首页 > 其他好文 > 详细

分布式系统生成唯一主键

时间：2017-06-10 13:58:21 阅读：367 评论：0 收藏：0 [点我收藏+]

标签：keyword blog process width stream pack lib int moved

原创文章，转载请注明出处：server非业余研究http://blog.csdn.net/erlib 作者Sunface

唯一ID在系统小的时候，随便整都可以。可是系统大了后呢？这个时候假设出现了问题，ID重构就是非常大的困难，因此对于不论什么一个系统，从最初架构时就进行好唯一ID设计是非常重要的。twitter的snowflake就攻克了这种需求。实现也还是非常easy的，除去配置信息，核心代码就是毫秒级时间41位+机器ID 10位+毫秒内序列12位。

该项目地址为：https://github.com/twitter/snowflake是用Scala实现的。

python版详见开源项目https://github.com/erans/pysnowflake。

核心代码为其IdWorker这个类实现，其原理结构例如以下，我分别用一个0表示一位。用—切割开部分的作用：

10---0000000000 0000000000 0000000000 0000000000 0 --- 00000 ---00000 ---000000000000

在上面的字符串中，第一位为未使用（实际上也可作为long的符号位），接下来的41位为毫秒级时间，然后5位datacenter标识位，5位机器ID（并不算标识符，实际是为线程标识），然后12位该毫秒内的当前毫秒内的计数。加起来刚好64位，为一个Long型。

这种优点是，总体上依照时间自增排序，而且整个分布式系统内不会产生ID碰撞（由datacenter和机器ID作区分）。而且效率较高。经測试，snowflake每秒可以产生26万ID左右。全然满足须要。

且看其核心代码：

1</pre>
2/** Copyright 2010-2012 Twitter, Inc.*/
3package com.twitter.service.snowflake
4 
5import com.twitter.ostrich.stats.Stats
6import com.twitter.service.snowflake.gen._
7import java.util.Random
8import com.twitter.logging.Logger
9 
10/**
11 * An object that generates IDs.
12 * This is broken into a separate class in case
13 * we ever want to support multiple worker threads
14 * per process
15 */
16class IdWorker(val workerId: Long, val datacenterId: Long, private val reporter: Reporter, var sequence: Long = 0L)
17extends Snowflake.Iface {
18  private[this] def genCounter(agent: String) = {
19    Stats.incr("ids_generated")
20    Stats.incr("ids_generated_%s".format(agent))
21  }
22  private[this] val exceptionCounter = Stats.getCounter("exceptions")
23  private[this] val log = Logger.get
24  private[this] val rand = new Random
25 
26  val twepoch = 1288834974657L
27 
28 //机器标识位数
29 
30  private[this] val workerIdBits = 5L
31 
32//数据中心标识位数
33  private[this] val datacenterIdBits = 5L
34 
35//机器ID最大值
36  private[this] val maxWorkerId = -1L ^ (-1L << workerIdBits)
37 
38//数据中心ID最大值
39  private[this] val maxDatacenterId = -1L ^ (-1L << datacenterIdBits)
40 
41//毫秒内自增位
42  private[this] val sequenceBits = 12L
43 
44//机器ID偏左移12位
45 
46  private[this] val workerIdShift = sequenceBits
47 
48//数据中心ID左移17位
49  private[this] val datacenterIdShift = sequenceBits + workerIdBits
50 
51//时间毫秒左移22位
52  private[this] val timestampLeftShift = sequenceBits + workerIdBits + datacenterIdBits
53  private[this] val sequenceMask = -1L ^ (-1L << sequenceBits)
54 
55  private[this] var lastTimestamp = -1L
56 
57  // sanity check for workerId
58  if (workerId > maxWorkerId || workerId < 0) {
59    exceptionCounter.incr(1)
60    throw new IllegalArgumentException("worker Id can‘t be greater than %d or less than 0".format(maxWorkerId))
61  }
62 
63  if (datacenterId > maxDatacenterId || datacenterId < 0) {
64    exceptionCounter.incr(1)
65    throw new IllegalArgumentException("datacenter Id can‘t be greater than %d or less than 0".format(maxDatacenterId))
66  }
67 
68  log.info("worker starting. timestamp left shift %d, datacenter id bits %d, worker id bits %d, sequence bits %d, workerid %d",
69    timestampLeftShift, datacenterIdBits, workerIdBits, sequenceBits, workerId)
70 
71  def get_id(useragent: String): Long = {
72    if (!validUseragent(useragent)) {
73      exceptionCounter.incr(1)
74      throw new InvalidUserAgentError
75    }
76 
77    val id = nextId()
78    genCounter(useragent)
79 
80    reporter.report(new AuditLogEntry(id, useragent, rand.nextLong))
81    id
82  }
83 
84  def get_worker_id(): Long = workerId
85  def get_datacenter_id(): Long = datacenterId
86  def get_timestamp() = System.currentTimeMillis
87 
88  protected[snowflake] def nextId(): Long = synchronized {
89    var timestamp = timeGen()
90 
91 //时间错误
92 
93    if (timestamp < lastTimestamp) {
94      exceptionCounter.incr(1)
95      log.error("clock is moving backwards.  Rejecting requests until %d.", lastTimestamp);
96      throw new InvalidSystemClock("Clock moved backwards.  Refusing to generate id for %d milliseconds".format(
97        lastTimestamp - timestamp))
98    }
99 
100    if (lastTimestamp == timestamp) {
101//当前毫秒内，则+1
102      sequence = (sequence + 1) & sequenceMask
103      if (sequence == 0) {
104//当前毫秒内计数满了，则等待下一秒
105        timestamp = tilNextMillis(lastTimestamp)
106      }
107    } else {
108      sequence = 0
109    }
110 
111    lastTimestamp = timestamp
112//ID偏移组合生成终于的ID。并返回ID   
113 
114((timestamp - twepoch) << timestampLeftShift) |
115      (datacenterId << datacenterIdShift) |
116      (workerId << workerIdShift) |
117      sequence
118  }
119 
120//等待下一个毫秒的到来 
121 
122protected def tilNextMillis(lastTimestamp: Long): Long = {
123    var timestamp = timeGen()
124    while (timestamp <= lastTimestamp) {
125      timestamp = timeGen()
126    }
127    timestamp
128  }
129 
130  protected def timeGen(): Long = System.currentTimeMillis()
131 
132  val AgentParser = """([a-zA-Z][a-zA-Z\-0-9]*)""".r
133 
134  def validUseragent(useragent: String): Boolean = useragent match {
135    case AgentParser(_) => true
136    case _ => false
137  }
138}
139<pre>

