Golang实现请求限流的几种办法

目录

简单的并发控制

使用计数器实现请求限流

使用golang官方包实现httpserver频率限制

使用Token Bucket(令牌桶算法)实现请求限流


简单的并发控制

利用 channel 的缓冲设定,我们就可以来实现并发的限制。我们只要在执行并发的同时,往一个带有缓冲的 channel 里写入点东西(随便写啥,内容不重要)。让并发的 goroutine在执行完成后把这个 channel 里的东西给读走。这样整个并发的数量就讲控制在这个 channel的缓冲区大小上。

比如我们可以用一个 bool 类型的带缓冲 channel 作为并发限制的计数器。

chLimit := make(chan bool, 1)

 然后在并发执行的地方,每创建一个新的 goroutine,都往 chLimit 里塞个东西。

  1.   for i, sleeptime := range input {
  2.   chs[i] = make(chan string, 1)
  3.   chLimit <- true
  4.   go limitFunc(chLimit, chs[i], i, sleeptime, timeout)
  5.   }

这里通过 go 关键字并发执行的是新构造的函数。他在执行完后,会把 chLimit的缓冲区里给消费掉一个。

  1.   limitFunc := func(chLimit chan bool, ch chan string, task_id, sleeptime, timeout int) {
  2.   Run(task_id, sleeptime, timeout, ch)
  3.   <-chLimit
  4.   }

这样一来,当创建的 goroutine 数量到达 chLimit 的缓冲区上限后。主 goroutine 就挂起阻塞了,直到这些 goroutine 执行完毕,消费掉了 chLimit 缓冲区中的数据,程序才会继续创建新的 goroutine 。我们并发数量限制的目的也就达到了。

以下是完整代码:

  1.   package main
  2.    
  3.   import (
  4.   "fmt"
  5.   "time"
  6.   )
  7.    
  8.   func Run(task_id, sleeptime, timeout int, ch chan string) {
  9.   ch_run := make(chan string)
  10.   go run(task_id, sleeptime, ch_run)
  11.   select {
  12.   case re := <-ch_run:
  13.   ch <- re
  14.   case <-time.After(time.Duration(timeout) * time.Second):
  15.   re := fmt.Sprintf("task id %d , timeout", task_id)
  16.   ch <- re
  17.   }
  18.   }
  19.    
  20.   func run(task_id, sleeptime int, ch chan string) {
  21.    
  22.   time.Sleep(time.Duration(sleeptime) * time.Second)
  23.   ch <- fmt.Sprintf("task id %d , sleep %d second", task_id, sleeptime)
  24.   return
  25.   }
  26.    
  27.   func main() {
  28.   input := []int{3, 2, 1}
  29.   timeout := 2
  30.   chLimit := make(chan bool, 1)
  31.   chs := make([]chan string, len(input))
  32.   limitFunc := func(chLimit chan bool, ch chan string, task_id, sleeptime, timeout int) {
  33.   Run(task_id, sleeptime, timeout, ch)
  34.   <-chLimit
  35.   }
  36.   startTime := time.Now()
  37.   fmt.Println("Multirun start")
  38.   for i, sleeptime := range input {
  39.   chs[i] = make(chan string, 1)
  40.   chLimit <- true
  41.   go limitFunc(chLimit, chs[i], i, sleeptime, timeout)
  42.   }
  43.    
  44.   for _, ch := range chs {
  45.   fmt.Println(<-ch)
  46.   }
  47.   endTime := time.Now()
  48.   fmt.Printf("Multissh finished. Process time %s. Number of task is %d", endTime.Sub(startTime), len(input))
  49.   }

运行结果:

  1.   Multirun start
  2.   task id 0 , timeout
  3.   task id 1 , timeout
  4.   task id 2 , sleep 1 second
  5.   Multissh finished. Process time 5s. Number of task is 3

如果修改并发限制为2:

chLimit := make(chan bool, 2)

运行结果:

  1.    
  2.   Multirun start
  3.   task id 0 , timeout
  4.   task id 1 , timeout
  5.   task id 2 , sleep 1 second
  6.   Multissh finished. Process time 3s. Number of task is 3

使用计数器实现请求限流

限流的要求是在指定的时间间隔内,server 最多只能服务指定数量的请求。实现的原理是我们启动一个计数器,每次服务请求会把计数器加一,同时到达指定的时间间隔后会把计数器清零;这个计数器的实现代码如下所示:

