Naveen Sampath | 04696f7 | 2022-06-13 15:19:14 +0530 | [diff] [blame] | 1 | package redis |
| 2 | |
| 3 | import ( |
| 4 | "context" |
| 5 | "crypto/tls" |
| 6 | "errors" |
| 7 | "net" |
| 8 | "strings" |
| 9 | "sync" |
| 10 | "time" |
| 11 | |
| 12 | "github.com/go-redis/redis/v8/internal" |
| 13 | "github.com/go-redis/redis/v8/internal/pool" |
| 14 | "github.com/go-redis/redis/v8/internal/rand" |
| 15 | ) |
| 16 | |
| 17 | //------------------------------------------------------------------------------ |
| 18 | |
| 19 | // FailoverOptions are used to configure a failover client and should |
| 20 | // be passed to NewFailoverClient. |
| 21 | type FailoverOptions struct { |
| 22 | // The master name. |
| 23 | MasterName string |
| 24 | // A seed list of host:port addresses of sentinel nodes. |
| 25 | SentinelAddrs []string |
| 26 | |
| 27 | // If specified with SentinelPassword, enables ACL-based authentication (via |
| 28 | // AUTH <user> <pass>). |
| 29 | SentinelUsername string |
| 30 | // Sentinel password from "requirepass <password>" (if enabled) in Sentinel |
| 31 | // configuration, or, if SentinelUsername is also supplied, used for ACL-based |
| 32 | // authentication. |
| 33 | SentinelPassword string |
| 34 | |
| 35 | // Allows routing read-only commands to the closest master or slave node. |
| 36 | // This option only works with NewFailoverClusterClient. |
| 37 | RouteByLatency bool |
| 38 | // Allows routing read-only commands to the random master or slave node. |
| 39 | // This option only works with NewFailoverClusterClient. |
| 40 | RouteRandomly bool |
| 41 | |
| 42 | // Route all commands to slave read-only nodes. |
| 43 | SlaveOnly bool |
| 44 | |
| 45 | // Use slaves disconnected with master when cannot get connected slaves |
| 46 | // Now, this option only works in RandomSlaveAddr function. |
| 47 | UseDisconnectedSlaves bool |
| 48 | |
| 49 | // Following options are copied from Options struct. |
| 50 | |
| 51 | Dialer func(ctx context.Context, network, addr string) (net.Conn, error) |
| 52 | OnConnect func(ctx context.Context, cn *Conn) error |
| 53 | |
| 54 | Username string |
| 55 | Password string |
| 56 | DB int |
| 57 | |
| 58 | MaxRetries int |
| 59 | MinRetryBackoff time.Duration |
| 60 | MaxRetryBackoff time.Duration |
| 61 | |
| 62 | DialTimeout time.Duration |
| 63 | ReadTimeout time.Duration |
| 64 | WriteTimeout time.Duration |
| 65 | |
| 66 | // PoolFIFO uses FIFO mode for each node connection pool GET/PUT (default LIFO). |
| 67 | PoolFIFO bool |
| 68 | |
| 69 | PoolSize int |
| 70 | MinIdleConns int |
| 71 | MaxConnAge time.Duration |
| 72 | PoolTimeout time.Duration |
| 73 | IdleTimeout time.Duration |
| 74 | IdleCheckFrequency time.Duration |
| 75 | |
| 76 | TLSConfig *tls.Config |
| 77 | } |
| 78 | |
| 79 | func (opt *FailoverOptions) clientOptions() *Options { |
| 80 | return &Options{ |
| 81 | Addr: "FailoverClient", |
| 82 | |
| 83 | Dialer: opt.Dialer, |
| 84 | OnConnect: opt.OnConnect, |
| 85 | |
| 86 | DB: opt.DB, |
| 87 | Username: opt.Username, |
| 88 | Password: opt.Password, |
| 89 | |
| 90 | MaxRetries: opt.MaxRetries, |
| 91 | MinRetryBackoff: opt.MinRetryBackoff, |
| 92 | MaxRetryBackoff: opt.MaxRetryBackoff, |
| 93 | |
| 94 | DialTimeout: opt.DialTimeout, |
| 95 | ReadTimeout: opt.ReadTimeout, |
| 96 | WriteTimeout: opt.WriteTimeout, |
| 97 | |
| 98 | PoolFIFO: opt.PoolFIFO, |
