hbase simple权限一

一:shell执行:

grant授权:

查看commands目录下grant.rb文件:

security_admin.grant(user, permissions, table_name, family, qualifier)

进入security.rb文件执行:

org.apache.hadoop.hbase.security.access.AccessControlClient.grant( @connection, tableName, user, fambytes, qualbytes, perm.getActions())

revoke解除权限:

commands目录下revoke.rb文件:

security_admin.revoke(user, table_name, family, qualifier)

紧接着进入security.rb文件执行:

org.apache.hadoop.hbase.security.access.AccessControlClient.revoke(@connection, tableName, user, fambytes, qualbytes)

二:hbase授权:

进入AccessController服务端处理请求类,(客户端类为AccessControllerClient类;)这个类执行hbase权限操作;找到授权代码:

public void grant(RpcController controller,
AccessControlProtos.GrantRequest request,
RpcCallback<AccessControlProtos.GrantResponse> done)

核心代码块:

User.runAsLoginUser(new PrivilegedExceptionAction<Void>() {
  @Override
  public Void run() throws Exception {
AccessControlLists.addUserPermission(regionEnv.getConfiguration(), perm,
  regionEnv.getTable(AccessControlLists.ACL_TABLE_NAME), request.getMergeExistingPermissions());
return null;
  }
});

执行添加用户权限到acl表一步,进入到addUserPermission方法,存储用户权限到表中:

看代码:

//获取权限行为

Permission.Action[] actions = userPerm.getActions();

//声明rowkey
//1.namespace存在,为namespace;2.表table与namespace不存在,为默认acl;3.否则为tableName

byte[] rowKey = userPermissionRowKey(userPerm);

//初始化put对象,为后面插入数据表准备

Put p = new Put(rowKey);

//插入权限行为对象的key

byte[] key = userPermissionKey(userPerm);

//action的判断

if ((actions == null) || (actions.length == 0)) {

...................中间处理,包含数据转byte等

p.addImmutable(ACL_LIST_FAMILY, key, value);

最后将权限数据插入acl表中:

t.put(p);

hbase授权结束。

三:hbase解除权限:

hbase解除权限与hbase授权一样;核心代码:

//扎到之前存储的匹配的权限移除
static void removeUserPermission(Configuration conf, UserPermission userPerm, Table t)
  throws IOException {
Delete d = new Delete(userPermissionRowKey(userPerm));
byte[] key = userPermissionKey(userPerm);

if (LOG.isDebugEnabled()) {
  LOG.debug("Removing permission "+ userPerm.toString());
}
d.addColumns(ACL_LIST_FAMILY, key);
try {
  t.delete(d);
} finally {
  t.close();
}
  }

四:hbase权限判断:

回头看AccessController类,

public class AccessController implements MasterObserver, RegionObserver, RegionServerObserver,
  AccessControlService.Interface, CoprocessorService, EndpointObserver, BulkLoadObserver

实现了CoprocessorService、AccessControlService.Interface等,会在scan,delete,merger,balance等操作前后触发操作:
例如:preScannerOpen,其他操作类似
内部调用permissionGranted(opType, user, env, families, Action.READ);检查当前user是否有privilege执行操作;
接着:

//meta表读权限所有用户都将允许
if (hri.isMetaRegion()) {
  if (permRequest == Action.READ) {
return AuthResult.allow(request, "All users allowed", user,
  permRequest, tableName, families);
  }
}
/**如果获取不到用户,权限拒绝;
其中用户没指定情况下三种情况获取:
    1.keberos中获取;
    2.hadoop用户;
    3.操作系统中的用户;
    **/
if (user == null) {
  return AuthResult.deny(request, "No user associated with request!", null,
permRequest, tableName, families);
}

检查是否有权限请求的表及列族;
如果列族存在且所有列族检查通过,则:

// all family checks passed
  return AuthResult.allow(request, "All family checks passed", user, permRequest,
  tableName, families);

否则:

// 4. no families to check and table level access failed
return AuthResult.deny(request, "No families to check and table permission failed",
user, permRequest, tableName, families);

中间权限检查通过TableAuthManager的authorize执行:
接着验证用户与验证用户组:

 if (authorizeUser(user, table, family, qualifier, action)) {
  return true;
}

String[] groups = user.getGroupNames();
if (groups != null) {
  for (String group : groups) {
if (authorizeGroup(group, table, family, qualifier, action)) {
  return true;
}
  }
}

查看验证用户:
核心代码:

// Global and namespace authorizations supercede table level
if (authorize(user, table.getNamespaceAsString(), action)) {
  return true;
}
// Check table permissions
return authorize(getTablePermissions(table).getUser(user.getShortName()), table, family,
qualifier, action);

其中检查namespace(namespace的来源就是zookeeper上注册监听的节点名nodename):权限:
从globalCache获取用户:
globalCache.getUser(user.getShortName()), action)
获取用户组:
globalCache.getGroup(group), action)
接着从nsCache获取namespace权限;
其中注册在ZooKeeperWatcher中节点数据变化(会将数据写到zookeeper中acl节点下,当然会解析acl表的数据成指定格式)的时候会refreshAuthManager,及会执行updateNsCache(namespace, perms);

    /**
    * Updates the internal permissions cache for a single table, splitting
    * the permissions listed into separate caches for users and groups to optimize
     * group lookups.
   */
private void updateNsCache(String namespace,
 ListMultimap<String, TablePermission> tablePerms) {
PermissionCache<TablePermission> newTablePerms = new PermissionCache<>();

for (Map.Entry<String, TablePermission> entry : tablePerms.entries()) {
  if (AuthUtil.isGroupPrincipal(entry.getKey())) {
newTablePerms.putGroup(AuthUtil.getGroupName(entry.getKey()), entry.getValue());
  } else {
newTablePerms.putUser(entry.getKey(), entry.getValue());
  }
}

nsCache.put(namespace, newTablePerms);
mtime.incrementAndGet();
  }

上面是cache的基本流程;
接着会用从cache中获取的权限逐个检查:

 for (Permission p : perms) {
if (p.implies(action)) {
  return true;
}
  }

结束;其中用户组的获取也是一样的道理。
这样就完成了用户操作钱权限的判断,具体请看源码。

五:hbase权限改进思路:

基于更严谨的hbase权限管理,可考虑:

1.HBASE中RWECA可以改善成更细粒度的权限,如增删改查权限;

2.连接hbase权限,用户名user+密码pwd认证;

3.grant授权与revoke解除权限,改进为user+pwd识别用户;

4.严谨的superuser用户定义及权限。

等等权限。

六:hbase权限其他方面:

1:Using Secure HTTP (HTTPS) for the Web UI

2:Using SPNEGO for Kerberos authentication with Web UIs

3:Thrift Gateway

4:REST Gateway

5:HDFS and ZooKeeper

6:Securing Access To Your Data

  • Role-based Access Control (RBAC) controls
  • Visibility Labels
  • Transparent encryption of data at rest
  • Secure Bulk Load
  • Secure Enable

七:参考:

http://hbase.apache.org/book.html#security

上一篇:python定时任务模块-crontab


下一篇:Period [HDU 1358]