  1.   type RequestLimitService struct {
  2.   Interval time.Duration
  3.   MaxCount int
  4.   Lock sync.Mutex
  5.   ReqCount int
  6.   }
  7.    
  8.   func NewRequestLimitService(interval time.Duration, maxCnt int) *RequestLimitService {
  9.   reqLimit := &RequestLimitService{
  10.   Interval: interval,
  11.   MaxCount: maxCnt,
  12.   }
  13.    
  14.   go func() {
  15.   ticker := time.NewTicker(interval)
  16.   for {
  17.   <-ticker.C
  18.   reqLimit.Lock.Lock()
  19.   fmt.Println("Reset Count...")
  20.   reqLimit.ReqCount = 0
  21.   reqLimit.Lock.Unlock()
  22.   }
  23.   }()
  24.    
  25.   return reqLimit
  26.   }
  27.    
  28.   func (reqLimit *RequestLimitService) Increase() {
  29.   reqLimit.Lock.Lock()
  30.   defer reqLimit.Lock.Unlock()
  31.    
  32.   reqLimit.ReqCount += 1
  33.   }
  34.    
  35.   func (reqLimit *RequestLimitService) IsAvailable() bool {
  36.   reqLimit.Lock.Lock()
  37.   defer reqLimit.Lock.Unlock()
  38.    
  39.   return reqLimit.ReqCount < reqLimit.MaxCount
  40.   }

在服务请求的时候, 我们会对当前计数器和阈值进行比较,只有未超过阈值时才进行服务:

  1.   var RequestLimit = NewRequestLimitService(10 * time.Second, 5)
  2.    
  3.   func helloHandler(w http.ResponseWriter, r *http.Request) {
  4.   if RequestLimit.IsAvailable() {
  5.   RequestLimit.Increase()
  6.   fmt.Println(RequestLimit.ReqCount)
  7.   io.WriteString(w, "Hello world!\n")
  8.   } else {
  9.   fmt.Println("Reach request limiting!")
  10.   io.WriteString(w, "Reach request limit!\n")
  11.   }
  12.   }
  13.    
  14.   func main() {
  15.   fmt.Println("Server Started!")
  16.   http.HandleFunc("/", helloHandler)
  17.   http.ListenAndServe(":8000", nil)
  18.   }

 完整代码url:https://github.com/hiberabyss/JustDoIt/blob/master/RequestLimit/request_limit.go

使用golang官方包实现httpserver频率限制

使用golang来编写httpserver时,可以使用官方已经有实现好的包:

  1.   import(
  2.   "fmt"
  3.   "net"
  4.   "golang.org/x/net/netutil"
  5.   )
  6.    
  7.   func main() {
  8.   l, err := net.Listen("tcp", "127.0.0.1:0")
  9.   if err != nil {
  10.   fmt.Fatalf("Listen: %v", err)
  11.   }
  12.   defer l.Close()
  13.   l = LimitListener(l, max)
  14.    
  15.   http.Serve(l, http.HandlerFunc())
  16.    
  17.   //bla bla bla.................
  18.   }

源码如下(url : https://github.com/golang/net/blob/master/netutil/listen.go),基本思路就是为连接数计数,通过make chan来建立一个最大连接数的channel, 每次accept就+1,close时候就-1. 当到达最大连接数时,就等待空闲连接出来之后再accept。

