Python 官方文档:入门教程 => 点击学习
目录背景技术方案分布式协调中间件基于ZooKeeper的技术方案服务注册目录设计服务分片处理流程编码实现总结背景 公司的一个服务需要做类似于分片的逻辑,一开始服务基于传统部署方式通过
公司的一个服务需要做类似于分片的逻辑,一开始服务基于传统部署方式通过本地配置文件配置的方式就可以指定该机器服务的分片内容如:0,1,2,3,随着系统的升级迭代,该服务进行了容器化部署,所以原来基于本地配置文件各自配置分片数据的方式就不适用了,原来的部署方式使得服务是有状态,是一种非云原生的方式,所以该服务要重新设计实现一套分布式服务分片
逻辑。
要实现分布式服务分片的能力,需要有一个分布式中间件,如:Redis
,Mysql
,Zookeeper
等等都可以,我们选用Zookeeper
。
使用Zookeeper
主要是基于Zookeeper
的临时节点和节点变化监听机制,具体的技术设计如下:
Zookeeper
的数据存储结构类似于目录,服务注册后的目录类似如下结构:
解释下该目录结构,首先/xxxx/xxxx/sharding
是区别于其他业务的的目录,该目录节点是持久的,service
是服务目录,标识一个服务,该节点也是持久的,ip1
,ip2
是该服务注册到Zookeeper
的机器列表节点,该节点是临时节点。
/xxxx/xxxx/sharding/service/ip1
-----|----|--------|-------/ip2
CuratorFramework
客户端,设置客户端连接状态监听;Zookeeper
注册该机器的信息,这里设计简单,机器信息就是ip
地址;Zookeeper
获取所有注册信息;Zookeeper
获取的所有注册机器信息根据分片算法进行分片计算。ZookeeperConfig
Zookeeper
的配置信息
@Data
public class ZookeeperConfig {
private String zkAddress;
private String nodePath;
private String serviceName;
private Integer shardinGCount;
public ZookeeperConfig(String zkAddress, String nodePath, String serviceName, Integer shardingCount) {
this.zkAddress = zkAddress;
this.nodePath = nodePath;
this.serviceName = "/" + serviceName;
this.shardingCount = shardingCount;
}
private int baseSleepTimeMilliseconds = 1000;
private int maxSleepTimeMilliseconds = 3000;
private int maxRetries = 3;
private int sessionTimeoutMilliseconds;
private int connectionTimeoutMilliseconds;
}
InstanceInfo注册机器
@AllArgsConstructor
@EqualsAndHashCode()
public class InstanceInfo {
private String ip;
public String getInstance() {
return ip;
}
}
ZookeeperShardingService分片服务
@Slf4j
public class ZookeeperShardingService {
public final Map<String, List<Integer>> caches = new HashMap<>(16);
private final CuratorFramework client;
private final ZookeeperConfig zkConfig;
private final ShardingStrategy shardingStrategy;
private final InstanceInfo instanceInfo;
private static final CountDownLatch COUNT_DOWN_LATCH = new CountDownLatch(1);
public ZookeeperShardingService(ZookeeperConfig zkConfig, ShardingStrategy shardingStrategy) {
this.zkConfig = zkConfig;
log.info("开始初始化zk, ip列表是: {}.", zkConfig.getZkAddress());
CuratorFrameworkFactory.Builder builder = CuratorFrameworkFactory.builder()
.connectString(zkConfig.getZkAddress())
.retryPolicy(new ExponentialBackoffRetry(zkConfig.getBaseSleepTimeMilliseconds(), zkConfig.getMaxRetries(), zkConfig.getMaxSleepTimeMilliseconds()));
if (0 != zkConfig.getSessionTimeoutMilliseconds()) {
builder.sessionTimeoutMs(zkConfig.getSessionTimeoutMilliseconds());
}
if (0 != zkConfig.getConnectionTimeoutMilliseconds()) {
builder.connectionTimeoutMs(zkConfig.getConnectionTimeoutMilliseconds());
}
this.shardingStrategy = shardingStrategy;
HostInfo host = new HostInfo();
this.instanceInfo = new InstanceInfo(host.getAddress());
client = builder.build();
client.getConnectionStateListenable().addListener(new ConnectionListener());
client.start();
try {
COUNT_DOWN_LATCH.await();
} catch (InterruptedException e) {
e.printStackTrace();
}
// 注册服务节点监听
reGISterPathChildListener(zkConfig.getNodePath() + zkConfig.getServiceName(), new ChildrenPathListener());
try {
if (!client.blockUntilConnected(zkConfig.getMaxSleepTimeMilliseconds() * zkConfig.getMaxRetries(), TimeUnit.MILLISECONDS)) {
client.close();
throw new KeeperException.OperationTimeoutException();
}
} catch (final Exception ex) {
ex.printStackTrace();
throw new RuntimeException(ex);
}
}
private void registerPathChildListener(String nodePath, PathChildrenCacheListener listener) {
try {
// 1. 创建一个PathChildrenCache
PathChildrenCache pathChildrenCache = new PathChildrenCache(client, nodePath, true);
// 2. 添加目录监听器
pathChildrenCache.getListenable().addListener(listener);
// 3. 启动监听器
pathChildrenCache.start(PathChildrenCache.StartMode.BUILD_INITIAL_CACHE);
} catch (Exception e) {
log.error("注册子目录监听器出现异常,nodePath:{}",nodePath,e);
throw new RuntimeException(e);
}
}
private void zkOp() throws Exception {
// 是否存在ruubypay-sharding主节点
if (null == client.checkExists().forPath(zkConfig.getNodePath())) {
client.create().creatingParentsIfNeeded().withMode(CreateMode.PERSISTENT).forPath(zkConfig.getNodePath(), Hashing.sha1().hashString("sharding", Charsets.UTF_8).toString().getBytes());
}
// 是否存服务主节点
