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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702
package redis

import (
	"context"
	"errors"
	"fmt"
	"math/rand"
	"strconv"
	"sync"
	"sync/atomic"
	"time"

	"github.com/go-redis/redis/internal"
	"github.com/go-redis/redis/internal/consistenthash"
	"github.com/go-redis/redis/internal/hashtag"
	"github.com/go-redis/redis/internal/pool"
)

// Hash is type of hash function used in consistent hash.
type Hash consistenthash.Hash

var errRingShardsDown = errors.New("redis: all ring shards are down")

// RingOptions are used to configure a ring client and should be
// passed to NewRing.
type RingOptions struct {
	// Map of name => host:port addresses of ring shards.
	Addrs map[string]string

	// Frequency of PING commands sent to check shards availability.
	// Shard is considered down after 3 subsequent failed checks.
	HeartbeatFrequency time.Duration

	// Hash function used in consistent hash.
	// Default is crc32.ChecksumIEEE.
	Hash Hash

	// Number of replicas in consistent hash.
	// Default is 100 replicas.
	//
	// Higher number of replicas will provide less deviation, that is keys will be
	// distributed to nodes more evenly.
	//
	// Following is deviation for common nreplicas:
	//  --------------------------------------------------------
	//  | nreplicas | standard error | 99% confidence interval |
	//  |     10    |     0.3152     |      (0.37, 1.98)       |
	//  |    100    |     0.0997     |      (0.76, 1.28)       |
	//  |   1000    |     0.0316     |      (0.92, 1.09)       |
	//  --------------------------------------------------------
	//
	//  See https://arxiv.org/abs/1406.2294 for reference
	HashReplicas int

	// Following options are copied from Options struct.

	OnConnect func(*Conn) error

	DB       int
	Password string

	MaxRetries      int
	MinRetryBackoff time.Duration
	MaxRetryBackoff time.Duration

	DialTimeout  time.Duration
	ReadTimeout  time.Duration
	WriteTimeout time.Duration

	PoolSize           int
	MinIdleConns       int
	MaxConnAge         time.Duration
	PoolTimeout        time.Duration
	IdleTimeout        time.Duration
	IdleCheckFrequency time.Duration
}

func (opt *RingOptions) init() {
	if opt.HeartbeatFrequency == 0 {
		opt.HeartbeatFrequency = 500 * time.Millisecond
	}

	if opt.HashReplicas == 0 {
		opt.HashReplicas = 100
	}

	switch opt.MinRetryBackoff {
	case -1:
		opt.MinRetryBackoff = 0
	case 0:
		opt.MinRetryBackoff = 8 * time.Millisecond
	}
	switch opt.MaxRetryBackoff {
	case -1:
		opt.MaxRetryBackoff = 0
	case 0:
		opt.MaxRetryBackoff = 512 * time.Millisecond
	}
}

func (opt *RingOptions) clientOptions() *Options {
	return &Options{
		OnConnect: opt.OnConnect,

		DB:       opt.DB,
		Password: opt.Password,

		DialTimeout:  opt.DialTimeout,
		ReadTimeout:  opt.ReadTimeout,
		WriteTimeout: opt.WriteTimeout,

		PoolSize:           opt.PoolSize,
		MinIdleConns:       opt.MinIdleConns,
		MaxConnAge:         opt.MaxConnAge,
		PoolTimeout:        opt.PoolTimeout,
		IdleTimeout:        opt.IdleTimeout,
		IdleCheckFrequency: opt.IdleCheckFrequency,
	}
}

//------------------------------------------------------------------------------

type ringShard struct {
	Client *Client
	down   int32
}

func (shard *ringShard) String() string {
	var state string
	if shard.IsUp() {
		state = "up"
	} else {
		state = "down"
	}
	return fmt.Sprintf("%s is %s", shard.Client, state)
}

func (shard *ringShard) IsDown() bool {
	const threshold = 3
	return atomic.LoadInt32(&shard.down) >= threshold
}

func (shard *ringShard) IsUp() bool {
	return !shard.IsDown()
}

// Vote votes to set shard state and returns true if state was changed.
func (shard *ringShard) Vote(up bool) bool {
	if up {
		changed := shard.IsDown()
		atomic.StoreInt32(&shard.down, 0)
		return changed
	}

	if shard.IsDown() {
		return false
	}

	atomic.AddInt32(&shard.down, 1)
	return shard.IsDown()
}

//------------------------------------------------------------------------------

type ringShards struct {
	opt *RingOptions

	mu     sync.RWMutex
	hash   *consistenthash.Map
	shards map[string]*ringShard // read only
	list   []*ringShard          // read only
	len    int
	closed bool
}