  1.   // Copyright 2013 The Go Authors. All rights reserved.
  2.   // Use of this source code is governed by a BSD-style
  3.   // license that can be found in the LICENSE file.
  4.    
  5.   // Package netutil provides network utility functions, complementing the more
  6.   // common ones in the net package.
  7.   package netutil // import "golang.org/x/net/netutil"
  8.    
  9.   import (
  10.   "net"
  11.   "sync"
  12.   )
  13.    
  14.   // LimitListener returns a Listener that accepts at most n simultaneous
  15.   // connections from the provided Listener.
  16.   func LimitListener(l net.Listener, n int) net.Listener {
  17.   return &limitListener{
  18.   Listener: l,
  19.   sem: make(chan struct{}, n),
  20.   done: make(chan struct{}),
  21.   }
  22.   }
  23.    
  24.   type limitListener struct {
  25.   net.Listener
  26.   sem chan struct{}
  27.   closeOnce sync.Once // ensures the done chan is only closed once
  28.   done chan struct{} // no values sent; closed when Close is called
  29.   }
  30.    
  31.   // acquire acquires the limiting semaphore. Returns true if successfully
  32.   // accquired, false if the listener is closed and the semaphore is not
  33.   // acquired.
  34.   func (l *limitListener) acquire() bool {
  35.   select {
  36.   case <-l.done:
  37.   return false
  38.   case l.sem <- struct{}{}:
  39.   return true
  40.   }
  41.   }
  42.   func (l *limitListener) release() { <-l.sem }
  43.    
  44.   func (l *limitListener) Accept() (net.Conn, error) {
  45.   //如果sem满了,就会阻塞在这
  46.   acquired := l.acquire()
  47.   // If the semaphore isn't acquired because the listener was closed, expect
  48.   // that this call to accept won't block, but immediately return an error.
  49.   c, err := l.Listener.Accept()
  50.   if err != nil {
  51.   if acquired {
  52.   l.release()
  53.   }
  54.   return nil, err
  55.   }
  56.   return &limitListenerConn{Conn: c, release: l.release}, nil
  57.   }
  58.    
  59.   func (l *limitListener) Close() error {
  60.   err := l.Listener.Close()
  61.   l.closeOnce.Do(func() { close(l.done) })
  62.   return err
  63.   }
  64.    
  65.   type limitListenerConn struct {
  66.   net.Conn
  67.   releaseOnce sync.Once
  68.   release func()
  69.   }
  70.    
  71.   func (l *limitListenerConn) Close() error {
  72.   err := l.Conn.Close()
  73.   //close时释放占用的sem
  74.   l.releaseOnce.Do(l.release)
  75.   return err
  76.   }

使用Token Bucket(令牌桶算法)实现请求限流

在开发高并发系统时有三把利器用来保护系统:缓存、降级和限流!为了保证在业务高峰期,线上系统也能保证一定的弹性和稳定性,最有效的方案就是进行服务降级了,而限流就是降级系统最常采用的方案之一。

这里为大家推荐一个开源库https://github.com/didip/tollbooth,但是,如果您想要一些简单的、轻量级的或者只是想要学习的东西,实现自己的中间件来处理速率限制并不困难。今天我们就来聊聊如何实现自己的一个限流中间件

首先我们需要安装一个提供了 Token bucket (令牌桶算法)的依赖包,上面提到的toolbooth 的实现也是基于它实现的:

$ go get golang.org/x/time/rate

先看Demo代码的实现:

  1.   package main
  2.    
  3.   import (
  4.   "net/http"
  5.   "golang.org/x/time/rate"
  6.   )
  7.    
  8.   var limiter = rate.NewLimiter(2, 5)
  9.   func limit(next http.Handler) http.Handler {
  10.   return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
  11.   if limiter.Allow() == false {
  12.   http.Error(w, http.StatusText(429), http.StatusTooManyRequests)
  13.   return
  14.   }
  15.   next.ServeHTTP(w, r)
  16.   })
  17.   }
  18.    
  19.   func main() {
  20.   mux := http.NewServeMux()
  21.   mux.HandleFunc("/", okHandler)
  22.   // Wrap the servemux with the limit middleware.
  23.   http.ListenAndServe(":4000", limit(mux))
  24.   }
  25.    
  26.   func okHandler(w http.ResponseWriter, r *http.Request) {
  27.   w.Write([]byte("OK"))
  28.   }

然后看看 rate.NewLimiter的源码:

算法描述:用户配置的平均发送速率为r,则每隔1/r秒一个令牌被加入到桶中(每秒会有r个令牌放入桶中),桶中最多可以存放b个令牌。如果令牌到达时令牌桶已经满了,那么这个令牌会被丢弃;

