PHP和zookeeper结合实践 - sfornt - 博客园
首页
新闻
博问
专区
闪存
班级
我的博客
我的园子
账号设置
简洁模式 ...
退出登录
注册
登录
sfornt
首页
订阅
管理
PHP和zookeeper结合实践
Zookeeper
简单介绍
Apache Zookeeper是开发和维护开源服务器的服务，它能够实现高度可靠的分布式协调。
安装Zookeeper(无需安装)
wget http://mirror.bit.edu.cn/apache/zookeeper/zookeeper-3.4.10/zookeeper-3.4.10.tar.gz
tar zxvf zookeeper-3.4.10.tar.gz
安装Zookeeper C扩展支持
cd zookeeper-3.4.8/src/c
./configure --prefix=/usr/
make
make install
安装php的zookeeper扩展
wget https://pecl.php.net/get/zookeeper-0.6.2.tgz
tar zxvf zookeeper-0.6.2.tgz
cd zookeeper-0.6.2
phpize
./configure --with-libzookeeper-dir=/usr/
$ make
$ make install
启动zookeeper server
# {zookeeperserver}代表你zookeeper解压的目录
mkdir -p /project/zookeeper/demo/data
#zoo.conf 是默认启动所加载的配置文件
cp /{zookeeperserver}/conf/zoo_sample.cfg /{zookeeperserver}/conf/zoo.cfg
/{zookeeperserver}/bin/zkServer start
zookeeper client 测试
/{zookeeperserver}/bin/zkCli.sh -server 127.0.0.1:2181
# [zk: 127.0.0.1:2181(CONNECTED) x] 客户端连接信息shell
[zk: 127.0.0.1:2181(CONNECTED) 3] ls /
[zookeeper]
[zk: 127.0.0.1:2181(CONNECTED) 4] create /test qkl001
Created /test
[zk: 127.0.0.1:2181(CONNECTED) 5] create /zgq zgq002
Created /zgq
[zk: 127.0.0.1:2181(CONNECTED) 6] ls /
[zookeeper, test, zgq]
[zk: 127.0.0.1:2181(CONNECTED) 5] delete /zgq
Created /zgq
[zk: 127.0.0.1:2181(CONNECTED) 7] quit
PHP下实践
zookeeper类信息
Zookeeper — Zookeeper类
-- Zookeeper::addAuth — 指定应用程序凭据
-- Zookeeper::connect — 创建与Zookeeper沟通的句柄
-- Zookeeper::__construct — 创建与Zookeeper沟通的句柄
-- Zookeeper::create — 同步创建节点
-- Zookeeper::delete — 同步删除Zookeeper中的一个节点
-- Zookeeper::exists — 同步检查Zookeeper节点的存在性
-- Zookeeper::get — 同步获取与节点关联的数据。
-- Zookeeper::getAcl — 同步地获取与节点关联的ACL。
-- Zookeeper::getChildren — 同步列出节点的子节点
-- Zookeeper::getClientId — 返回客户端会话ID，仅在连接当前连接时才有效（即最后观察者状态为ZooOnCeleDelphi状态）
-- Zookeeper::getRecvTimeout — 返回此会话的超时，如果连接当前连接（只有上次观察者状态为ZooOnCeleTytStand状态）才有效。此值可能在服务器重新连接后发生更改。
-- Zookeeper::getState — 获取Zookeeper连接的状态
-- Zookeeper::isRecoverable — 检查当前的Zookeeper连接状态是否可以恢复
-- Zookeeper::set — 设置与节点关联的数据
-- Zookeeper::setAcl — 同步设置与节点关联的ACL
-- Zookeeper::setDebugLevel — 设置库的调试级别
-- Zookeeper::setDeterministicConnOrder — 启用/禁用仲裁端点顺序随机化
-- Zookeeper::setLogStream — 设置库用于日志记录的流
-- Zookeeper::setWatcher — 设置观察函数
客户端操作
zkCli.cmd
# output 表示连接成功
WatchedEvent state:SyncConnected type:None path:null
[zk: localhost:2181(CONNECTED) 0]
# 以下是操作命令
ls /
create /ztest 1
获取节点信息(Zookeeper::get)
# get.php
$zoo = new Zookeeper('192.168.1.45:2181');
$r = $zoo->get( '/ztest');
var_dump($r);
获取节点信息(Zookeeper::get)并设置watcher
# get.php
class ZookeeperDemo extends Zookeeper
public function watcher($type, $state, $key)
var_dump($type);
var_dump($state);
var_dump($key);
if ($type == 3) {
var_dump($this->get('/zgetw'));
