ZooKeeper is a distributed, open source distributed application coordination service, an open source implementation of Google's Chubby, and an important component of Hadoop and Hbase. It is a software that provides consistent services for distributed applications. The functions provided include: configuration maintenance, domain name service, distributed synchronization, group service, etc.
The architecture of ZooKeeper achieves high availability through redundant services. Therefore, if there is no answer the first time, the client can ask another ZooKeeper host. ZooKeeper nodes store their data in a hierarchical namespace, much like a file system or a prefix tree structure. Clients can read and write on the node, thus having a shared configuration service in this way. Updates are totally ordered.
Process based on ZooKeeper distributed lock
- Create a temporary sequential node node_n under the zookeeper specified node (locks)
- Get all child nodes children under locks
- Sort child nodes according to the node's self-incrementing sequence number from small to large
- Determine whether this node is the first child node, if so, acquire the lock; if not, listen for the deletion event of the node smaller than the node
- If the listening event takes effect, go back to the second step to re-judgment until the lock is acquired
Implementation
The following is the specific use of java and zookeeper to implement distributed locks, and the zookeeper package provided by apache is used to operate zookeeper.
- By implementing the Watch interface, implement the process (WatchedEvent event) method to implement monitoring, use CountDownLatch to complete the monitoring, use CountDownLatch to count when waiting for the lock, and then perform countDown after waiting, stop waiting, and continue to run.
- The following overall process is basically the same as the above description process, except that CountDownLatch is used to monitor the previous node when monitoring.
Distributed lock
import org.apache.zookeeper.*;
import org.apache.zookeeper.data.Stat;
import java.io.IOException;
import java.util.ArrayList;
import java.util.Collections;
import java.util.List;
import java.util.concurrent.CountDownLatch;
import java.util.concurrent.TimeUnit;
import java.util.concurrent.locks.Condition;
import java.util.concurrent.locks.Lock;
/**
* Created by liuyang on 2017/4/20.
*/
public class DistributedLock implements Lock, Watcher {
private ZooKeeper zk = null;
// 根节点
private String ROOT_LOCK = "/locks";
// 竞争的资源
private String lockName;
// 等待的前一个锁
private String WAIT_LOCK;
// 当前锁
private String CURRENT_LOCK;
// 计数器
private CountDownLatch countDownLatch;
private int sessionTimeout = 30000;
private List<Exception> exceptionList = new ArrayList<Exception>();
/**
* 配置分布式锁
* @param config 连接的url
* @param lockName 竞争资源
*/
public DistributedLock(String config, String lockName) {
this.lockName = lockName;
try {
// 连接zookeeper
zk = new ZooKeeper(config, sessionTimeout, this);
Stat stat = zk.exists(ROOT_LOCK, false);
if (stat == null) {
// 如果根节点不存在，则创建根节点
zk.create(ROOT_LOCK, new byte[0], ZooDefs.Ids.OPEN_ACL_UNSAFE, CreateMode.PERSISTENT);
}
} catch (IOException e) {
e.printStackTrace();
} catch (InterruptedException e) {
e.printStackTrace();
} catch (KeeperException e) {
e.printStackTrace();
}
}
// 节点监视器
public void process(WatchedEvent event) {
if (this.countDownLatch != null) {
this.countDownLatch.countDown();
}
}
public void lock() {
if (exceptionList.size() > 0) {
throw new LockException(exceptionList.get(0));
}
try {
if (this.tryLock()) {
System.out.println(Thread.currentThread().getName() + " " + lockName + "获得了锁");