| 99 | PoolSize: opt.PoolSize, |
| 100 | PoolTimeout: opt.PoolTimeout, |
| 101 | IdleTimeout: opt.IdleTimeout, |
| 102 | IdleCheckFrequency: opt.IdleCheckFrequency, |
| 103 | MinIdleConns: opt.MinIdleConns, |
| 104 | MaxConnAge: opt.MaxConnAge, |
| 105 | |
| 106 | TLSConfig: opt.TLSConfig, |
| 107 | } |
| 108 | } |
| 109 | |
| 110 | func (opt *FailoverOptions) sentinelOptions(addr string) *Options { |
| 111 | return &Options{ |
| 112 | Addr: addr, |
| 113 | |
| 114 | Dialer: opt.Dialer, |
| 115 | OnConnect: opt.OnConnect, |
| 116 | |
| 117 | DB: 0, |
| 118 | Username: opt.SentinelUsername, |
| 119 | Password: opt.SentinelPassword, |
| 120 | |
| 121 | MaxRetries: opt.MaxRetries, |
| 122 | MinRetryBackoff: opt.MinRetryBackoff, |
| 123 | MaxRetryBackoff: opt.MaxRetryBackoff, |
| 124 | |
| 125 | DialTimeout: opt.DialTimeout, |
| 126 | ReadTimeout: opt.ReadTimeout, |
| 127 | WriteTimeout: opt.WriteTimeout, |
| 128 | |
| 129 | PoolFIFO: opt.PoolFIFO, |
| 130 | PoolSize: opt.PoolSize, |
| 131 | PoolTimeout: opt.PoolTimeout, |
| 132 | IdleTimeout: opt.IdleTimeout, |
| 133 | IdleCheckFrequency: opt.IdleCheckFrequency, |
| 134 | MinIdleConns: opt.MinIdleConns, |
| 135 | MaxConnAge: opt.MaxConnAge, |
| 136 | |
| 137 | TLSConfig: opt.TLSConfig, |
| 138 | } |
| 139 | } |
| 140 | |
| 141 | func (opt *FailoverOptions) clusterOptions() *ClusterOptions { |
| 142 | return &ClusterOptions{ |
| 143 | Dialer: opt.Dialer, |
| 144 | OnConnect: opt.OnConnect, |
| 145 | |
| 146 | Username: opt.Username, |
| 147 | Password: opt.Password, |
| 148 | |
| 149 | MaxRedirects: opt.MaxRetries, |
| 150 | |
| 151 | RouteByLatency: opt.RouteByLatency, |
| 152 | RouteRandomly: opt.RouteRandomly, |
| 153 | |
| 154 | MinRetryBackoff: opt.MinRetryBackoff, |
| 155 | MaxRetryBackoff: opt.MaxRetryBackoff, |
| 156 | |
| 157 | DialTimeout: opt.DialTimeout, |
| 158 | ReadTimeout: opt.ReadTimeout, |
| 159 | WriteTimeout: opt.WriteTimeout, |
| 160 | |
| 161 | PoolFIFO: opt.PoolFIFO, |
| 162 | PoolSize: opt.PoolSize, |
| 163 | PoolTimeout: opt.PoolTimeout, |
| 164 | IdleTimeout: opt.IdleTimeout, |
| 165 | IdleCheckFrequency: opt.IdleCheckFrequency, |
| 166 | MinIdleConns: opt.MinIdleConns, |
| 167 | MaxConnAge: opt.MaxConnAge, |
| 168 | |
| 169 | TLSConfig: opt.TLSConfig, |
| 170 | } |
| 171 | } |
| 172 | |
| 173 | // NewFailoverClient returns a Redis client that uses Redis Sentinel |
| 174 | // for automatic failover. It's safe for concurrent use by multiple |
| 175 | // goroutines. |
| 176 | func NewFailoverClient(failoverOpt *FailoverOptions) *Client { |
| 177 | if failoverOpt.RouteByLatency { |
| 178 | panic("to route commands by latency, use NewFailoverClusterClient") |
| 179 | } |
| 180 | if failoverOpt.RouteRandomly { |
| 181 | panic("to route commands randomly, use NewFailoverClusterClient") |
| 182 | } |
| 183 | |
| 184 | sentinelAddrs := make([]string, len(failoverOpt.SentinelAddrs)) |
| 185 | copy(sentinelAddrs, failoverOpt.SentinelAddrs) |
| 186 | |
| 187 | rand.Shuffle(len(sentinelAddrs), func(i, j int) { |
| 188 | sentinelAddrs[i], sentinelAddrs[j] = sentinelAddrs[j], sentinelAddrs[i] |
| 189 | }) |
| 190 | |
| 191 | failover := &sentinelFailover{ |
| 192 | opt: failoverOpt, |
| 193 | sentinelAddrs: sentinelAddrs, |
| 194 | } |
| 195 | |
| 196 | opt := failoverOpt.clientOptions() |
| 197 | opt.Dialer = masterSlaveDialer(failover) |
| 198 | opt.init() |
