kafka清空topics

1、缘由（需求）：因为测试时不小心，写了一大堆topics，看监测时一顿揪心。

2、解决方案：kafka是被zookeeper监管的，topics是在zookeeper上的（描述可能不是很准确），我发现kakfa原生api好像没有批量删除topics的方式，我登陆到zookeeper后，在zookeeper的目录下发现了topics。于是写了个递归程序将其删除。以下是代码：

public class App {
	static String path = "/brokers";
	static ZKConnection zkc = null;

	public static void main(String[] args) throws Exception {
		zkc = new ZKConnection();
		zkc.connect();
		get(zkc.zk.getChildren(path, true), path);

	}

	public static void get(List<String> paths, String path) throws Exception {
		for (String p : paths) {
			p = path + "/" + p;
			if (null == zkc.zk.exists(p, true)) {
				System.out.println(p);
				//zkc.zk.delete(p, 0);
			} else {
				get(zkc.zk.getChildren(p, true), p);
				int i =zkc.zk.exists(p,true).getVersion();
				zkc.zk.delete(p,i);
			}

			System.out.println(p);
		}
	}
}

class ZKConnection {
	/**
	 * server列表, 以逗号分割
	 */
	protected String hosts = "server1:2181,server2:2181,server3:2181";
	/**
	 * 连接的超时时间, 毫秒
	 */
	private static final int SESSION_TIMEOUT = 5000;
	private CountDownLatch connectedSignal = new CountDownLatch(1);
	public ZooKeeper zk;

	/**
	 * 连接zookeeper server
	 */
	public void connect() throws Exception {
		zk = new ZooKeeper(hosts, SESSION_TIMEOUT, new ConnWatcher());
		// 等待连接完成
		connectedSignal.await();
	}

	public class ConnWatcher implements Watcher {
		public void process(WatchedEvent event) {
			// 连接建立, 回调process接口时, 其event.getState()为KeeperState.SyncConnected
			if (event.getState() == KeeperState.SyncConnected) {
				// 放开闸门, wait在connect方法上的线程将被唤醒
				connectedSignal.countDown();
			}
		}
	}

注：连接代码是网上得的

3.重启kafka，完毕。