return;
} else {
// 等待锁
waitForLock(WAIT_LOCK, sessionTimeout);
}
} catch (InterruptedException e) {
e.printStackTrace();
} catch (KeeperException e) {
e.printStackTrace();
}
}
public boolean tryLock() {
try {
String splitStr = "_lock_";
if (lockName.contains(splitStr)) {
throw new LockException("锁名有误");
}
// 创建临时有序节点
CURRENT_LOCK = zk.create(ROOT_LOCK + "/" + lockName + splitStr, new byte[0],
ZooDefs.Ids.OPEN_ACL_UNSAFE, CreateMode.EPHEMERAL_SEQUENTIAL);
System.out.println(CURRENT_LOCK + " 已经创建");
// 取所有子节点
List<String> subNodes = zk.getChildren(ROOT_LOCK, false);
// 取出所有lockName的锁
List<String> lockObjects = new ArrayList<String>();
for (String node : subNodes) {
String _node = node.split(splitStr)[0];
if (_node.equals(lockName)) {
lockObjects.add(node);
}
}
Collections.sort(lockObjects);
System.out.println(Thread.currentThread().getName() + " 的锁是 " + CURRENT_LOCK);
// 若当前节点为最小节点，则获取锁成功
if (CURRENT_LOCK.equals(ROOT_LOCK + "/" + lockObjects.get(0))) {
return true;
}
// 若不是最小节点，则找到自己的前一个节点
String prevNode = CURRENT_LOCK.substring(CURRENT_LOCK.lastIndexOf("/") + 1);
WAIT_LOCK = lockObjects.get(Collections.binarySearch(lockObjects, prevNode) - 1);
} catch (InterruptedException e) {
e.printStackTrace();
} catch (KeeperException e) {
e.printStackTrace();
}
return false;
}
public boolean tryLock(long timeout, TimeUnit unit) {
try {
if (this.tryLock()) {
return true;
}
return waitForLock(WAIT_LOCK, timeout);
} catch (Exception e) {
e.printStackTrace();
}
return false;
}
// 等待锁
private boolean waitForLock(String prev, long waitTime) throws KeeperException, InterruptedException {
Stat stat = zk.exists(ROOT_LOCK + "/" + prev, true);
if (stat != null) {
System.out.println(Thread.currentThread().getName() + "等待锁 " + ROOT_LOCK + "/" + prev);
this.countDownLatch = new CountDownLatch(1);
// 计数等待，若等到前一个节点消失，则precess中进行countDown，停止等待，获取锁
this.countDownLatch.await(waitTime, TimeUnit.MILLISECONDS);
this.countDownLatch = null;
System.out.println(Thread.currentThread().getName() + " 等到了锁");
}
return true;
}
public void unlock() {
try {
System.out.println("释放锁 " + CURRENT_LOCK);
zk.delete(CURRENT_LOCK, -1);
CURRENT_LOCK = null;
zk.close();
} catch (InterruptedException e) {
e.printStackTrace();
} catch (KeeperException e) {
e.printStackTrace();
}
}
public Condition newCondition() {
return null;
}
public void lockInterruptibly() throws InterruptedException {
this.lock();
}
public class LockException extends RuntimeException {
private static final long serialVersionUID = 1L;
public LockException(String e){
super(e);
}
public LockException(Exception e){
super(e);
}
}
}test code
public class Test {
static int n = 500;
public static void secskill() {
System.out.println(--n);
}
public static void main(String[] args) {
Runnable runnable = new Runnable() {
public void run() {
DistributedLock lock = null;
try {
lock = new DistributedLock("127.0.0.1:2181", "test1");
lock.lock();
secskill();
System.out.println(Thread.currentThread().getName() + "正在运行");
} finally {
if (lock != null) {
lock.unlock();
}
}
}
};
for (int i = 0; i < 10; i++) {
Thread t = new Thread(runnable);
t.start();
}
}
}operation result:
In general, if you understand the entire implementation process, it is not very difficult to use zookeeper to implement distributed locks, but this is just a simple implementation. Compared with the previous implementation of Redis, this implementation is more stable, because this is because Due to the characteristics of zookeeper, from the outside, every node in the zookeeper cluster is consistent.
The full code can be viewed in my GitHub: https://github.com/yangliu0/DistributedLock