if (null == client.checkExists().forPath(zkConfig.getNodePath() + zkConfig.getServiceName())) {
// 创建服务主节点
client.create().creatingParentsIfNeeded().withMode(CreateMode.PERSISTENT).forPath(zkConfig.getNodePath() + zkConfig.getServiceName());
}
// 检查是否存在临时节点
if (null == client.checkExists().forPath(zkConfig.getNodePath() + zkConfig.getServiceName() + "/" + instanceInfo.getInstance())) {
System.out.println(zkConfig.getNodePath() + zkConfig.getServiceName() + "/" + instanceInfo.getInstance());
// 创建临时节点
client.create().creatingParentsIfNeeded().withMode(CreateMode.EPHEMERAL).forPath(zkConfig.getNodePath() + zkConfig.getServiceName() +
"/" + instanceInfo.getInstance(), zkConfig.getShardingCount().toString().getBytes(StandardCharsets.UTF_8));
}
shardingFromZk();
}
private void shardingFromZk() throws Exception {
// 从 serviceName 节点下获取所有Ip列表
final GetChildrenBuilder childrenBuilder = client.getChildren();
final List<String> instanceList = childrenBuilder.watched().forPath(zkConfig.getNodePath() + zkConfig.getServiceName());
List<InstanceInfo> res = new ArrayList<>();
instanceList.forEach(s -> {
res.add(new InstanceInfo(s));
});
Map<InstanceInfo, List<Integer>> shardingResult = shardingStrategy.sharding(res, zkConfig.getShardingCount());
// 先清一遍缓存
caches.clear();
shardingResult.forEach((k, v) -> {
caches.put(k.getInstance().split("-")[0], v);
});
}
private class ConnectionListener implements ConnectionStateListener {
@Override
public void stateChanged(CuratorFramework client, ConnectionState newState) {
if (newState == ConnectionState.CONNECTED || newState == ConnectionState.LOST || newState == ConnectionState.RECONNECTED) {
try {
zkOp();
} catch (Exception e) {
e.printStackTrace();
throw new RuntimeException(e);
} finally {
COUNT_DOWN_LATCH.countDown();
}
}
}
}
private class ChildrenPathListener implements PathChildrenCacheListener {
@Override
public void childEvent(CuratorFramework client, PathChildrenCacheEvent event) {
PathChildrenCacheEvent.Type type = event.getType();
if (PathChildrenCacheEvent.Type.CHILD_ADDED == type || PathChildrenCacheEvent.Type.CHILD_REMOVED == type) {
try {
shardingFromZk();
} catch (Exception e) {
e.printStackTrace();
throw new RuntimeException(e);
}
}
}
}
}
分片算法
采用平均分配的算法
public interface ShardingStrategy {
Map<InstanceInfo, List<Integer>> sharding(final List<InstanceInfo> list, Integer shardingCount);
}
public class AverageAllocationShardingStrategy implements ShardingStrategy {
@Override
public Map<InstanceInfo, List<Integer>> sharding(List<InstanceInfo> list, Integer shardingCount) {
if (list.isEmpty()) {
return null;
}
Map<InstanceInfo, List<Integer>> result = shardingAliquot(list, shardingCount);
addAliquant(list, shardingCount, result);
return result;
}
private Map<InstanceInfo, List<Integer>> shardingAliquot(final List<InstanceInfo> instanceInfos, final int shardingTotalCount) {
Map<InstanceInfo, List<Integer>> result = new LinkedHashMap<>(shardingTotalCount, 1);
int itemCountPerSharding = shardingTotalCount / instanceInfos.size();
int count = 0;
for (InstanceInfo each : instanceInfos) {
List<Integer> shardingitems = new ArrayList<>(itemCountPerSharding + 1);
for (int i = count * itemCountPerSharding; i < (count + 1) * itemCountPerSharding; i++) {
shardingItems.add(i);
}
result.put(each, shardingItems);
count++;
}
return result;
}
private void addAliquant(final List<InstanceInfo> instanceInfos, final int shardingTotalCount, final Map<InstanceInfo, List<Integer>> shardingResults) {
int aliquant = shardingTotalCount % instanceInfos.size();
int count = 0;
for (Map.Entry<InstanceInfo, List<Integer>> entry : shardingResults.entrySet()) {
if (count < aliquant) {
entry.getValue().add(shardingTotalCount / instanceInfos.size() * instanceInfos.size() + count);
}
count++;
}
}
}
基于Zookeeper
和简单的平均分配算法实现了一个简单的分布式分片服务,该分片服务目前满足公司需求,因为其简单,所以不一定满足其他场景,针对其他场景还需考虑其他因素,该示例供参考。
到此这篇关于Java Zookeeper分布式分片算法超详细讲解流程的文章就介绍到这了,更多相关Java Zookeeper分布式分片算法内容请搜索编程网以前的文章或继续浏览下面的相关文章希望大家以后多多支持编程网!
--结束END--
本文标题: Java Zookeeper分布式分片算法超详细讲解流程
本文链接: https://lsjlt.com/news/197961.html(转载时请注明来源链接)
有问题或投稿请发送至: 邮箱/279061341@qq.com QQ/279061341
2024-03-01
2024-03-01
2024-03-01
2024-02-29
2024-02-29
2024-02-29
2024-02-29
2024-02-29
2024-02-29
2024-02-29
回答
回答
回答
回答
回答
回答
回答
回答
回答
回答
0