| 199 | |
| 200 | connPool := newConnPool(opt) |
| 201 | |
| 202 | failover.mu.Lock() |
| 203 | failover.onFailover = func(ctx context.Context, addr string) { |
| 204 | _ = connPool.Filter(func(cn *pool.Conn) bool { |
| 205 | return cn.RemoteAddr().String() != addr |
| 206 | }) |
| 207 | } |
| 208 | failover.mu.Unlock() |
| 209 | |
| 210 | c := Client{ |
| 211 | baseClient: newBaseClient(opt, connPool), |
| 212 | ctx: context.Background(), |
| 213 | } |
| 214 | c.cmdable = c.Process |
| 215 | c.onClose = failover.Close |
| 216 | |
| 217 | return &c |
| 218 | } |
| 219 | |
| 220 | func masterSlaveDialer( |
| 221 | failover *sentinelFailover, |
| 222 | ) func(ctx context.Context, network, addr string) (net.Conn, error) { |
| 223 | return func(ctx context.Context, network, _ string) (net.Conn, error) { |
| 224 | var addr string |
| 225 | var err error |
| 226 | |
| 227 | if failover.opt.SlaveOnly { |
| 228 | addr, err = failover.RandomSlaveAddr(ctx) |
| 229 | } else { |
| 230 | addr, err = failover.MasterAddr(ctx) |
| 231 | if err == nil { |
| 232 | failover.trySwitchMaster(ctx, addr) |
| 233 | } |
| 234 | } |
| 235 | if err != nil { |
| 236 | return nil, err |
| 237 | } |
| 238 | if failover.opt.Dialer != nil { |
| 239 | return failover.opt.Dialer(ctx, network, addr) |
| 240 | } |
| 241 | |
| 242 | netDialer := &net.Dialer{ |
| 243 | Timeout: failover.opt.DialTimeout, |
| 244 | KeepAlive: 5 * time.Minute, |
| 245 | } |
| 246 | if failover.opt.TLSConfig == nil { |
| 247 | return netDialer.DialContext(ctx, network, addr) |
| 248 | } |
| 249 | return tls.DialWithDialer(netDialer, network, addr, failover.opt.TLSConfig) |
| 250 | } |
| 251 | } |
| 252 | |
| 253 | //------------------------------------------------------------------------------ |
| 254 | |
| 255 | // SentinelClient is a client for a Redis Sentinel. |
| 256 | type SentinelClient struct { |
| 257 | *baseClient |
| 258 | hooks |
| 259 | ctx context.Context |
| 260 | } |
| 261 | |
| 262 | func NewSentinelClient(opt *Options) *SentinelClient { |
| 263 | opt.init() |
| 264 | c := &SentinelClient{ |
| 265 | baseClient: &baseClient{ |
| 266 | opt: opt, |
| 267 | connPool: newConnPool(opt), |
| 268 | }, |
| 269 | ctx: context.Background(), |
| 270 | } |
| 271 | return c |
| 272 | } |
| 273 | |
| 274 | func (c *SentinelClient) Context() context.Context { |
| 275 | return c.ctx |
| 276 | } |
| 277 | |
| 278 | func (c *SentinelClient) WithContext(ctx context.Context) *SentinelClient { |
| 279 | if ctx == nil { |
| 280 | panic("nil context") |
| 281 | } |
| 282 | clone := *c |
| 283 | clone.ctx = ctx |
| 284 | return &clone |
| 285 | } |
| 286 | |
| 287 | func (c *SentinelClient) Process(ctx context.Context, cmd Cmder) error { |
| 288 | return c.hooks.process(ctx, cmd, c.baseClient.process) |
| 289 | } |
| 290 | |
| 291 | func (c *SentinelClient) pubSub() *PubSub { |
| 292 | pubsub := &PubSub{ |
| 293 | opt: c.opt, |
| 294 | |
| 295 | newConn: func(ctx context.Context, channels []string) (*pool.Conn, error) { |
| 296 | return c.newConn(ctx) |
| 297 | }, |
| 298 | closeConn: c.connPool.CloseConn, |
| 299 | } |
| 300 | pubsub.init() |
| 301 | return pubsub |
| 302 | } |
| 303 | |
| 304 | // Ping is used to test if a connection is still alive, or to |
| 305 | // measure latency. |
| 306 | func (c *SentinelClient) Ping(ctx context.Context) *StringCmd { |