  1.   // Copyright 2015 The Go Authors. All rights reserved.
  2.   // Use of this source code is governed by a BSD-style
  3.   // license that can be found in the LICENSE file.
  4.   // Package rate provides a rate limiter.
  5.   package rate
  6.    
  7.   import (
  8.   "fmt"
  9.   "math"
  10.   "sync"
  11.   "time"
  12.    
  13.   "golang.org/x/net/context"
  14.   )
  15.    
  16.   // Limit defines the maximum frequency of some events.
  17.   // Limit is represented as number of events per second.
  18.   // A zero Limit allows no events.
  19.   type Limit float64
  20.    
  21.   // Inf is the infinite rate limit; it allows all events (even if burst is zero).
  22.   const Inf = Limit(math.MaxFloat64)
  23.    
  24.   // Every converts a minimum time interval between events to a Limit.
  25.   func Every(interval time.Duration) Limit {
  26.   if interval <= 0 {
  27.   return Inf
  28.   }
  29.   return 1 / Limit(interval.Seconds())
  30.   }
  31.    
  32.   // A Limiter controls how frequently events are allowed to happen.
  33.   // It implements a "token bucket" of size b, initially full and refilled
  34.   // at rate r tokens per second.
  35.   // Informally, in any large enough time interval, the Limiter limits the
  36.   // rate to r tokens per second, with a maximum burst size of b events.
  37.   // As a special case, if r == Inf (the infinite rate), b is ignored.
  38.   // See https://en.wikipedia.org/wiki/Token_bucket for more about token buckets.
  39.   //
  40.   // The zero value is a valid Limiter, but it will reject all events.
  41.   // Use NewLimiter to create non-zero Limiters.
  42.   //
  43.   // Limiter has three main methods, Allow, Reserve, and Wait.
  44.   // Most callers should use Wait.
  45.   //
  46.   // Each of the three methods consumes a single token.
  47.   // They differ in their behavior when no token is available.
  48.   // If no token is available, Allow returns false.
  49.   // If no token is available, Reserve returns a reservation for a future token
  50.   // and the amount of time the caller must wait before using it.
  51.   // If no token is available, Wait blocks until one can be obtained
  52.   // or its associated context.Context is canceled.
  53.   //
  54.   // The methods AllowN, ReserveN, and WaitN consume n tokens.
  55.   type Limiter struct {
  56.   //maximum token, token num per second
  57.   limit Limit
  58.   //burst field, max token num
  59.   burst int
  60.   mu sync.Mutex
  61.   //tokens num, change
  62.   tokens float64
  63.   // last is the last time the limiter's tokens field was updated
  64.   last time.Time
  65.   // lastEvent is the latest time of a rate-limited event (past or future)
  66.   lastEvent time.Time
  67.   }
  68.    
  69.   // Limit returns the maximum overall event rate.
  70.   func (lim *Limiter) Limit() Limit {
  71.   lim.mu.Lock()
  72.   defer lim.mu.Unlock()
  73.   return lim.limit
  74.   }
  75.    
  76.   // Burst returns the maximum burst size. Burst is the maximum number of tokens
  77.   // that can be consumed in a single call to Allow, Reserve, or Wait, so higher
  78.   // Burst values allow more events to happen at once.
  79.   // A zero Burst allows no events, unless limit == Inf.
  80.   func (lim *Limiter) Burst() int {
  81.   return lim.burst
  82.   }
  83.    
  84.   // NewLimiter returns a new Limiter that allows events up to rate r and permits
  85.   // bursts of at most b tokens.
  86.   func NewLimiter(r Limit, b int) *Limiter {
  87.   return &Limiter{
  88.   limit: r,
  89.   burst: b,
  90.   }
  91.   }
  92.    
  93.   // Allow is shorthand for AllowN(time.Now(), 1).
  94.   func (lim *Limiter) Allow() bool {
  95.   return lim.AllowN(time.Now(), 1)
  96.   }
  97.    
  98.   // AllowN reports whether n events may happen at time now.
  99.   // Use this method if you intend to drop / skip events that exceed the rate limit.
  100.   // Otherwise use Reserve or Wait.
  101.   func (lim *Limiter) AllowN(now time.Time, n int) bool {
  102.   return lim.reserveN(now, n, 0).ok
  103.   }
  104.    
  105.   // A Reservation holds information about events that are permitted by a Limiter to happen after a delay.
  106.   // A Reservation may be canceled, which may enable the Limiter to permit additional events.
  107.   type Reservation struct {
  108.   ok bool
  109.   lim *Limiter
  110.   tokens int
  111.   //This is the time to action
  112.   timeToAct time.Time
  113.   // This is the Limit at reservation time, it can change later.
  114.   limit Limit
  115.   }
  116.    
  117.   // OK returns whether the limiter can provide the requested number of tokens
