转载

Redis从入门到放弃系列(十) Cluster

本文例子基于：5.0.4

Redis Cluster集群高可用方案,去中心化,最基本三主多从,主从切换类似Sentinel,关于Sentinel内容可以查看编者另外一篇【 Redis从入门到放弃系列(九) Sentinel 】.

在Redis Cluster中,只存在index为0的数据库,而且其实Redis作为单线程,如果在同一个实例上创建多个库的话,也是需要上下文切换的.

slot

由于Redis Cluster是采用16384个slot来划分数据的,也就是说你当前插入的数据会存在不同的节点上,简而言之不支持比较复杂的多建操作(可以对key打上hash tags来解决).

我们说Cluster是按照16384个slot来划分数据的,那么是如何来确定一个key落在那个节点上呢?

//计算slot
HASH_SLOT = CRC16(key) mod 16384
复制代码

每个节点会拥有一部分的slot,通过上述获取到具体key的slot即知道应该去哪儿找对应的节点啦.可是在网络中,一切都会有不存稳定因素,网络抖动.

当在Cluster中存在网络抖动的时候,当时间过长,有可能产生下线,其实原理跟Sentinel里面讲的很相似,因为都是依赖Gossip协议来实现的.可以通过以下配置来设置确定下线的时间.

//节点持续timeout的时间,才认定该节点出现故障,需要进行主从切换,
cluster-node-timeout
//作为上面timeout的系数来放大时间
cluster-replica-validity-factor
复制代码

由于数据是按照16384个slot去划分的,那么当我们在请求某个key到错误的节点,这时候key不在该节点上,Redis会向我们发送一个错误

-MOVED 3999 127.0.0.1:6381
复制代码

该消息是提示我们该key应该是存在 127.0.0.1 这台服务器上面的3999slot,这时候就需要我们的redis客户端去纠正本地的slot映射表,然后请求对应的地址.

增删集群节点

当我们在增加或者删除某个节点的时候,其实就只是将slot从某个节点移动到另外一个节点.可以使用一下命令来完成这一件事

CLUSTER ADDSLOTS slot1 [slot2] ... [slotN]
CLUSTER DELSLOTS slot1 [slot2] ... [slotN]
CLUSTER SETSLOT slot NODE node
CLUSTER SETSLOT slot MIGRATING node
CLUSTER SETSLOT slot IMPORTING node 有时候运维需要对redis节点的某些数据做迁移,官方提供了redis-trib工具来完成这件事情。

在迁移的时候,redis节点会存在两种状态,一种是MIGRATING和IMPORTING,用于将slot从一个节点迁移到另外一个节点.

节点状态设置为MIGRATING时，将接受与此散列槽有关的所有查询，但仅当有问题的key存在时才能接受，否则将使用-Ask重定向将查询转发到作为迁移目标的节点。
节点状态设置为IMPORTING时，节点将接受与此哈希槽有关的所有查询，但前提是请求前面有ASKING命令。如果客户端没有发出ASKING命令，查询将通过-MOVED重定向错误重定向到真正的散列槽所有者

多线程批量获取/删除

public class RedisUtils {

	private static final String LOCK_SUCCESS = "OK";
	private static final String SET_IF_NOT_EXIST = "NX";
	private static final String SET_WITH_EXPIRE_TIME = "PX";
	private static final Long RELEASE_SUCCESS = 1L;

	private final ThreadLocal<String> requestId = new ThreadLocal<>();

	private final static ExecutorService executorService = new ThreadPoolExecutor(
			//核心线程数量
			1,
			//最大线程数量
			8,
			//当线程空闲时，保持活跃的时间
			1000,
			//时间单元 ，毫秒级
			TimeUnit.MILLISECONDS,
			//线程任务队列
			new LinkedBlockingQueue<>(1024),
			//创建线程的工厂
			new RedisTreadFactory("redis-batch"));