| 307 | cmd := NewStringCmd(ctx, "ping") |
| 308 | _ = c.Process(ctx, cmd) |
| 309 | return cmd |
| 310 | } |
| 311 | |
| 312 | // Subscribe subscribes the client to the specified channels. |
| 313 | // Channels can be omitted to create empty subscription. |
| 314 | func (c *SentinelClient) Subscribe(ctx context.Context, channels ...string) *PubSub { |
| 315 | pubsub := c.pubSub() |
| 316 | if len(channels) > 0 { |
| 317 | _ = pubsub.Subscribe(ctx, channels...) |
| 318 | } |
| 319 | return pubsub |
| 320 | } |
| 321 | |
| 322 | // PSubscribe subscribes the client to the given patterns. |
| 323 | // Patterns can be omitted to create empty subscription. |
| 324 | func (c *SentinelClient) PSubscribe(ctx context.Context, channels ...string) *PubSub { |
| 325 | pubsub := c.pubSub() |
| 326 | if len(channels) > 0 { |
| 327 | _ = pubsub.PSubscribe(ctx, channels...) |
| 328 | } |
| 329 | return pubsub |
| 330 | } |
| 331 | |
| 332 | func (c *SentinelClient) GetMasterAddrByName(ctx context.Context, name string) *StringSliceCmd { |
| 333 | cmd := NewStringSliceCmd(ctx, "sentinel", "get-master-addr-by-name", name) |
| 334 | _ = c.Process(ctx, cmd) |
| 335 | return cmd |
| 336 | } |
| 337 | |
| 338 | func (c *SentinelClient) Sentinels(ctx context.Context, name string) *SliceCmd { |
| 339 | cmd := NewSliceCmd(ctx, "sentinel", "sentinels", name) |
| 340 | _ = c.Process(ctx, cmd) |
| 341 | return cmd |
| 342 | } |
| 343 | |
| 344 | // Failover forces a failover as if the master was not reachable, and without |
| 345 | // asking for agreement to other Sentinels. |
| 346 | func (c *SentinelClient) Failover(ctx context.Context, name string) *StatusCmd { |
| 347 | cmd := NewStatusCmd(ctx, "sentinel", "failover", name) |
| 348 | _ = c.Process(ctx, cmd) |
| 349 | return cmd |
| 350 | } |
| 351 | |
| 352 | // Reset resets all the masters with matching name. The pattern argument is a |
| 353 | // glob-style pattern. The reset process clears any previous state in a master |
| 354 | // (including a failover in progress), and removes every slave and sentinel |
| 355 | // already discovered and associated with the master. |
| 356 | func (c *SentinelClient) Reset(ctx context.Context, pattern string) *IntCmd { |
| 357 | cmd := NewIntCmd(ctx, "sentinel", "reset", pattern) |
| 358 | _ = c.Process(ctx, cmd) |
| 359 | return cmd |
| 360 | } |
| 361 | |
| 362 | // FlushConfig forces Sentinel to rewrite its configuration on disk, including |
| 363 | // the current Sentinel state. |
| 364 | func (c *SentinelClient) FlushConfig(ctx context.Context) *StatusCmd { |
| 365 | cmd := NewStatusCmd(ctx, "sentinel", "flushconfig") |
| 366 | _ = c.Process(ctx, cmd) |
| 367 | return cmd |
| 368 | } |
| 369 | |
| 370 | // Master shows the state and info of the specified master. |
| 371 | func (c *SentinelClient) Master(ctx context.Context, name string) *StringStringMapCmd { |
| 372 | cmd := NewStringStringMapCmd(ctx, "sentinel", "master", name) |
| 373 | _ = c.Process(ctx, cmd) |
| 374 | return cmd |
| 375 | } |
| 376 | |
| 377 | // Masters shows a list of monitored masters and their state. |
| 378 | func (c *SentinelClient) Masters(ctx context.Context) *SliceCmd { |
| 379 | cmd := NewSliceCmd(ctx, "sentinel", "masters") |
| 380 | _ = c.Process(ctx, cmd) |
| 381 | return cmd |
| 382 | } |
| 383 | |
| 384 | // Slaves shows a list of slaves for the specified master and their state. |