  118.   // within the maximum wait time. If OK is false, Delay returns InfDuration, and
  119.   // Cancel does nothing.
  120.   func (r *Reservation) OK() bool {
  121.   return r.ok
  122.   }
  123.    
  124.   // Delay is shorthand for DelayFrom(time.Now()).
  125.   func (r *Reservation) Delay() time.Duration {
  126.   return r.DelayFrom(time.Now())
  127.   }
  128.    
  129.   // InfDuration is the duration returned by Delay when a Reservation is not OK.
  130.   const InfDuration = time.Duration(1<<63 - 1)
  131.    
  132.   // DelayFrom returns the duration for which the reservation holder must wait
  133.   // before taking the reserved action. Zero duration means act immediately.
  134.   // InfDuration means the limiter cannot grant the tokens requested in this
  135.   // Reservation within the maximum wait time.
  136.   func (r *Reservation) DelayFrom(now time.Time) time.Duration {
  137.   if !r.ok {
  138.   return InfDuration
  139.   }
  140.   delay := r.timeToAct.Sub(now)
  141.   if delay < 0 {
  142.   return 0
  143.   }
  144.   return delay
  145.   }
  146.    
  147.   // Cancel is shorthand for CancelAt(time.Now()).
  148.   func (r *Reservation) Cancel() {
  149.   r.CancelAt(time.Now())
  150.   return
  151.   }
  152.    
  153.   // CancelAt indicates that the reservation holder will not perform the reserved action
  154.   // and reverses the effects of this Reservation on the rate limit as much as possible,
  155.   // considering that other reservations may have already been made.
  156.   func (r *Reservation) CancelAt(now time.Time) {
  157.   if !r.ok {
  158.   return
  159.   }
  160.   r.lim.mu.Lock()
  161.   defer r.lim.mu.Unlock()
  162.   if r.lim.limit == Inf || r.tokens == 0 || r.timeToAct.Before(now) {
  163.   return
  164.   }
  165.   // calculate tokens to restore
  166.   // The duration between lim.lastEvent and r.timeToAct tells us how many tokens were reserved
  167.   // after r was obtained. These tokens should not be restored.
  168.   restoreTokens := float64(r.tokens) - r.limit.tokensFromDuration(r.lim.lastEvent.Sub(r.timeToAct))
  169.   if restoreTokens <= 0 {
  170.   return
  171.   }
  172.   // advance time to now
  173.   now, _, tokens := r.lim.advance(now)
  174.   // calculate new number of tokens
  175.   tokens += restoreTokens
  176.   if burst := float64(r.lim.burst); tokens > burst {
  177.   tokens = burst
  178.   }
  179.   // update state
  180.   r.lim.last = now
  181.   r.lim.tokens = tokens
  182.   if r.timeToAct == r.lim.lastEvent {
  183.   prevEvent := r.timeToAct.Add(r.limit.durationFromTokens(float64(-r.tokens)))
  184.   if !prevEvent.Before(now) {
  185.   r.lim.lastEvent = prevEvent
  186.   }
  187.   }
  188.   return
  189.   }
  190.    
  191.   // Reserve is shorthand for ReserveN(time.Now(), 1).
  192.   func (lim *Limiter) Reserve() *Reservation {
  193.   return lim.ReserveN(time.Now(), 1)
  194.   }
  195.    
  196.   // ReserveN returns a Reservation that indicates how long the caller must wait before n events happen.
  197.   // The Limiter takes this Reservation into account when allowing future events.
  198.   // ReserveN returns false if n exceeds the Limiter's burst size.
  199.   // Usage example:
  200.   // r, ok := lim.ReserveN(time.Now(), 1)
  201.   // if !ok {
  202.   // // Not allowed to act! Did you remember to set lim.burst to be > 0 ?
  203.   // }
  204.   // time.Sleep(r.Delay())
  205.   // Act()
  206.   // Use this method if you wish to wait and slow down in accordance with the rate limit without dropping events.
  207.   // If you need to respect a deadline or cancel the delay, use Wait instead.
  208.   // To drop or skip events exceeding rate limit, use Allow instead.
  209.   func (lim *Limiter) ReserveN(now time.Time, n int) *Reservation {
  210.   r := lim.reserveN(now, n, InfDuration)
  211.   return &r
  212.   }
  213.    
  214.   // Wait is shorthand for WaitN(ctx, 1).
  215.   func (lim *Limiter) Wait(ctx context.Context) (err error) {
  216.   return lim.WaitN(ctx, 1)
  217.   }
  218.    
  219.   // WaitN blocks until lim permits n events to happen.
  220.   // It returns an error if n exceeds the Limiter's burst size, the Context is
  221.   // canceled, or the expected wait time exceeds the Context's Deadline.
  222.   func (lim *Limiter) WaitN(ctx context.Context, n int) (err error) {
  223.   if n > lim.burst {
  224.   return fmt.Errorf("rate: Wait(n=%d) exceeds limiter's burst %d", n, lim.burst)
  225.   }
  226.   // Check if ctx is already cancelled
  227.   select {
  228.   case <-ctx.Done():