func newRingShards(opt *RingOptions) *ringShards {
	return &ringShards{
		opt: opt,

		hash:   newConsistentHash(opt),
		shards: make(map[string]*ringShard),
	}
}

func (c *ringShards) Add(name string, cl *Client) {
	shard := &ringShard{Client: cl}
	c.hash.Add(name)
	c.shards[name] = shard
	c.list = append(c.list, shard)
}

func (c *ringShards) List() []*ringShard {
	c.mu.RLock()
	list := c.list
	c.mu.RUnlock()
	return list
}

func (c *ringShards) Hash(key string) string {
	c.mu.RLock()
	hash := c.hash.Get(key)
	c.mu.RUnlock()
	return hash
}

func (c *ringShards) GetByKey(key string) (*ringShard, error) {
	key = hashtag.Key(key)

	c.mu.RLock()

	if c.closed {
		c.mu.RUnlock()
		return nil, pool.ErrClosed
	}

	hash := c.hash.Get(key)
	if hash == "" {
		c.mu.RUnlock()
		return nil, errRingShardsDown
	}

	shard := c.shards[hash]
	c.mu.RUnlock()

	return shard, nil
}

func (c *ringShards) GetByHash(name string) (*ringShard, error) {
	if name == "" {
		return c.Random()
	}

	c.mu.RLock()
	shard := c.shards[name]
	c.mu.RUnlock()
	return shard, nil
}

func (c *ringShards) Random() (*ringShard, error) {
	return c.GetByKey(strconv.Itoa(rand.Int()))
}

// heartbeat monitors state of each shard in the ring.
func (c *ringShards) Heartbeat(frequency time.Duration) {
	ticker := time.NewTicker(frequency)
	defer ticker.Stop()
	for range ticker.C {
		var rebalance bool

		c.mu.RLock()

		if c.closed {
			c.mu.RUnlock()
			break
		}

		shards := c.list
		c.mu.RUnlock()

		for _, shard := range shards {
			err := shard.Client.Ping().Err()
			if shard.Vote(err == nil || err == pool.ErrPoolTimeout) {
				internal.Logf("ring shard state changed: %s", shard)
				rebalance = true
			}
		}

		if rebalance {
			c.rebalance()
		}
	}
}

// rebalance removes dead shards from the Ring.
func (c *ringShards) rebalance() {
	c.mu.RLock()
	shards := c.shards
	c.mu.RUnlock()

	hash := newConsistentHash(c.opt)
	var shardsNum int
	for name, shard := range shards {
		if shard.IsUp() {
			hash.Add(name)
			shardsNum++
		}
	}

	c.mu.Lock()
	c.hash = hash
	c.len = shardsNum
	c.mu.Unlock()
}

func (c *ringShards) Len() int {
	c.mu.RLock()
	l := c.len
	c.mu.RUnlock()
	return l
}

func (c *ringShards) Close() error {
	c.mu.Lock()
	defer c.mu.Unlock()

	if c.closed {
		return nil
	}
	c.closed = true

	var firstErr error
	for _, shard := range c.shards {
		if err := shard.Client.Close(); err != nil && firstErr == nil {
			firstErr = err
		}
	}
	c.hash = nil
	c.shards = nil
	c.list = nil

	return firstErr
}

//------------------------------------------------------------------------------

// Ring is a Redis client that uses consistent hashing to distribute
// keys across multiple Redis servers (shards). It's safe for
// concurrent use by multiple goroutines.
//
// Ring monitors the state of each shard and removes dead shards from
// the ring. When a shard comes online it is added back to the ring. This
// gives you maximum availability and partition tolerance, but no
// consistency between different shards or even clients. Each client
// uses shards that are available to the client and does not do any
// coordination when shard state is changed.
//
// Ring should be used when you need multiple Redis servers for caching
// and can tolerate losing data when one of the servers dies.
// Otherwise you should use Redis Cluster.
type Ring struct {
	cmdable

	ctx context.Context

	opt           *RingOptions
	shards        *ringShards
	cmdsInfoCache *cmdsInfoCache

	process         func(Cmder) error
	processPipeline func([]Cmder) error
}

func NewRing(opt *RingOptions) *Ring {
	opt.init()

	ring := &Ring{
		opt:    opt,
		shards: newRingShards(opt),
	}
	ring.cmdsInfoCache = newCmdsInfoCache(ring.cmdsInfo)

	ring.process = ring.defaultProcess
	ring.processPipeline = ring.defaultProcessPipeline