上述为twitter的实现，以下且看一个Java实现，貌似为淘宝的朋友写的。

1public class IdWorker {
2 private final long workerId;
3 private final static long twepoch = 1361753741828L;
4 private long sequence = 0L;
5 private final static long workerIdBits = 4L;
6 public final static long maxWorkerId = -1L ^ -1L << workerIdBits;
7 private final static long sequenceBits = 10L;
8 
9 private final static long workerIdShift = sequenceBits;
10 private final static long timestampLeftShift = sequenceBits + workerIdBits;
11 public final static long sequenceMask = -1L ^ -1L << sequenceBits;
12 
13 private long lastTimestamp = -1L;
14 
15 public IdWorker(final long workerId) {
16  super();
17  if (workerId > this.maxWorkerId || workerId < 0) {
18   throw new IllegalArgumentException(String.format(
19     "worker Id can‘t be greater than %d or less than 0",
20     this.maxWorkerId));
21  }
22  this.workerId = workerId;
23 }
24 
25 public synchronized long nextId() {
26  long timestamp = this.timeGen();
27  if (this.lastTimestamp == timestamp) {
28   this.sequence = (this.sequence + 1) & this.sequenceMask;
29   if (this.sequence == 0) {
30    System.out.println("###########" + sequenceMask);
31    timestamp = this.tilNextMillis(this.lastTimestamp);
32   }
33  } else {
34   this.sequence = 0;
35  }
36  if (timestamp < this.lastTimestamp) {
37   try {
38    throw new Exception(
39      String.format(
40        "Clock moved backwards.  Refusing to generate id for %d milliseconds",
41        this.lastTimestamp - timestamp));
42   } catch (Exception e) {
43    e.printStackTrace();
44   }
45  }
46 
47  this.lastTimestamp = timestamp;
48  long nextId = ((timestamp - twepoch << timestampLeftShift))
49    | (this.workerId << this.workerIdShift) | (this.sequence);
50//  System.out.println("timestamp:" + timestamp + ",timestampLeftShift:"
51//    + timestampLeftShift + ",nextId:" + nextId + ",workerId:"
52//    + workerId + ",sequence:" + sequence);
53  return nextId;
54 }
55 
56 private long tilNextMillis(final long lastTimestamp) {
57  long timestamp = this.timeGen();
58  while (timestamp <= lastTimestamp) {
59   timestamp = this.timeGen();
60  }
61  return timestamp;
62 }
63 
64 private long timeGen() {
65  return System.currentTimeMillis();
66 }
67  
68  
69 public static void main(String[] args){
70  IdWorker worker2 = new IdWorker(2);
71  System.out.println(worker2.nextId());
72 
73   
74 }
75 
76}