  229.   return ctx.Err()
  230.   default:
  231.   }
  232.   // Determine wait limit
  233.   now := time.Now()
  234.   waitLimit := InfDuration
  235.   if deadline, ok := ctx.Deadline(); ok {
  236.   waitLimit = deadline.Sub(now)
  237.   }
  238.   // Reserve
  239.   r := lim.reserveN(now, n, waitLimit)
  240.   if !r.ok {
  241.   return fmt.Errorf("rate: Wait(n=%d) would exceed context deadline", n)
  242.   }
  243.   // Wait
  244.   t := time.NewTimer(r.DelayFrom(now))
  245.   defer t.Stop()
  246.   select {
  247.   case <-t.C:
  248.   // We can proceed.
  249.   return nil
  250.   case <-ctx.Done():
  251.   // Context was canceled before we could proceed. Cancel the
  252.   // reservation, which may permit other events to proceed sooner.
  253.   r.Cancel()
  254.   return ctx.Err()
  255.   }
  256.   }
  257.    
  258.   // SetLimit is shorthand for SetLimitAt(time.Now(), newLimit).
  259.   func (lim *Limiter) SetLimit(newLimit Limit) {
  260.   lim.SetLimitAt(time.Now(), newLimit)
  261.   }
  262.    
  263.   // SetLimitAt sets a new Limit for the limiter. The new Limit, and Burst, may be violated
  264.   // or underutilized by those which reserved (using Reserve or Wait) but did not yet act
  265.   // before SetLimitAt was called.
  266.   func (lim *Limiter) SetLimitAt(now time.Time, newLimit Limit) {
  267.   lim.mu.Lock()
  268.   defer lim.mu.Unlock()
  269.   now, _, tokens := lim.advance(now)
  270.   lim.last = now
  271.   lim.tokens = tokens
  272.   lim.limit = newLimit
  273.   }
  274.    
  275.   // reserveN is a helper method for AllowN, ReserveN, and WaitN.
  276.   // maxFutureReserve specifies the maximum reservation wait duration allowed.
  277.   // reserveN returns Reservation, not *Reservation, to avoid allocation in AllowN and WaitN.
  278.   func (lim *Limiter) reserveN(now time.Time, n int, maxFutureReserve time.Duration) Reservation {
  279.   lim.mu.Lock()
  280.   defer lim.mu.Unlock()
  281.   if lim.limit == Inf {
  282.   return Reservation{
  283.   ok: true,
  284.   lim: lim,
  285.   tokens: n,
  286.   timeToAct: now,
  287.   }
  288.   }
  289.   now, last, tokens := lim.advance(now)
  290.   // Calculate the remaining number of tokens resulting from the request.
  291.   tokens -= float64(n)
  292.   // Calculate the wait duration
  293.   var waitDuration time.Duration
  294.   if tokens < 0 {
  295.   waitDuration = lim.limit.durationFromTokens(-tokens)
  296.   }
  297.   // Decide result
  298.   ok := n <= lim.burst && waitDuration <= maxFutureReserve
  299.   // Prepare reservation
  300.   r := Reservation{
  301.   ok: ok,
  302.   lim: lim,
  303.   limit: lim.limit,
  304.   }
  305.   if ok {
  306.   r.tokens = n
  307.   r.timeToAct = now.Add(waitDuration)
  308.   }
  309.   // Update state
  310.   if ok {
  311.   lim.last = now
  312.   lim.tokens = tokens
  313.   lim.lastEvent = r.timeToAct
  314.   } else {
  315.   lim.last = last
  316.   }
  317.   return r
  318.   }
  319.    
  320.   // advance calculates and returns an updated state for lim resulting from the passage of time.
  321.   // lim is not changed.
  322.   func (lim *Limiter) advance(now time.Time) (newNow time.Time, newLast time.Time, newTokens float64) {
  323.   last := lim.last
  324.   if now.Before(last) {
  325.   last = now
  326.   }
  327.   // Avoid making delta overflow below when last is very old.
  328.   maxElapsed := lim.limit.durationFromTokens(float64(lim.burst) - lim.tokens)
  329.   elapsed := now.Sub(last)
  330.   if elapsed > maxElapsed {
  331.   elapsed = maxElapsed
  332.   }
  333.   // Calculate the new number of tokens, due to time that passed.
  334.   delta := lim.limit.tokensFromDuration(elapsed)
  335.   tokens := lim.tokens + delta
  336.   if burst := float64(lim.burst); tokens > burst {
  337.   tokens = burst
  338.   }
  339.   return now, last, tokens
  340.   }
  341.    
  342.   // durationFromTokens is a unit conversion function from the number of tokens to the duration
  343.   // of time it takes to accumulate them at a rate of limit tokens per second.
  344.   func (limit Limit) durationFromTokens(tokens float64) time.Duration {
  345.   seconds := tokens / float64(limit)
  346.   return time.Nanosecond * time.Duration(1e9*seconds)
  347.   }
  348.    
  349.   // tokensFromDuration is a unit conversion function from a time duration to the number of tokens
  350.   // which could be accumulated during that duration at a rate of limit tokens per second.
  351.   func (limit Limit) tokensFromDuration(d time.Duration) float64 {
  352.   return d.Seconds() * float64(limit)
  353.   }

