使用zookeeper管理多个hbase集群

zookeeper是hbase集群的"协调器"。由于zookeeper的轻量级特性，因此我们可以将多个hbase集群共用一个zookeeper集 群，以节约大量的服务器。多个hbase集群共用zookeeper集群的方法是使用同一组ip，修改不同hbase集群 的"zookeeper.znode.parent"属性，让它们使用不同的根目录。比如cluster1使用/hbase-c1,cluster2使用 /hbase-c2，等等。

 

    使用以上方法有一个现实的问题：如何避免各集群的相互干扰？因为client的配置权是在用户手上，并不能保证用户永远是配置正确的，那么会产生某个用户 访问了不该他访问的hbase集群。此时数据安全性成了很大的问题，甚至可能出现误删除数据。我们需要在zookeeper层屏弊掉该问题。

 

    zookeeper3.x版本起自带了简单的ACL功能(注意3.3.x版本起不再支持按hostname来分配权限)。见：http://zookeeper.apache.org/doc/r3.3.2/zookeeperProgrammers.html#sc_ZooKeeperAccessControl。进行权限配置主要使用digest和ip两种方法。其中digest是用户密码方式，对用户来说使用上并不透明。ip配置最简单，对用户也是透明的，用户并不知道的情况下就能限制它的访问权限。

 

    zookeeper将访问权限分为了五类:READ/WRITE/DELETE/CREATE/ADMIN，其中admin为最高权限。zookeeper的权限是到znode级别的，限制了某一个node的权限并不能限制它的子节点权限。

 

    不过使用IP做权限配置方案有一个缺陷：必须指定具体的ip，而不能使用通配符或者范围一类的。这样对于大规模的权限设置是非常不方便的一件事，因此作者略调整了一下zookeeper的代码：

  IPAuthenticationProvider.java

/** 
 * Licensed to the Apache Software Foundation (ASF) under one 
 * or more contributor license agreements.  See the NOTICE file 
 * distributed with this work for additional information 
 * regarding copyright ownership.  The ASF licenses this file 
 * to you under the Apache License, Version 2.0 (the 
 * "License"); you may not use this file except in compliance 
 * with the License.  You may obtain a copy of the License at 
 * 
 *     http://www.apache.org/licenses/LICENSE-2.0 
 * 
 * Unless required by applicable law or agreed to in writing, software 
 * distributed under the License is distributed on an "AS IS" BASIS, 
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 
 * See the License for the specific language governing permissions and 
 * limitations under the License. 
 */  
  
package org.apache.zookeeper.server.auth;  
  
import org.apache.zookeeper.data.Id;  
import org.apache.zookeeper.server.ServerCnxn;  
import org.apache.zookeeper.KeeperException;  
  
public class IPAuthenticationProvider implements AuthenticationProvider {  
  
    public String getScheme() {  
        return "ip";  
    }  
  
    public KeeperException.Code  
        handleAuthentication(ServerCnxn cnxn, byte[] authData)  
    {  
        String id = cnxn.getRemoteAddress().getAddress().getHostAddress();  
        cnxn.getAuthInfo().add(new Id(getScheme(), id));  
        return KeeperException.Code.OK;  
    }  
  
    // This is a bit weird but we need to return the address and the number of  
    // bytes (to distinguish between IPv4 and IPv6  
    private byte[] addr2Bytes(String addr) {  
        byte b[] = v4addr2Bytes1(addr);  
        // TODO Write the v6addr2Bytes  
        return b;  
    }  
  
    private byte v4addr2Bytes(String part) throws NumberFormatException{  
        try {  
           int v = Integer.parseInt(part);  
           if (v >= 0 && v <= 255) {  
                  byte  b = (byte) v;  
                  return b;  
                } else {  
                    throw new NumberFormatException("v < 0 or v > 255!");  
                }  
            } catch (NumberFormatException e) {  
                throw e;  
            }  
    }  
      
    private byte[] v4addr2Bytes1(String addr) {  
      String parts[] = addr.split("\\.", -1);  
      if (parts.length != 4) {  
          return null;  
      }  
      byte b[] = new byte[4];  
      for (int i = 0; i < 4; i++) {  
        try {  
            if(parts[i].split("/").length == 2){  
                v4addr2Bytes(parts[i].split("/")[0]);  
                v4addr2Bytes(parts[i].split("/")[1]);  
                continue;  
            }else{  
            b[i] = v4addr2Bytes(parts[i]);  
            }  
        } catch (NumberFormatException e) {  
          return null;  
            }  
      }  
      return b;  
  }  
      
    public boolean matches(String id, String aclExpr) {  
      String parts[] = aclExpr.split("/", 2);  
      byte aclAddr[] = addr2Bytes(parts[0]);  
      if (aclAddr == null) {  
          return false;  
      }  
      byte endAclAddr[] = new byte[aclAddr.length];  
      for(int i = 0; i < aclAddr.length; i ++){  
        endAclAddr[i] = aclAddr[i];  
      }  
      if (parts.length == 2) {  
          try {  
                int end = Integer.parseInt(parts[1]);  
                int e = endAclAddr[endAclAddr.length-1]<=0?endAclAddr[endAclAddr.length-1]+256:endAclAddr[endAclAddr.length-1];  
                if(end <  e|| end < 0 || end > 255)  
                    return false;  
              endAclAddr[endAclAddr.length-1] = (byte)end;  
          } catch (NumberFormatException e) {  
              return false;  
          }  
      }  
      byte remoteAddr[] = addr2Bytes(id);  
      if (remoteAddr == null) {  
          return false;  
      }  
      for (int i = 0; i < remoteAddr.length; i++) {  
        int r = remoteAddr[i]<=0?(int)remoteAddr[i]+256:remoteAddr[i];  
        int a = aclAddr[i]<=0?(int)aclAddr[i]+256:aclAddr[i];  
        int e = endAclAddr[i]<=0?(int)endAclAddr[i]+256:endAclAddr[i];  
        if (r < a || r > e) {  
          return false;  
        }  
      }  
      return true;  
  }  
  
    public boolean isAuthenticated() {  
        return false;  
    }  
  
    public boolean isValid(String id) {  
        return addr2Bytes(id) != null;  
    }  
}  

支持了使用/做为范围标识，比如进入hbase zkcli，执行：setAcl /test ip:192.168.0.3/10:cd，则将读写权限赋给了192.168.0.3-192.168.0.10这8台机器，其它机器将没有任何权限。

    这样用同一个zookeeper管理多个集群、海量机器将不再有困扰。

    最后写了一个帮助运维同学自动化管理zookeeper集群下多个hbase集群的ACL权限的工具，像以下这样：

java -Djava.ext.dirs=libs/ -cp hbase-tools.jar dwbasis.hbase.tools.client.ZookeeperAcl aclFile.json  
Usage: ZookeeperAcl acljsonfile [-plan]  
/test/t ==> 'ip,'192.168.0.1  
:cdrwa  
/test ==> 'ip,'192.168.0.1/3  
:cdrwa  
/test ==> 'ip,'192.168.0.5  
:cdrwa  
do you really setAcl as above?(y/n)  

    补充：多集群共用zk后，每个集群的启动和停止不应该影响zk的稳定。因此请配置hbase-env.sh中export HBASE_MANAGES_ZK=false