| 385 | func (c *SentinelClient) Slaves(ctx context.Context, name string) *SliceCmd { |
| 386 | cmd := NewSliceCmd(ctx, "sentinel", "slaves", name) |
| 387 | _ = c.Process(ctx, cmd) |
| 388 | return cmd |
| 389 | } |
| 390 | |
| 391 | // CkQuorum checks if the current Sentinel configuration is able to reach the |
| 392 | // quorum needed to failover a master, and the majority needed to authorize the |
| 393 | // failover. This command should be used in monitoring systems to check if a |
| 394 | // Sentinel deployment is ok. |
| 395 | func (c *SentinelClient) CkQuorum(ctx context.Context, name string) *StringCmd { |
| 396 | cmd := NewStringCmd(ctx, "sentinel", "ckquorum", name) |
| 397 | _ = c.Process(ctx, cmd) |
| 398 | return cmd |
| 399 | } |
| 400 | |
| 401 | // Monitor tells the Sentinel to start monitoring a new master with the specified |
| 402 | // name, ip, port, and quorum. |
| 403 | func (c *SentinelClient) Monitor(ctx context.Context, name, ip, port, quorum string) *StringCmd { |
| 404 | cmd := NewStringCmd(ctx, "sentinel", "monitor", name, ip, port, quorum) |
| 405 | _ = c.Process(ctx, cmd) |
| 406 | return cmd |
| 407 | } |
| 408 | |
| 409 | // Set is used in order to change configuration parameters of a specific master. |
| 410 | func (c *SentinelClient) Set(ctx context.Context, name, option, value string) *StringCmd { |
| 411 | cmd := NewStringCmd(ctx, "sentinel", "set", name, option, value) |
| 412 | _ = c.Process(ctx, cmd) |
| 413 | return cmd |
| 414 | } |
| 415 | |
| 416 | // Remove is used in order to remove the specified master: the master will no |
| 417 | // longer be monitored, and will totally be removed from the internal state of |
| 418 | // the Sentinel. |
| 419 | func (c *SentinelClient) Remove(ctx context.Context, name string) *StringCmd { |
| 420 | cmd := NewStringCmd(ctx, "sentinel", "remove", name) |
| 421 | _ = c.Process(ctx, cmd) |
| 422 | return cmd |
| 423 | } |
| 424 | |
| 425 | //------------------------------------------------------------------------------ |
| 426 | |
| 427 | type sentinelFailover struct { |
| 428 | opt *FailoverOptions |
| 429 | |
| 430 | sentinelAddrs []string |
| 431 | |
| 432 | onFailover func(ctx context.Context, addr string) |
| 433 | onUpdate func(ctx context.Context) |
| 434 | |
| 435 | mu sync.RWMutex |
| 436 | _masterAddr string |
| 437 | sentinel *SentinelClient |
| 438 | pubsub *PubSub |
| 439 | } |
| 440 | |
| 441 | func (c *sentinelFailover) Close() error { |
| 442 | c.mu.Lock() |
| 443 | defer c.mu.Unlock() |
| 444 | if c.sentinel != nil { |
| 445 | return c.closeSentinel() |
| 446 | } |
| 447 | return nil |
| 448 | } |
| 449 | |
| 450 | func (c *sentinelFailover) closeSentinel() error { |
| 451 | firstErr := c.pubsub.Close() |
| 452 | c.pubsub = nil |
| 453 | |
| 454 | err := c.sentinel.Close() |
| 455 | if err != nil && firstErr == nil { |
| 456 | firstErr = err |
| 457 | } |
| 458 | c.sentinel = nil |
| 459 | |
| 460 | return firstErr |
| 461 | } |
| 462 | |
| 463 | func (c *sentinelFailover) RandomSlaveAddr(ctx context.Context) (string, error) { |
| 464 | if c.opt == nil { |
| 465 | return "", errors.New("opt is nil") |
| 466 | } |
| 467 | |
| 468 | addresses, err := c.slaveAddrs(ctx, false) |
| 469 | if err != nil { |
| 470 | return "", err |
| 471 | } |
| 472 | |
| 473 | if len(addresses) == 0 && c.opt.UseDisconnectedSlaves { |