虽然在某些情况下使用单个全局速率限制器非常有用,但另一种常见情况是基于IP地址或API密钥等标识符为每个用户实施速率限制器。我们将使用IP地址作为标识符。简单实现代码如下:

  1.   package main
  2.   import (
  3.   "net/http"
  4.   "sync"
  5.   "time"
  6.   "golang.org/x/time/rate"
  7.   )
  8.   // Create a custom visitor struct which holds the rate limiter for each
  9.   // visitor and the last time that the visitor was seen.
  10.   type visitor struct {
  11.   limiter *rate.Limiter
  12.   lastSeen time.Time
  13.   }
  14.   // Change the the map to hold values of the type visitor.
  15.   var visitors = make(map[string]*visitor)
  16.   var mtx sync.Mutex
  17.   // Run a background goroutine to remove old entries from the visitors map.
  18.   func init() {
  19.   go cleanupVisitors()
  20.   }
  21.   func addVisitor(ip string) *rate.Limiter {
  22.   limiter := rate.NewLimiter(2, 5)
  23.   mtx.Lock()
  24.   // Include the current time when creating a new visitor.
  25.   visitors[ip] = &visitor{limiter, time.Now()}
  26.   mtx.Unlock()
  27.   return limiter
  28.   }
  29.   func getVisitor(ip string) *rate.Limiter {
  30.   mtx.Lock()
  31.   v, exists := visitors[ip]
  32.   if !exists {
  33.   mtx.Unlock()
  34.   return addVisitor(ip)
  35.   }
  36.   // Update the last seen time for the visitor.
  37.   v.lastSeen = time.Now()
  38.   mtx.Unlock()
  39.   return v.limiter
  40.   }
  41.   // Every minute check the map for visitors that haven't been seen for
  42.   // more than 3 minutes and delete the entries.
  43.   func cleanupVisitors() {
  44.   for {
  45.   time.Sleep(time.Minute)
  46.   mtx.Lock()
  47.   for ip, v := range visitors {
  48.   if time.Now().Sub(v.lastSeen) > 3*time.Minute {
  49.   delete(visitors, ip)
  50.   }
  51.   }
  52.   mtx.Unlock()
  53.   }
  54.   }
  55.   func limit(next http.Handler) http.Handler {
  56.   return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
  57.   limiter := getVisitor(r.RemoteAddr)
  58.   if limiter.Allow() == false {
  59.   http.Error(w, http.StatusText(429), http.StatusTooManyRequests)
  60.   return
  61.   }
  62.   next.ServeHTTP(w, r)
  63.   })
  64.   }

 

转载: https://blog.csdn.net/micl200110041/article/details/82013032

上一篇:Cauchy收敛准则


下一篇:Linux下安装SVN服务端