基于 ZooKeeper 的分布式锁和队列( 上 )

Posted ImportNew

tags:

篇首语:本文由小常识网(cha138.com)小编为大家整理,主要介绍了基于 ZooKeeper 的分布式锁和队列( 上 )相关的知识,希望对你有一定的参考价值。


来源:阿凡卢,

www.cnblogs.com/luxiaoxun/p/4889764.html


在分布式系统中,往往需要一些分布式同步原语来做一些协同工作,上一篇文章介绍了Zookeeper的基本原理,本文介绍下基于Zookeeper的Lock和Queue的实现,主要代码都来自Zookeeper的官方recipe。


锁(Lock)


完全分布式锁是全局同步的,这意味着在任何时刻没有两个客户端会同时认为它们都拥有相同的锁,使用 Zookeeper 可以实现分布式锁,需要首先定义一个锁节点(lock root node)。


需要获得锁的客户端按照以下步骤来获取锁:


  1. 保证锁节点(lock root node)这个父根节点的存在,这个节点是每个要获取lock客户端共用的,这个节点是PERSISTENT的。

  2. 第一次需要创建本客户端要获取lock的节点,调用 create( ),并设置 节点为EPHEMERAL_SEQUENTIAL类型,表示该节点为临时的和顺序的。如果获取锁的节点挂掉,则该节点自动失效,可以让其他节点获取锁。

  3. 在父锁节点(lock root node)上调用 getChildren( ) ,不需要设置监视标志。 (为了避免“羊群效应”).

  4. 按照Fair竞争的原则,将步骤3中的子节点(要获取锁的节点)按照节点顺序的大小做排序,取出编号最小的一个节点做为lock的owner,判断自己的节点id

  5. 是否就为owner id,如果是则返回,lock成功。如果不是则调用 exists( )监听比自己小的前一位的id,关注它锁释放的操作(也就是exist watch)。

  6. 如果第4步监听exist的watch被触发,则继续按4中的原则判断自己是否能获取到lock。


释放锁:需要释放锁的客户端只需要删除在第2步中创建的节点即可。


注意事项:


一个节点的删除只会导致一个客户端被唤醒,因为每个节点只被一个客户端watch,这避免了“羊群效应”。


一个分布式lock的实现:


package org.apache.zookeeper.recipes.lock;

 

import org.slf4j.Logger;

import org.slf4j.LoggerFactory;

import org.apache.zookeeper.KeeperException;

import org.apache.zookeeper.WatchedEvent;

import org.apache.zookeeper.Watcher;

import static org.apache.zookeeper.CreateMode.EPHEMERAL_SEQUENTIAL;

import org.apache.zookeeper.ZooKeeper;

import org.apache.zookeeper.data.ACL;

import org.apache.zookeeper.data.Stat;

 

import java.util.List;

import java.util.SortedSet;

import java.util.TreeSet;

 

/**

 * A <a href="package.html">protocol to implement an exclusive

 *  write lock or to elect a leader</a>. <p/> You invoke {@link #lock()} to 

 *  start the process of grabbing the lock; you may get the lock then or it may be 

 *  some time later. <p/> You can register a listener so that you are invoked 

 *  when you get the lock; otherwise you can ask if you have the lock

 *  by calling {@link #isOwner()}

 *

 */

public class WriteLock extends ProtocolSupport {

    private static final Logger LOG = LoggerFactory.getLogger(WriteLock.class);

 

    private final String dir;

    private String id;

    private ZNodeName idName;

    private String ownerId;

    private String lastChildId;

    private byte[] data = {0x12, 0x34};

    private LockListener callback;

    private LockZooKeeperOperation zop;

     

    /**

     * zookeeper contructor for writelock

     * @param zookeeper zookeeper client instance

     * @param dir the parent path you want to use for locking

     * @param acls the acls that you want to use for all the paths, 

     * if null world read/write is used.

     */

    public WriteLock(ZooKeeper zookeeper, String dir, List<ACL> acl) {

        super(zookeeper);

        this.dir = dir;

        if (acl != null) {

            setAcl(acl);

        }

        this.zop = new LockZooKeeperOperation();

    }

     

    /**

     * zookeeper contructor for writelock with callback

     * @param zookeeper the zookeeper client instance

     * @param dir the parent path you want to use for locking

     * @param acl the acls that you want to use for all the paths

     * @param callback the call back instance

     */

    public WriteLock(ZooKeeper zookeeper, String dir, List<ACL> acl, 

            LockListener callback) {

        this(zookeeper, dir, acl);

        this.callback = callback;

    }

 

    /**

     * return the current locklistener

     * @return the locklistener

     */

    public LockListener getLockListener() {

        return this.callback;

    }

     

    /**

     * register a different call back listener

     * @param callback the call back instance

     */

    public void setLockListener(LockListener callback) {

        this.callback = callback;

    }

 

    /**

     * Removes the lock or associated znode if 

     * you no longer require the lock. this also 

     * removes your request in the queue for locking

     * in case you do not already hold the lock.