	ring.init()

	for name, addr := range opt.Addrs {
		clopt := opt.clientOptions()
		clopt.Addr = addr
		ring.shards.Add(name, NewClient(clopt))
	}

	go ring.shards.Heartbeat(opt.HeartbeatFrequency)

	return ring
}

func (c *Ring) init() {
	c.cmdable.setProcessor(c.Process)
}

func (c *Ring) Context() context.Context {
	if c.ctx != nil {
		return c.ctx
	}
	return context.Background()
}

func (c *Ring) WithContext(ctx context.Context) *Ring {
	if ctx == nil {
		panic("nil context")
	}
	c2 := c.clone()
	c2.ctx = ctx
	return c2
}

func (c *Ring) clone() *Ring {
	cp := *c
	cp.init()

	return &cp
}

// Options returns read-only Options that were used to create the client.
func (c *Ring) Options() *RingOptions {
	return c.opt
}

func (c *Ring) retryBackoff(attempt int) time.Duration {
	return internal.RetryBackoff(attempt, c.opt.MinRetryBackoff, c.opt.MaxRetryBackoff)
}

// PoolStats returns accumulated connection pool stats.
func (c *Ring) PoolStats() *PoolStats {
	shards := c.shards.List()
	var acc PoolStats
	for _, shard := range shards {
		s := shard.Client.connPool.Stats()
		acc.Hits += s.Hits
		acc.Misses += s.Misses
		acc.Timeouts += s.Timeouts
		acc.TotalConns += s.TotalConns
		acc.IdleConns += s.IdleConns
	}
	return &acc
}

// Len returns the current number of shards in the ring.
func (c *Ring) Len() int {
	return c.shards.Len()
}

// Subscribe subscribes the client to the specified channels.
func (c *Ring) Subscribe(channels ...string) *PubSub {
	if len(channels) == 0 {
		panic("at least one channel is required")
	}

	shard, err := c.shards.GetByKey(channels[0])
	if err != nil {
		// TODO: return PubSub with sticky error
		panic(err)
	}
	return shard.Client.Subscribe(channels...)
}

// PSubscribe subscribes the client to the given patterns.
func (c *Ring) PSubscribe(channels ...string) *PubSub {
	if len(channels) == 0 {
		panic("at least one channel is required")
	}

	shard, err := c.shards.GetByKey(channels[0])
	if err != nil {
		// TODO: return PubSub with sticky error
		panic(err)
	}
	return shard.Client.PSubscribe(channels...)
}

// ForEachShard concurrently calls the fn on each live shard in the ring.
// It returns the first error if any.
func (c *Ring) ForEachShard(fn func(client *Client) error) error {
	shards := c.shards.List()
	var wg sync.WaitGroup
	errCh := make(chan error, 1)
	for _, shard := range shards {
		if shard.IsDown() {
			continue
		}

		wg.Add(1)
		go func(shard *ringShard) {
			defer wg.Done()
			err := fn(shard.Client)
			if err != nil {
				select {
				case errCh <- err:
				default:
				}
			}
		}(shard)
	}
	wg.Wait()

	select {
	case err := <-errCh:
		return err
	default:
		return nil
	}
}

func (c *Ring) cmdsInfo() (map[string]*CommandInfo, error) {
	shards := c.shards.List()
	firstErr := errRingShardsDown
	for _, shard := range shards {
		cmdsInfo, err := shard.Client.Command().Result()
		if err == nil {
			return cmdsInfo, nil
		}
		if firstErr == nil {
			firstErr = err
		}
	}
	return nil, firstErr
}

func (c *Ring) cmdInfo(name string) *CommandInfo {
	cmdsInfo, err := c.cmdsInfoCache.Get()
	if err != nil {
		return nil
	}
	info := cmdsInfo[name]
	if info == nil {
		internal.Logf("info for cmd=%s not found", name)
	}
	return info
}

func (c *Ring) cmdShard(cmd Cmder) (*ringShard, error) {
	cmdInfo := c.cmdInfo(cmd.Name())
	pos := cmdFirstKeyPos(cmd, cmdInfo)
	if pos == 0 {
		return c.shards.Random()
	}
	firstKey := cmd.stringArg(pos)
	return c.shards.GetByKey(firstKey)
}