// Watcher gets consumed so we need to set a new one
$this->get('/zgetw', array($this, 'watcher'));
$zoo = new ZookeeperDemo('192.168.1.45:2181');
$zoo->get('/zgetw', [$zoo, 'watcher']);
while (true) {
echo '.';
sleep(1);
客户端操作
set /ztest 2
set /ztest 3
set /ztest 5
获取节点信息(Zookeeper::get) 结果
php get.php
#output
......int(3)
int(3)
string(6) "/ztest"
string(1) "2"
..int(3)
int(3)
string(6) "/ztest"
string(1) "3"
..int(3)
int(3)
string(6) "/ztest"
string(1) "4"
.int(3)
int(3)
string(6) "/ztest"
string(1) "5"
创建子节点
create /ztest/child1 c1
# output: Created /ztest/child1
create /ztest/child2 c2
# output: Created /ztest/child2
ls /ztest
# output: [child2, child1]
php下获取子节点
# getChild.php
$zookeeper = new Zookeeper('192.168.1.45:2181');
$path = '/ztest';
$r = $zookeeper->getchildren($path);
if (!empty($r)) {
foreach ($r as $c)
var_dump($c);
# php getChild.php
string(6) "child2"
string(6) "child1"
创建顺序节点
[zk: localhost:2181(CONNECTED) 22] create /stest 1
Created /stest
[zk: localhost:2181(CONNECTED) 23] create -s /stest/seq 1
Created /stest/seq0000000000
[zk: localhost:2181(CONNECTED) 24] create -s /stest/seq 1
Created /stest/seq0000000001
[zk: localhost:2181(CONNECTED) 25] create -s /stest/seq 1
Created /stest/seq0000000002
[zk: localhost:2181(CONNECTED) 26] create -s /stest/seq 1
Created /stest/seq0000000003
[zk: localhost:2181(CONNECTED) 27] create -s /stest/seq 1
Created /stest/seq0000000004
[zk: localhost:2181(CONNECTED) 28] create -s /stest/seq 1
Created /stest/seq0000000005
[zk: localhost:2181(CONNECTED) 29] create -s /stest/seq 1
Created /stest/seq0000000006
[zk: localhost:2181(CONNECTED) 30] create -s /stest/seq 1
Created /stest/seq0000000007
[zk: localhost:2181(CONNECTED) 31] create -s /stest/seq 2
Created /stest/seq0000000008
[zk: localhost:2181(CONNECTED) 32] create -s /stest/seq 2
Created /stest/seq0000000009
[zk: localhost:2181(CONNECTED) 33] create -s /stest/seq 2
Created /stest/seq0000000010
[zk: localhost:2181(CONNECTED) 34] create -s /stest/seq 2
Created /stest/seq0000000011
[zk: localhost:2181(CONNECTED) 35] create -s /stest/seq 2
Created /stest/seq0000000012
[zk: localhost:2181(CONNECTED) 36] create -s /stest/seq 2
Created /stest/seq0000000013
[zk: localhost:2181(CONNECTED) 37] create -s /stest/seq 2
Created /stest/seq0000000014
代码
$zookeeper = new Zookeeper('192.168.1.45:2181');
$aclArray = array(
array(
'perms' => Zookeeper::PERM_ALL,
'scheme' => 'world',
'id' => 'anyone',
);
$path = '/t3';
//ZOO_EPHEMERAL = 1
//ZOO_SEQUENCE = 2
//这里这里的flag=NULL,flag=0 表示创建永久节点，=1创建临时节点，=2创建seq 顺序节点
$realPath = $zookeeper->create($path, null, $aclArray, 1);
var_dump($realPath);
worker
一个利用顺序临时节点的leader迁移实现
# worker.php
<?php
/**
* Created by PhpStorm.