	@Autowired
	private JedisCluster jedisCluster;

	public String set(String key, String value) {
		return jedisCluster.set(key, value);
	}

	public String get(String key) {
		return jedisCluster.get(key);
	}

	public Map<String, String> getBatchKey(List<String> keys) {
		Map<Jedis, List<String>> nodeKeyListMap = jedisKeys(keys);
		//结果集
		Map<String, String> resultMap = new HashMap<>();
		CompletionService<Map<String,String>> batchService = new ExecutorCompletionService(executorService);
		nodeKeyListMap.forEach((k,v)->{
			batchService.submit(new BatchGetTask(k,v));
		});
		nodeKeyListMap.forEach((k,v)->{
			try {
				resultMap.putAll(batchService.take().get());
			} catch (InterruptedException | ExecutionException e) {
				e.printStackTrace();
			}
		});
		return resultMap;
	}

	public boolean lock(String lockKey, long expireTime){
		String uuid = UUID.randomUUID().toString();
		requestId.set(uuid);
		String result = jedisCluster.set(lockKey, uuid, SET_IF_NOT_EXIST, SET_WITH_EXPIRE_TIME, expireTime);
		return LOCK_SUCCESS.equals(result);
	}

	public boolean unLock(String lockKey){
		String uuid = requestId.get();
		String script = "if redis.call('get', KEYS[1]) == ARGV[1] then return redis.call('del', KEYS[1]) else return 0 end";
		Object result = jedisCluster.eval(script, Collections.singletonList(lockKey), Collections.singletonList(uuid));
		requestId.remove();
		return RELEASE_SUCCESS.equals(result);
	}

	private Map<Jedis, List<String>> jedisKeys(List<String> keys){
		Map<Jedis, List<String>> nodeKeyListMap = new HashMap<>();
		for (String key : keys) {
			//计算slot
			int slot = JedisClusterCRC16.getSlot(key);
			Jedis jedis = jedisCluster.getConnectionFromSlot(slot);
			if (nodeKeyListMap.containsKey(jedis)) {
				nodeKeyListMap.get(jedis).add(key);
			} else {
				nodeKeyListMap.put(jedis, Arrays.asList(key));
			}
		}
		return nodeKeyListMap;
	}

	public long delBatchKey(List<String> keys){
		Map<Jedis, List<String>> nodeKeyListMap = jedisKeys(keys);
		CompletionService<Long> batchService = new ExecutorCompletionService(executorService);
		nodeKeyListMap.forEach((k,v)->{
			batchService.submit(new BatchDelTask(k,v));
		});
		Long result = 0L;
		for (int i=0;i<nodeKeyListMap.size();i++){
			try {
				result += batchService.take().get();
			} catch (InterruptedException | ExecutionException e) {
				e.printStackTrace();
			}
		}
		return result;
	}

	class BatchGetTask implements Callable<Map<String,String>>{

		private Jedis jedis;

		private List<String> keys;

		private BatchGetTask(Jedis jedis, List<String> keys) {
			this.jedis = jedis;
			this.keys = keys;
		}

		@Override
		public Map<String, String> call() throws Exception {
			Map<String, String> resultMap = new HashMap<>();
			String[] keyArray = keys.toArray(new String[]{});
			try {
				List<String> nodeValueList = jedis.mget(keyArray);
				for (int i = 0; i < keys.size(); i++) {
					resultMap.put(keys.get(i),nodeValueList.get(i));
				}
			}finally {
				jedis.close();
			}
			return resultMap;
		}
	}

	class BatchDelTask implements Callable<Long>{

		private Jedis jedis;

		private List<String> keys;

		private BatchDelTask(Jedis jedis, List<String> keys) {
			this.jedis = jedis;
			this.keys = keys;
		}

		@Override
		public Long call() throws Exception {
			String[] keyArray = keys.toArray(new String[]{});
			try {
				return jedis.del(keyArray);
			}finally {
				jedis.close();
			}
		}
	}

	 static class RedisTreadFactory implements ThreadFactory{

		private final AtomicInteger threadNumber = new AtomicInteger(0);

		private final String namePredix;

		public RedisTreadFactory(String namePredix) {
			this.namePredix = namePredix +"-";
		}

		@Override
		public Thread newThread(Runnable r) {
			Thread t = new Thread( r,namePredix + threadNumber.getAndIncrement());
			if (t.isDaemon())
				t.setDaemon(true);
			if (t.getPriority() != Thread.NORM_PRIORITY)
				t.setPriority(Thread.NORM_PRIORITY);
			return t;
		}
	}
}
复制代码