadPref, isOutputAggregate bool, t Topology, candidates []Server) ([]Server, error) {
if err := verifyMaxStaleness(rp, t); err != nil {
return nil, err
}
// If underlying operation is an aggregate with an output stage, only apply read preference
// if all candidates are 5.0+. Otherwise, operate under primary read preference.
if isOutputAggregate {
for _, s := range candidates {
if s.WireVersion.Max < 13 {
return selectByKind(candidates, RSPrimary), nil
}
}
}
switch rp.Mode() {
case readpref.PrimaryMode:
return selectByKind(candidates, RSPrimary), nil
case readpref.PrimaryPreferredMode:
selected := selectByKind(candidates, RSPrimary)
if len(selected) == 0 {
selected = selectSecondaries(rp, candidates)
return selectByTagSet(selected, rp.TagSets()), nil
}
return selected, nil
case readpref.SecondaryPreferredMode:
selected := selectSecondaries(rp, candidates)
selected = selectByTagSet(selected, rp.TagSets())
if len(selected) > 0 {
return selected, nil
}
return selectByKind(candidates, RSPrimary), nil
case readpref.SecondaryMode:
selected := selectSecondaries(rp, candidates)
return selectByTagSet(selected, rp.TagSets()), nil
case readpref.NearestMode:
selected := selectByKind(candidates, RSPrimary)
selected = append(selected, selectSecondaries(rp, candidates)...)
return selectByTagSet(selected, rp.TagSets()), nil
}
return nil, fmt.Errorf("unsupported mode: %d", rp.Mode())
}
func selectSecondaries(rp *readpref.ReadPref, candidates []Server) []Server {
secondaries := selectByKind(candidates, RSSecondary)
if len(secondaries) == 0 {
return secondaries
}
if maxStaleness, set := rp.MaxStaleness(); set {
primaries := selectByKind(candidates, RSPrimary)
if len(primaries) == 0 {
baseTime := secondaries[0].LastWriteTime
for i := 1; i < len(secondaries); i++ {
if secondaries[i].LastWriteTime.After(baseTime) {
baseTime = secondaries[i].LastWriteTime
}
}
var selected []Server
for _, secondary := range secondaries {
estimatedStaleness := baseTime.Sub(secondary.LastWriteTime) + secondary.HeartbeatInterval
if estimatedStaleness <= maxStaleness {
selected = append(selected, secondary)
}
}
return selected
}
primary := primaries[0]
var selected []Server
for _, secondary := range secondaries {
estimatedStaleness := secondary.LastUpdateTime.Sub(secondary.LastWriteTime) - primary.LastUpdateTime.Sub(primary.LastWriteTime) + secondary.HeartbeatInterval
if estimatedStaleness <= maxStaleness {
selected = append(selected, secondary)
}
}
return selected
}
return secondaries
}
func selectByTagSet(candidates []Server, tagSets []tag.Set) []Server {
if len(tagSets) == 0 {
return candidates
}
for _, ts := range tagSets {
// If this tag set is empty, we can take a fast path because the empty list is a subset of all tag sets, so
// all candidate servers will be selected.
if len(ts) == 0 {
return candidates
}
var results []Server
for _, s := range candidates {
// ts is non-empty, so only servers with a non-empty set of tags need to be checked.
if len(s.Tags) > 0 && s.Tags.ContainsAll(ts) {
results = append(results, s)
}
}
if len(results) > 0 {
return results
}
}
return []Server{}
}
func selectByKind(candidates []Server, kind ServerKind) []Server {
// Record the indices of viable candidates first and then append those to the returned slice
// to avoid appending costly Server structs directly as an optimization.
viableIndexes := make([]int, 0, len(candidates))
for i, s := range candidates {
if s.Kind == kind {
viableIndexes = append(viableIndexes, i)
}
}
if len(viableIndexes) == len(candidates) {
return candidates
}
result := make([]Server, len(viableIndexes))
for i, idx := range viableIndexes {
result[i] = candidates[idx]
}
return result
}
func verifyMaxStaleness(rp *readpref.ReadPref, t Topology) error {
maxStaleness, set := rp.MaxStaleness()
if !set {
return nil
}
if maxStaleness < 90*time.Second {
return fmt.Errorf("max staleness (%s) must be greater than or equal to 90s", maxStaleness)
}
if len(t.Servers) < 1 {
// Maybe we should return an error here instead?
return nil
}
// we'll assume all candidates have the same heartbeat interval.
s := t.Servers[0]
idleWritePeriod := 10 * time.Second
if maxStaleness < s.HeartbeatInterval+idleWritePeriod {
return fmt.Errorf(
"max staleness (%s) must be greater than or equal to the heartbeat interval (%s) plus idle write period (%s)",
maxStaleness, s.HeartbeatInterval, idleWritePeriod,
)
}
return nil
}
|