* User: qkl
* Date: 2018/8/27
* Time: 14:25
*/
class Worker
const CONTAINER = '/worker_test';
protected $acl = [
'perms' => Zookeeper::PERM_ALL,
'scheme' => 'world',
'id' => 'anyone'
];
private $isLeader = false;
private $znode = '';
private $prevNode = '';
public function __construct($host = '', $watcher_cb = null, $recv_timeout = 10000)
$this->zk = new Zookeeper($host, $watcher_cb, $recv_timeout);
public function register()
if (!$this->zk->exists(self::CONTAINER)) {
$this->zk->create(self::CONTAINER, null, $this->acl);
# 创建一个临时的顺序节点
$this->znode = $this->zk->create(self::CONTAINER . '/w-',
null,
$this->acl,
Zookeeper::EPHEMERAL | Zookeeper::SEQUENCE);
//获取顺序节点名，去除父路径
$this->znode = str_replace(self::CONTAINER . '/', '', $this->znode);
printf("我创建了节点: %s\n", self::CONTAINER . '/' . $this->znode);
$this->prevNode = $this->getPrev();
if (is_null($this->prevNode)) {
$this->setLeader(true);
} else {
$this->zk->get(self::CONTAINER . "/" . $this->prevNode, [$this, 'watcher']);
public function getPrev()
$workers = $this->zk->getChildren(self::CONTAINER);
sort($workers);
$size = count($workers);
for ($i = 0; $i < $size; $i++) {
if ($this->znode == $workers[$i] && $i > 0) {
return $workers[$i - 1];
return null;
public function watcher($type, $state, $key)
$prevNode = $this->prevNode;
printf("watchNode-getPrevious：%s\n", self::CONTAINER . "/" . $prevNode);
if (!is_null($prevNode)) {
if ($type == 2) {
printf($prevNode . " had deleted\n");
if ($type == 3) {
printf("我重新监控节点：%s\n", self::CONTAINER . "/" . $prevNode);
$this->zk->get(self::CONTAINER . "/" . $prevNode, [$this, 'watcher']);
public function isLeader()
return $this->isLeader;
public function setLeader($flag)
$this->isLeader = $flag;
public function run()
$this->register();
while (true) {
if ($this->isLeader()) {
$this->leaderJob();
} else {
$this->watcherJob();
sleep(2);
public function leaderJob()
echo "现在我是Leader\n";
public function watcherJob()
echo "我在监控：".(self::CONTAINER . "/" . $this->prevNode)."\n";
$worker = new Worker('192.168.1.45:2181');
$worker->run();
尝试
# 终端1
php worker.php
# 终端2
php worker.php
# 终端3
php worker.php
# 此处运行不会被节点变化不会被监控到
再次尝试
# zkCli
create /worker_test/w-
/worker_test/0000000020
php worker.php
# output
我创建了节点: /worker_test/w-0000000022
我在监控：/worker_test/w-0000000020
我在监控：/worker_test/w-0000000020
watchNode-getPrevious：/worker_test/w-0000000020
我重新监控节点：/worker_test/w-0000000020
我在监控：/worker_test/w-0000000020
我在监控：/worker_test/w-0000000020
我在监控：/worker_test/w-0000000020
我在监控：/worker_test/w-0000000020
我在监控：/worker_test/w-0000000020
watchNode-getPrevious：/worker_test/w-0000000020
我重新监控节点：/worker_test/w-0000000020
2018-08-28 02:11:46,684:2486(0x7f28ed0a1700):ZOO_WARN@zookeeper_interest@1570: Exceeded deadline by 15ms
我在监控：/worker_test/w-0000000020
我在监控：/worker_test/w-0000000020
watchNode-getPrevious：/worker_test/w-0000000020
w-0000000020 had deleted
我在监控：/worker_test/w-0000000020
这边发现一个类似bug问题
如果我们直接运行worker.php
在worker.php运行创建的临时顺序节点是不会被watcher到的
我们必须先首先创建好相关的节点再启动监控，不知道这里是不是php版本的zookeeper的bug
有了解的小伙伴可以告之下
Watcher通知状态与事件类型一览
ZOO_CREATED_EVENT(value=1)：节点创建事件，需要watch一个不存在的节点，当节点被创建时触发，此watch通过zoo_exists()设置ZOO_DELETED_EVENT(value=2)：节点删除事件，此watch通过zoo_exists()或zoo_get()设置ZOO_CHANGED_EVENT(value=3)：节点数据改变事件，此watch通过zoo_exists()或zoo_get()设置ZOO_CHILD_EVENT(value=4)：子节点列表改变事件，此watch通过zoo_get_children()或zoo_get_children2()设置ZOO_SESSION_EVENT(value=-1)：会话事件，客户端与服务端断开或重连时触发ZOO_NOTWATCHING_EVENT(value=-2)：watch移除事件，服务端出于某些原因不再为客户端watch节点时触发
原文地址：https://segmentfault.com/a/1190000016173878
posted @
2018-11-17 13:43
sfornt
阅读(2193)
评论(0)
编辑
收藏
举报
刷新评论刷新页面返回顶部
Copyright 2022 sfornt
Powered by .NET 7.0 on Kubernetes