再来看一个php的实现

1<?php
2class Idwork
3{
4const debug = 1;
5static $workerId;
6static $twepoch = 1361775855078;
7static $sequence = 0;
8const workerIdBits = 4;
9static $maxWorkerId = 15;
10const sequenceBits = 10;
11static $workerIdShift = 10;
12static $timestampLeftShift = 14;
13static $sequenceMask = 1023;
14private  static $lastTimestamp = -1;
15 
16function __construct($workId){
17if($workId > self::$maxWorkerId || $workId< 0 )
18{
19throw new Exception("worker Id can‘t be greater than 15 or less than 0");
20}
21self::$workerId=$workId;
22 
23echo ‘logdebug->__construct()->self::$workerId:‘.self::$workerId;
24echo ‘</br>‘;
25 
26}
27 
28function timeGen(){
29//获得当前时间戳
30$time = explode(‘ ‘, microtime());
31$time2= substr($time[0], 2, 3);
32$timestramp = $time[1].$time2;
33echo ‘logdebug->timeGen()->$timestramp:‘.$time[1].$time2;
34echo ‘</br>‘;
35return  $time[1].$time2;
36}
37function  tilNextMillis($lastTimestamp) {
38$timestamp = $this->timeGen();
39while ($timestamp <= $lastTimestamp) {
40$timestamp = $this->timeGen();
41}
42 
43echo ‘logdebug->tilNextMillis()->$timestamp:‘.$timestamp;
44echo ‘</br>‘;
45return $timestamp;
46}
47 
48function  nextId()
49{
50$timestamp=$this->timeGen();
51echo ‘logdebug->nextId()->self::$lastTimestamp1:‘.self::$lastTimestamp;
52echo ‘</br>‘;
53if(self::$lastTimestamp == $timestamp) {
54self::$sequence = (self::$sequence + 1) & self::$sequenceMask;
55if (self::$sequence == 0) {
56    echo "###########".self::$sequenceMask;
57    $timestamp = $this->tilNextMillis(self::$lastTimestamp);
58    echo ‘logdebug->nextId()->self::$lastTimestamp2:‘.self::$lastTimestamp;
59    echo ‘</br>‘;
60  }
61} else {
62self::$sequence  = 0;
63    echo ‘logdebug->nextId()->self::$sequence:‘.self::$sequence;
64    echo ‘</br>‘;
65}
66if ($timestamp < self::$lastTimestamp) {
67   throw new Excwption("Clock moved backwards.  Refusing to generate id for ".(self::$lastTimestamp-$timestamp)." milliseconds");
68   }
69self::$lastTimestamp  = $timestamp;
70echo ‘logdebug->nextId()->self::$lastTimestamp3:‘.self::$lastTimestamp;
71echo ‘</br>‘;
72 
73echo ‘logdebug->nextId()->(($timestamp - self::$twepoch << self::$timestampLeftShift )):‘.((sprintf(‘%.0f‘, $timestamp) - sprintf(‘%.0f‘, self::$twepoch) ));
74echo ‘</br>‘;
75$nextId = ((sprintf(‘%.0f‘, $timestamp) - sprintf(‘%.0f‘, self::$twepoch)  )) | ( self::$workerId << self::$workerIdShift ) | self::$sequence;
76echo ‘timestamp:‘.$timestamp.‘-----‘;
77echo ‘twepoch:‘.sprintf(‘%.0f‘, self::$twepoch).‘-----‘;
78echo ‘timestampLeftShift =‘.self::$timestampLeftShift.‘-----‘;
79echo ‘nextId:‘.$nextId.‘----‘;
80echo ‘workId:‘.self::$workerId.‘-----‘;
81echo ‘workerIdShift:‘.self::$workerIdShift.‘-----‘;
82return $nextId;
83}
84 
85}
86$Idwork = new Idwork(1);
87$a= $Idwork->nextId();
88$Idwork = new Idwork(2);
89$a= $Idwork->nextId();
90?>

分布式系统生成唯一主键

标签：keyword blog process width stream pack lib int moved

原文地址：http://www.cnblogs.com/brucemengbm/p/6978296.html

踩

(0)

评论一句话评论（0）

分享档案

更多>

2021年07月29日 (22)
2021年07月28日 (40)
2021年07月27日 (32)
2021年07月26日 (79)
2021年07月23日 (29)
2021年07月22日 (30)
2021年07月21日 (42)
2021年07月20日 (16)
2021年07月19日 (90)
2021年07月16日 (35)

周排行

友情链接

兰亭集智国之画百度统计站长统计阿里云 chrome插件新版天听网

关于我们 - 联系我们 - 留言反馈

迷上了代码！