// Do creates a Cmd from the args and processes the cmd.
func (c *Ring) Do(args ...interface{}) *Cmd {
	cmd := NewCmd(args...)
	c.Process(cmd)
	return cmd
}

func (c *Ring) WrapProcess(
	fn func(oldProcess func(cmd Cmder) error) func(cmd Cmder) error,
) {
	c.process = fn(c.process)
}

func (c *Ring) Process(cmd Cmder) error {
	return c.process(cmd)
}

func (c *Ring) defaultProcess(cmd Cmder) error {
	for attempt := 0; attempt <= c.opt.MaxRetries; attempt++ {
		if attempt > 0 {
			time.Sleep(c.retryBackoff(attempt))
		}

		shard, err := c.cmdShard(cmd)
		if err != nil {
			cmd.setErr(err)
			return err
		}

		err = shard.Client.Process(cmd)
		if err == nil {
			return nil
		}
		if !internal.IsRetryableError(err, cmd.readTimeout() == nil) {
			return err
		}
	}
	return cmd.Err()
}

func (c *Ring) Pipeline() Pipeliner {
	pipe := Pipeline{
		exec: c.processPipeline,
	}
	pipe.cmdable.setProcessor(pipe.Process)
	return &pipe
}

func (c *Ring) Pipelined(fn func(Pipeliner) error) ([]Cmder, error) {
	return c.Pipeline().Pipelined(fn)
}

func (c *Ring) WrapProcessPipeline(
	fn func(oldProcess func([]Cmder) error) func([]Cmder) error,
) {
	c.processPipeline = fn(c.processPipeline)
}

func (c *Ring) defaultProcessPipeline(cmds []Cmder) error {
	cmdsMap := make(map[string][]Cmder)
	for _, cmd := range cmds {
		cmdInfo := c.cmdInfo(cmd.Name())
		hash := cmd.stringArg(cmdFirstKeyPos(cmd, cmdInfo))
		if hash != "" {
			hash = c.shards.Hash(hashtag.Key(hash))
		}
		cmdsMap[hash] = append(cmdsMap[hash], cmd)
	}

	for attempt := 0; attempt <= c.opt.MaxRetries; attempt++ {
		if attempt > 0 {
			time.Sleep(c.retryBackoff(attempt))
		}

		var mu sync.Mutex
		var failedCmdsMap map[string][]Cmder
		var wg sync.WaitGroup

		for hash, cmds := range cmdsMap {
			wg.Add(1)
			go func(hash string, cmds []Cmder) {
				defer wg.Done()

				shard, err := c.shards.GetByHash(hash)
				if err != nil {
					setCmdsErr(cmds, err)
					return
				}

				cn, err := shard.Client.getConn()
				if err != nil {
					setCmdsErr(cmds, err)
					return
				}

				canRetry, err := shard.Client.pipelineProcessCmds(cn, cmds)
				shard.Client.releaseConnStrict(cn, err)

				if canRetry && internal.IsRetryableError(err, true) {
					mu.Lock()
					if failedCmdsMap == nil {
						failedCmdsMap = make(map[string][]Cmder)
					}
					failedCmdsMap[hash] = cmds
					mu.Unlock()
				}
			}(hash, cmds)
		}

		wg.Wait()
		if len(failedCmdsMap) == 0 {
			break
		}
		cmdsMap = failedCmdsMap
	}

	return cmdsFirstErr(cmds)
}

func (c *Ring) TxPipeline() Pipeliner {
	panic("not implemented")
}

func (c *Ring) TxPipelined(fn func(Pipeliner) error) ([]Cmder, error) {
	panic("not implemented")
}

// Close closes the ring client, releasing any open resources.
//
// It is rare to Close a Ring, as the Ring is meant to be long-lived
// and shared between many goroutines.
func (c *Ring) Close() error {
	return c.shards.Close()
}

func (c *Ring) Watch(fn func(*Tx) error, keys ...string) error {
	if len(keys) == 0 {
		return fmt.Errorf("redis: Watch requires at least one key")
	}

	var shards []*ringShard
	for _, key := range keys {
		if key != "" {
			shard, err := c.shards.GetByKey(hashtag.Key(key))
			if err != nil {
				return err
			}

			shards = append(shards, shard)
		}
	}

	if len(shards) == 0 {
		return fmt.Errorf("redis: Watch requires at least one shard")
	}

	if len(shards) > 1 {
		for _, shard := range shards[1:] {
			if shard.Client != shards[0].Client {
				err := fmt.Errorf("redis: Watch requires all keys to be in the same shard")
				return err
			}
		}
	}

	return shards[0].Client.Watch(fn, keys...)
}

func newConsistentHash(opt *RingOptions) *consistenthash.Map {
	return consistenthash.New(opt.HashReplicas, consistenthash.Hash(opt.Hash))
}