| 474 | addresses, err = c.slaveAddrs(ctx, true) |
| 475 | if err != nil { |
| 476 | return "", err |
| 477 | } |
| 478 | } |
| 479 | |
| 480 | if len(addresses) == 0 { |
| 481 | return c.MasterAddr(ctx) |
| 482 | } |
| 483 | return addresses[rand.Intn(len(addresses))], nil |
| 484 | } |
| 485 | |
| 486 | func (c *sentinelFailover) MasterAddr(ctx context.Context) (string, error) { |
| 487 | c.mu.RLock() |
| 488 | sentinel := c.sentinel |
| 489 | c.mu.RUnlock() |
| 490 | |
| 491 | if sentinel != nil { |
| 492 | addr := c.getMasterAddr(ctx, sentinel) |
| 493 | if addr != "" { |
| 494 | return addr, nil |
| 495 | } |
| 496 | } |
| 497 | |
| 498 | c.mu.Lock() |
| 499 | defer c.mu.Unlock() |
| 500 | |
| 501 | if c.sentinel != nil { |
| 502 | addr := c.getMasterAddr(ctx, c.sentinel) |
| 503 | if addr != "" { |
| 504 | return addr, nil |
| 505 | } |
| 506 | _ = c.closeSentinel() |
| 507 | } |
| 508 | |
| 509 | for i, sentinelAddr := range c.sentinelAddrs { |
| 510 | sentinel := NewSentinelClient(c.opt.sentinelOptions(sentinelAddr)) |
| 511 | |
| 512 | masterAddr, err := sentinel.GetMasterAddrByName(ctx, c.opt.MasterName).Result() |
| 513 | if err != nil { |
| 514 | internal.Logger.Printf(ctx, "sentinel: GetMasterAddrByName master=%q failed: %s", |
| 515 | c.opt.MasterName, err) |
| 516 | _ = sentinel.Close() |
| 517 | continue |
| 518 | } |
| 519 | |
| 520 | // Push working sentinel to the top. |
| 521 | c.sentinelAddrs[0], c.sentinelAddrs[i] = c.sentinelAddrs[i], c.sentinelAddrs[0] |
| 522 | c.setSentinel(ctx, sentinel) |
| 523 | |
| 524 | addr := net.JoinHostPort(masterAddr[0], masterAddr[1]) |
| 525 | return addr, nil |
| 526 | } |
| 527 | |
| 528 | return "", errors.New("redis: all sentinels specified in configuration are unreachable") |
| 529 | } |
| 530 | |
| 531 | func (c *sentinelFailover) slaveAddrs(ctx context.Context, useDisconnected bool) ([]string, error) { |
| 532 | c.mu.RLock() |
| 533 | sentinel := c.sentinel |
| 534 | c.mu.RUnlock() |
| 535 | |
| 536 | if sentinel != nil { |
| 537 | addrs := c.getSlaveAddrs(ctx, sentinel) |
| 538 | if len(addrs) > 0 { |
| 539 | return addrs, nil |
| 540 | } |
| 541 | } |
| 542 | |
| 543 | c.mu.Lock() |
| 544 | defer c.mu.Unlock() |
| 545 | |
| 546 | if c.sentinel != nil { |
| 547 | addrs := c.getSlaveAddrs(ctx, c.sentinel) |
| 548 | if len(addrs) > 0 { |
| 549 | return addrs, nil |
| 550 | } |
| 551 | _ = c.closeSentinel() |
| 552 | } |
| 553 | |
| 554 | var sentinelReachable bool |
| 555 | |
| 556 | for i, sentinelAddr := range c.sentinelAddrs { |
| 557 | sentinel := NewSentinelClient(c.opt.sentinelOptions(sentinelAddr)) |
| 558 | |
| 559 | slaves, err := sentinel.Slaves(ctx, c.opt.MasterName).Result() |
| 560 | if err != nil { |
| 561 | internal.Logger.Printf(ctx, "sentinel: Slaves master=%q failed: %s", |
| 562 | c.opt.MasterName, err) |
| 563 | _ = sentinel.Close() |
| 564 | continue |
| 565 | } |
| 566 | sentinelReachable = true |
| 567 | addrs := parseSlaveAddrs(slaves, useDisconnected) |
| 568 | if len(addrs) == 0 { |
| 569 | continue |
| 570 | } |
| 571 | // Push working sentinel to the top. |
| 572 | c.sentinelAddrs[0], c.sentinelAddrs[i] = c.sentinelAddrs[i], c.sentinelAddrs[0] |
| 573 | c.setSentinel(ctx, sentinel) |
| 574 | |
| 575 | return addrs, nil |
| 576 | } |
| 577 | |
| 578 | if sentinelReachable { |
| 579 | return []string{}, nil |
| 580 | } |