     * @throws RuntimeException throws a runtime exception

     * if it cannot connect to zookeeper.

     */

    public synchronized void unlock() throws RuntimeException {

         

        if (!isClosed() && id != null) {

            // we don't need to retry this operation in the case of failure

            // as ZK will remove ephemeral files and we don't wanna hang

            // this process when closing if we cannot reconnect to ZK

            try {

                ZooKeeperOperation zopdel = new ZooKeeperOperation() {

                    public boolean execute() throws KeeperException,

                        InterruptedException {

                        zookeeper.delete(id, -1);   

                        return Boolean.TRUE;

                    }

                };

                zopdel.execute();

            } catch (InterruptedException e) {

                LOG.warn("Caught: " + e, e);

                //set that we have been interrupted.

               Thread.currentThread().interrupt();

            } catch (KeeperException.NoNodeException e) {

                // do nothing

            } catch (KeeperException e) {

                LOG.warn("Caught: " + e, e);

                throw (RuntimeException) new RuntimeException(e.getMessage()).

                    initCause(e);

            }

            finally {

                if (callback != null) {

                    callback.lockReleased();

                }

                id = null;

            }

        }

    }

     

    /** 

     * the watcher called on  

     * getting watch while watching 

     * my predecessor

     */

    private class LockWatcher implements Watcher {

        public void process(WatchedEvent event) {

            // lets either become the leader or watch the new/updated node

            LOG.debug("Watcher fired on path: " + event.getPath() + " state: " + 

                    event.getState() + " type " + event.getType());

            try {

                lock();

            } catch (Exception e) {

                LOG.warn("Failed to acquire lock: " + e, e);

            }

        }

    }

     

    /**

     * a zoookeeper operation that is mainly responsible

     * for all the magic required for locking.

     */

    private  class LockZooKeeperOperation implements ZooKeeperOperation {

         

        /** find if we have been created earler if not create our node

         * 

         * @param prefix the prefix node

         * @param zookeeper teh zookeeper client

         * @param dir the dir paretn

         * @throws KeeperException

         * @throws InterruptedException

         */

        private void findPrefixInChildren(String prefix, ZooKeeper zookeeper, String dir) 

            throws KeeperException, InterruptedException {

            List<String> names = zookeeper.getChildren(dir, false);

            for (String name : names) {

                if (name.startsWith(prefix)) {

                    id = dir + "/" + name;

                    if (LOG.isDebugEnabled()) {

                        LOG.debug("Found id created last time: " + id);

                    }

                    break;

                }

            }

            if (id == null) {

                id = zookeeper.create(dir + "/" + prefix, data, 

                        getAcl(), EPHEMERAL_SEQUENTIAL);

 

                if (LOG.isDebugEnabled()) {

                    LOG.debug("Created id: " + id);

                }

            }

 

        }

         

        /**

         * the command that is run and retried for actually 

         * obtaining the lock

         * @return if the command was successful or not

         */

        public boolean execute() throws KeeperException, InterruptedException {

            do {

                if (id == null) {

                    long sessionId = zookeeper.getSessionId();

                    String prefix = "x-" + sessionId + "-";

                    // lets try look up the current ID if we failed 

                    // in the middle of creating the znode

                    findPrefixInChildren(prefix, zookeeper, dir);

                    idName = new ZNodeName(id);

                }

                if (id != null) {

                    List<String> names = zookeeper.getChildren(dir, false);

                    if (names.isEmpty()) {

                        LOG.warn("No children in: " + dir + " when we've just " +

                        "created one! Lets recreate it...");

                        // lets force the recreation of the id

                        id = null;

                    } else {

                        // lets sort them explicitly (though they do seem to come back in order ususally :)

                        SortedSet<ZNodeName> sortedNames = new TreeSet<ZNodeName>();

                        for (String name : names) {

                            sortedNames.add(new ZNodeName(dir + "/" + name));

                        }

                        ownerId = sortedNames.first().getName();

                        SortedSet<ZNodeName> lessThanMe = sortedNames.headSet(idName);

                        if (!lessThanMe.isEmpty()) {

                            ZNodeName lastChildName = lessThanMe.last();

                            lastChildId = lastChildName.getName();

                            if (LOG.isDebugEnabled()) {

                                LOG.debug("watching less than me node: " + lastChildId);