| 581 | return []string{}, errors.New("redis: all sentinels specified in configuration are unreachable") |
| 582 | } |
| 583 | |
| 584 | func (c *sentinelFailover) getMasterAddr(ctx context.Context, sentinel *SentinelClient) string { |
| 585 | addr, err := sentinel.GetMasterAddrByName(ctx, c.opt.MasterName).Result() |
| 586 | if err != nil { |
| 587 | internal.Logger.Printf(ctx, "sentinel: GetMasterAddrByName name=%q failed: %s", |
| 588 | c.opt.MasterName, err) |
| 589 | return "" |
| 590 | } |
| 591 | return net.JoinHostPort(addr[0], addr[1]) |
| 592 | } |
| 593 | |
| 594 | func (c *sentinelFailover) getSlaveAddrs(ctx context.Context, sentinel *SentinelClient) []string { |
| 595 | addrs, err := sentinel.Slaves(ctx, c.opt.MasterName).Result() |
| 596 | if err != nil { |
| 597 | internal.Logger.Printf(ctx, "sentinel: Slaves name=%q failed: %s", |
| 598 | c.opt.MasterName, err) |
| 599 | return []string{} |
| 600 | } |
| 601 | return parseSlaveAddrs(addrs, false) |
| 602 | } |
| 603 | |
| 604 | func parseSlaveAddrs(addrs []interface{}, keepDisconnected bool) []string { |
| 605 | nodes := make([]string, 0, len(addrs)) |
| 606 | for _, node := range addrs { |
| 607 | ip := "" |
| 608 | port := "" |
| 609 | flags := []string{} |
| 610 | lastkey := "" |
| 611 | isDown := false |
| 612 | |
| 613 | for _, key := range node.([]interface{}) { |
| 614 | switch lastkey { |
| 615 | case "ip": |
| 616 | ip = key.(string) |
| 617 | case "port": |
| 618 | port = key.(string) |
| 619 | case "flags": |
| 620 | flags = strings.Split(key.(string), ",") |
| 621 | } |
| 622 | lastkey = key.(string) |
| 623 | } |
| 624 | |
| 625 | for _, flag := range flags { |
| 626 | switch flag { |
| 627 | case "s_down", "o_down": |
| 628 | isDown = true |
| 629 | case "disconnected": |
| 630 | if !keepDisconnected { |
| 631 | isDown = true |
| 632 | } |
| 633 | } |
| 634 | } |
| 635 | |
| 636 | if !isDown { |
| 637 | nodes = append(nodes, net.JoinHostPort(ip, port)) |
| 638 | } |
| 639 | } |
| 640 | |
| 641 | return nodes |
| 642 | } |
| 643 | |
| 644 | func (c *sentinelFailover) trySwitchMaster(ctx context.Context, addr string) { |
| 645 | c.mu.RLock() |
| 646 | currentAddr := c._masterAddr //nolint:ifshort |
| 647 | c.mu.RUnlock() |
| 648 | |
| 649 | if addr == currentAddr { |
| 650 | return |
| 651 | } |
| 652 | |
| 653 | c.mu.Lock() |
| 654 | defer c.mu.Unlock() |
| 655 | |
| 656 | if addr == c._masterAddr { |
| 657 | return |
| 658 | } |
| 659 | c._masterAddr = addr |
| 660 | |
| 661 | internal.Logger.Printf(ctx, "sentinel: new master=%q addr=%q", |
| 662 | c.opt.MasterName, addr) |
| 663 | if c.onFailover != nil { |
| 664 | c.onFailover(ctx, addr) |
| 665 | } |
| 666 | } |
| 667 | |
| 668 | func (c *sentinelFailover) setSentinel(ctx context.Context, sentinel *SentinelClient) { |
| 669 | if c.sentinel != nil { |
| 670 | panic("not reached") |
| 671 | } |
| 672 | c.sentinel = sentinel |
| 673 | c.discoverSentinels(ctx) |
| 674 | |
| 675 | c.pubsub = sentinel.Subscribe(ctx, "+switch-master", "+slave-reconf-done") |
| 676 | go c.listen(c.pubsub) |
| 677 | } |
| 678 | |
| 679 | func (c *sentinelFailover) discoverSentinels(ctx context.Context) { |
| 680 | sentinels, err := c.sentinel.Sentinels(ctx, c.opt.MasterName).Result() |
| 681 | if err != nil { |
| 682 | internal.Logger.Printf(ctx, "sentinel: Sentinels master=%q failed: %s", c.opt.MasterName, err) |
| 683 | return |
| 684 | } |
| 685 | for _, sentinel := range sentinels { |
| 686 | vals := sentinel.([]interface{}) |
| 687 | var ip, port string |
| 688 | for i := 0; i < len(vals); i += 2 { |
| 689 | key := vals[i].(string) |
| 690 | switch key { |
| 691 | case "ip": |
| 692 | ip = vals[i+1].(string) |
| 693 | case "port": |
| 694 | port = vals[i+1].(string) |
| 695 | } |
| 696 | } |
| 697 | if ip != "" && port != "" { |
| 698 | sentinelAddr := net.JoinHostPort(ip, port) |
| 699 | if !contains(c.sentinelAddrs, sentinelAddr) { |
| 700 | internal.Logger.Printf(ctx, "sentinel: discovered new sentinel=%q for master=%q", |
| 701 | sentinelAddr, c.opt.MasterName) |
| 702 | c.sentinelAddrs = append(c.sentinelAddrs, sentinelAddr) |
| 703 | } |
| 704 | } |
| 705 | } |
| 706 | } |
| 707 | |
| 708 | func (c *sentinelFailover) listen(pubsub *PubSub) { |
| 709 | ctx := context.TODO() |
| 710 | |
| 711 | if c.onUpdate != nil { |
| 712 | c.onUpdate(ctx) |
| 713 | } |
| 714 | |
| 715 | ch := pubsub.Channel() |
| 716 | for msg := range ch { |
| 717 | if msg.Channel == "+switch-master" { |
| 718 | parts := strings.Split(msg.Payload, " ") |
| 719 | if parts[0] != c.opt.MasterName { |
| 720 | internal.Logger.Printf(pubsub.getContext(), "sentinel: ignore addr for master=%q", parts[0]) |
| 721 | continue |
| 722 | } |
| 723 | addr := net.JoinHostPort(parts[3], parts[4]) |
| 724 | c.trySwitchMaster(pubsub.getContext(), addr) |
| 725 | } |
| 726 | |
| 727 | if c.onUpdate != nil { |
| 728 | c.onUpdate(ctx) |
| 729 | } |
| 730 | } |
| 731 | } |
| 732 | |
| 733 | func contains(slice []string, str string) bool { |
| 734 | for _, s := range slice { |
| 735 | if s == str { |
| 736 | return true |
| 737 | } |
| 738 | } |
| 739 | return false |
| 740 | } |
| 741 | |
| 742 | //------------------------------------------------------------------------------ |
| 743 | |
| 744 | // NewFailoverClusterClient returns a client that supports routing read-only commands |
| 745 | // to a slave node. |
| 746 | func NewFailoverClusterClient(failoverOpt *FailoverOptions) *ClusterClient { |
| 747 | sentinelAddrs := make([]string, len(failoverOpt.SentinelAddrs)) |
| 748 | copy(sentinelAddrs, failoverOpt.SentinelAddrs) |
| 749 | |
| 750 | failover := &sentinelFailover{ |
| 751 | opt: failoverOpt, |
| 752 | sentinelAddrs: sentinelAddrs, |
| 753 | } |
| 754 | |
| 755 | opt := failoverOpt.clusterOptions() |
| 756 | opt.ClusterSlots = func(ctx context.Context) ([]ClusterSlot, error) { |
| 757 | masterAddr, err := failover.MasterAddr(ctx) |
| 758 | if err != nil { |
| 759 | return nil, err |
| 760 | } |
| 761 | |
| 762 | nodes := []ClusterNode{{ |
| 763 | Addr: masterAddr, |
| 764 | }} |
| 765 | |
| 766 | slaveAddrs, err := failover.slaveAddrs(ctx, false) |
| 767 | if err != nil { |
| 768 | return nil, err |
| 769 | } |
| 770 | |
| 771 | for _, slaveAddr := range slaveAddrs { |
| 772 | nodes = append(nodes, ClusterNode{ |
| 773 | Addr: slaveAddr, |
| 774 | }) |
| 775 | } |
| 776 | |
| 777 | slots := []ClusterSlot{ |
| 778 | { |
| 779 | Start: 0, |
| 780 | End: 16383, |
| 781 | Nodes: nodes, |
| 782 | }, |
| 783 | } |
| 784 | return slots, nil |
| 785 | } |
| 786 | |
| 787 | c := NewClusterClient(opt) |
| 788 | |
| 789 | failover.mu.Lock() |
| 790 | failover.onUpdate = func(ctx context.Context) { |
| 791 | c.ReloadState(ctx) |
| 792 | } |
| 793 | failover.mu.Unlock() |
| 794 | |
| 795 | return c |
| 796 | } |