                            }

                            Stat stat = zookeeper.exists(lastChildId, new LockWatcher());

                            if (stat != null) {

                                return Boolean.FALSE;

                            } else {

                                LOG.warn("Could not find the" +

                                        " stats for less than me: " + lastChildName.getName());

                            }

                        } else {

                            if (isOwner()) {

                                if (callback != null) {

                                    callback.lockAcquired();

                                }

                                return Boolean.TRUE;

                            }

                        }

                    }

                }

            }

            while (id == null);

            return Boolean.FALSE;

        }

    };

 

    /**

     * Attempts to acquire the exclusive write lock returning whether or not it was

     * acquired. Note that the exclusive lock may be acquired some time later after

     * this method has been invoked due to the current lock owner going away.

     */

    public synchronized boolean lock() throws KeeperException, InterruptedException {

        if (isClosed()) {

            return false;

        }

        ensurePathExists(dir);

 

        return (Boolean) retryOperation(zop);

    }

 

    /**

     * return the parent dir for lock

     * @return the parent dir used for locks.

     */

    public String getDir() {

        return dir;

    }

 

    /**

     * Returns true if this node is the owner of the

     *  lock (or the leader)

     */

    public boolean isOwner() {

        return id != null && ownerId != null && id.equals(ownerId);

    }

 

    /**

     * return the id for this lock

     * @return the id for this lock

     */

    public String getId() {

       return this.id;

    }

}


注意这里的lock,可能会失败,会尝试多次,每次失败后会Sleep一段时间。


PS:官方的代码有个小bug,id和ownerId应该都是全路径,即id = dir + “/” + name;原代码在findPrefixInChildren里有问题。


用于辅助节点大小顺序排序的类:


package org.apache.zookeeper.recipes.lock;

 

import org.slf4j.Logger;

import org.slf4j.LoggerFactory;

 

/**

 * Represents an ephemeral znode name which has an ordered sequence number and

 * can be sorted in order

 * 

 */

class ZNodeName implements Comparable<ZNodeName> {

    private final String name;

    private String prefix;

    private int sequence = -1;

    private static final Logger LOG = LoggerFactory.getLogger(ZNodeName.class);

 

    public ZNodeName(String name) {

        if (name == null) {

            throw new NullPointerException("id cannot be null");

        }

        this.name = name;

        this.prefix = name;

        int idx = name.lastIndexOf('-');

        if (idx >= 0) {

            this.prefix = name.substring(0, idx);

            try {

                this.sequence = Integer.parseInt(name.substring(idx + 1));

                // If an exception occurred we misdetected a sequence suffix,

                // so return -1.

            } catch (NumberFormatException e) {

                LOG.info("Number format exception for " + idx, e);

            } catch (ArrayIndexOutOfBoundsException e) {

                LOG.info("Array out of bounds for " + idx, e);

            }

        }

    }

 

    @Override

    public String toString() {

        return name.toString();

    }

 

    @Override

    public boolean equals(Object o) {

        if (this == o)

            return true;

        if (o == null || getClass() != o.getClass())

            return false;

 

        ZNodeName sequence = (ZNodeName) o;

 

        if (!name.equals(sequence.name))

            return false;

 

        return true;

    }

 

    @Override

    public int hashCode() {

        return name.hashCode() + 37;

    }

 

    public int compareTo(ZNodeName that) {

        int s1 = this.sequence;

        int s2 = that.sequence;

        if (s1 == -1 && s2 == -1) {

            return this.name.compareTo(that.name);

        }

        if (s1 == -1) {

            return -1;

        } else if (s2 == -1) {

            return 1;

        } else {

            return s1 - s2;

        }

    }

 

    /**

     * Returns the name of the znode

     */

    public String getName() {

        return name;

    }

 

    /**

     * Returns the sequence number

     */

    public int getZNodeName() {

        return sequence;

    }

 

    /**

     * Returns the text prefix before the sequence number

     */

    public String getPrefix() {

        return prefix;

    }

}


PS:这个ZNodeName类是被我修改过的,官方的代码比较有问题,官方的先用了节点路径的前缀prefix比较,再去比较sequence序号是不对的,这样会导致sessionid小的总是能拿到锁。应该直接比较全局有序的sequence序号,小的先拿到锁,先到先得。


看完本文有收获?请转发分享给更多人

关注「ImportNew」,看技术干货

以上是关于基于 ZooKeeper 的分布式锁和队列( 上 )的主要内容,如果未能解决你的问题,请参考以下文章

使用ZooKeeper实现分布式队列分布式锁和选举详解

Zookeeper入门

Zookeeper 概念

快速系统理解 ZooKeeper 的原理

ZooKeeper

Zookeeper