Python 通过thrift接口连接Hbase读取存储数据
来源：互联网 发布：数据挖掘导论英文版 编辑：程序博客网 时间：2024/05/29 11:43
介绍：
Hbase：开源的分布式数据库
资料介绍:http://www.oschina.net/p/hbase
Thrift:一个软件框架，用来进行可扩展且跨语言的服务的开发。最初由Facebook开发，作为Hadoop的一个工具，提供跨语言服务开发；
资料介绍：http://dongxicheng.org/search-engine/thrift-guide/
官方使用手册：http://download.csdn.net/detail/wyjzt999/5141006从安装到使用都很全面
我们项目里客户端是用python开发，因此需要Thrift提供server端，经过thrift对Hbase进行数据读写操作，性能非常不错，并且可以在Hadoop集群上做并行拓展，稳定性高，Facebook内部通信也是采用thrift来做；
首先学习一下Hbase的表结构：
Row Key
Row key行键 (Row key)可以是任意字符串(最大长度是 64KB，实际应用中长度一般为 10-100bytes)，在hbase内部，row key保存为字节数组。
列族 (column family)
hbase表中的每个列，都归属与某个列族。列族是表的chema的一部分(而列不是)，必须在使用表之前定义。列名都以列族作为前缀。例如courses:history ， courses:math 都属于 courses 这个列族。
时间戳
HBase中通过row和columns确定的为一个存贮单元称为cell。每个 cell都保存着同一份数据的多个版本。版本通过时间戳来索引。时间戳的类型是 64位整型。时间戳可以由hbase(在数据写入时自动 )赋值，此时时间戳是精确到毫秒的当前系统时间。时间戳也可以由客户显式赋值。如果应用程序要避免数据版本冲突，就必须自己生成具有唯一性的时间戳。每个 cell中，不同版本的数据按照时间倒序排序，即最新的数据排在最前面。
为了避免数据存在过多版本造成的的管理 (包括存贮和索引)负担，hbase提供了两种数据版本回收方式。一是保存数据的最后n个版本，二是保存最近一段时间内的版本（比如最近七天）。用户可以针对每个列族进行设置。
对Hbase而言，表结构设计会对系统的性能以及开销上造成很大的区别;
1.首先建立与thriftserver端的连接
from thrift import Thriftfrom thrift.transport import TSocket, TTransportfrom thrift.protocol import TBinaryProtocolfrom hbase import Hbase#server端地址和端口transport = TSocket.TSocket(host, port)#可以设置超时transport.setTimeout(5000)#设置传输方式（TFramedTransport或TBufferedTransport）trans = TTransport.TBufferedTransport(transport)#设置传输协议protocol = TBinaryProtocol.TBinaryProtocol(trans)#确定客户端client = Hbase.Client(protocol)#打开连接transport.open()
2.然后就可以做具体的操作，比如查表，删表，插入Row Key等等
from hbase.ttypes import ColumnDescriptor, Mutation, BatchMutation, TRegionInfofrom hbase.ttypes import IOError, AlreadyExists#获取表名client.getTableNames()#创建新表_TABLE = "keyword"
demo = ColumnDescriptor(name='data:',maxVersions = 10)#列族data能保留最近的10个数据，每个列名后面要跟:号createTable(_TABLE, [demo])
#创建列名2个data:url data:word tmp1= [Mutation(column="data:url", value="www.baidu.com")]tmp2= [Mutation(column="data:word", value="YaGer")]#新建2个列 (表名，行键， 列名)client.mutateRow(_TABLE, row, tmp1)client.mutateRow(_TABLE, row, tmp1)
#从表中取数据#通过最大版本数取数据client.getByMaxver(_TABLE,'00001','data:word', 10)#一次取10个版本#取列族内数据client.getColumns(_TABLE, '00001')
3.支持thrift并行拓展和失效转移Failover机制
#file name：hbaseconn.py#! /usr/bin/env python # -*- coding: utf-8 -*-#提供建立连接的方法和取数据操作的方法import loggingimport tracebackimport time,sysfrom unittest import TestCase, mainimport socketfrom thrift import Thrift from thrift.transport import TSocket from thrift.transport import TTransport from thrift.protocol import TBinaryProtocol from hbase import Hbase from hbase.ttypes import IOError as HbaseIOError, AlreadyExistsfrom hbase.ttypes import ColumnDescriptor, Mutation, BatchMutation_TABLE = 'thrift_check'_ROW = 'test_for_thrift'_DATA = 'data:word'_VALUE = 'Flag'class DbConnError(Exception): """Conn Db exception. Timeout or any except for connection from low layer api """ passclass Connection: def __init__(self, trans, client, addr, port): self.trans = trans self.client = client self.hp = addr self.port = port passclass CenterDb: @classmethod def open(cls, host_port): cls.tc_list = [] for hp in host_port: trans, client = cls._open(*hp) cls.tc_list.append(Connection(trans, client, hp[0], hp[1])) return cls.tc_list @classmethod def _open(cls, host, port): transport = TSocket.TSocket(host, port) transport.setTimeout(5000) trans = TTransport.TBufferedTransport(transport) protocol = TBinaryProtocol.TBinaryProtocol(trans) client = Hbase.Client(protocol) ok = False try: trans.open() ok = True except TSocket.TTransportException, e: logerr('CenterDb(Hbase) Open error(%s, %d)' % (host, port)) ok = False else: pass# dbg('CenterDb connected (%s, %d)' % (host, port)) return trans, client @classmethod def _initTable(cls, client): dat = ColumnDescriptor(name = 'data', maxVersions = 1) tmp = [Mutation(column = _DATA, value = _VALUE)] try: client.createTable(_TABLE, [dat]) client.mutateRow(_TABLE, _ROW, tmp) dbg("Create Table For Thrift Test Success!") return True except AlreadyExists: return True return False @classmethod def _get(cls, client): client.getVer(_TABLE,_ROW,_DATA, 1) return True @classmethod def _reconnect(cls, trans): trans.close() trans.open() def __init__(self, transport, client): self.t = transport self.c = client def __str__(self): return 'CenterDb.open(%s, %d)' % (self.t, self.c) def __del__(self): self.t.close() def getColumns(self, table, row): tr_list = [] tr_list = self._failover('getRow', table, row) if (not tr_list): return {} return tr_list[0].columns #file name: thriftmanage.py#! /usr/bin/python# -*- coding: utf-8 -*-#提供管理thrift连接的方法#建一个线程循环检测thrift的可使用性，非连接池import loggingimport time,sys,randomimport threadingfrom hbase import Hbase from hbase.ttypes import IOError as HbaseIOErrorfrom hbase.ttypes import ColumnDescriptor, Mutation, BatchMutationimport hbaseconnfrom hbaseconn import CenterDbclass Failover(threading.Thread): def __init__(self,serverid, serverlist): threading.Thread.__init__(self) self.serverlist = serverlist self.serverid = serverid self.lenlist = len(serverlist) self.conn_list = [] self.invalid_con_list = [] ''' cur_conn : now oper is using my_conn : this schd server should use [(trans1,client1),(trans2,client2),...] ''' self.cur_conn = None self.my_conn = None self._makeConn() self._getConn() self._init_for_check() self.switched_flag = False def _makeConn(self): '''make all the conntion to thrift serverlist ''' self.conn_list = CenterDb.open(self.serverlist) def _getConn(self): '''get only one connection from the conn_list,confirm cur_conn and my_conn ''' self.cur_conn = self.my_conn = self.conn_list[(int(self.serverid) % self.lenlist)] self.other_conn = self.conn_list[:] self.other_conn.remove(self.my_conn) return True def _init_for_check(self): '''init _TABLLE test_for_flag ''' try: if not CenterDb._initTable( self.my_conn.client ): dbg("Error In Create Table For Thrift Test!") except Exception, e: dbg("init_for_check thrift:%s" % e) '''make the only conn for check proc ''' self.check_conn = CenterDb.open( ((self.my_conn.hp, self.my_conn.port),))[0] def _switch(self): '''when my_conn failed, choose the other client randomly; when my_conn is reset OK, cur_conn will use my_conn again ''' #print 'Schd%s come in _switch' % self.serverid if 0 == len(self.other_conn): return False trycount = 0 while True: try: if trycount == 3*self.lenlist :#try 3*length times return False tmp_conn = random.choice(self.other_conn) #CenterDb._reconnect(tmp_conn.trans) #DEBUG if self._checker(tmp_conn): self.cur_conn = tmp_conn dbg('Schd%s _switch cur_conn: %s' % (self.serverid, self.cur_conn.hp)) return True else: trycount += 1 logerr('Schd%s _switch for %d times' % (self.serverid, trycount)) CenterDb._reconnect(tmp_conn.trans)#close this trans and try again;breakdown early if self._checker(tmp_conn): self.cur_conn = tmp_conn return True else: continue#can't be used except Exception, e: continue def _failover(self,oper, *args, **kwargs): result = [] try : result = getattr(self.cur_conn.client, oper)(*args, **kwargs) return result except HbaseIOError, e: logerr("_failover : %s " % e) except Exception, e: logerr( 'Schd%s _failover : Connect to %s thrift Server closed! Choose another......Reason:%s ' % \ (self.serverid, self.cur_conn.hp, e)) self.cur_conn.trans.close() self.switched_flag = True if self._switch(): logerr('Schd%s _failover : Now using %s server' % (self.serverid, self.cur_conn.hp)) result = getattr(self.cur_conn.client, oper)(*args, **kwargs) else: logerr( 'Schd%s _failover : Switch 3 rotate Find No Healthy Thrift server !' % self.serverid) return result def getByMaxver(self, table, row, column, max_ver): """Get cell list by ver Get cell list no more than max versions Args: table: table name row: row key column: column name max_ver: max version to retrieve cell """ cell_list = [] start_time = time.time() cell_list = self._failover('getVer', table, row, column, max_ver) if (not cell_list): return [] take = time.time() - start_time if take > 0.015: logerr('Hbase over 15ms:take %s ms' % str("%.3f" % (take*1000))) return map(lambda x: (x.value, x.timestamp), cell_list) def getColumns(self, table, row): tr_list = [] tr_list = self._failover('getRow', table, row) if (not tr_list): return {} return tr_list[0].columns def _checker(self,conn): '''check my_conn , be sure it is connected and get data OK ''' try: if conn.trans.isOpen(): CenterDb._get(conn.client) return True else: return False except Exception, e: logerr( 'Schd%s _checker : Connect to %s closed! Please Restart now..... Reason: %s ' % \ (self.serverid, conn.hp, e)) return False def _restart(self): '''if my_conn failed, restart it's trans ''' while True: #print 'come in _restart %s ' % self.my_conn.hp self.my_conn.trans.close() self.check_conn.trans.close() try: time.sleep(2) self.my_conn.trans.open() self.check_conn.trans.open() if self.my_conn.trans.isOpen() and self.check_conn.trans.isOpen(): CenterDb._get(self.check_conn.client) #self.check_conn.client.getVer('keywordurl','test_for_thrift','data:word', 1) return True else: continue except Exception, e: logerr('Schd%s _restart : Connect to %s is not restart yet ... Reason:%s ' % \ (self.serverid, self.my_conn.hp, e)) self.my_conn.trans.close() self.check_conn.trans.close() continue def run(self): while True: time.sleep(1) if self._checker(self.check_conn) and (not self.switched_flag): continue else: if self._restart(): logerr( 'Schd%s Connection from %s to my:%s Recovered ! ' % \ (self.serverid,self.cur_conn.hp,self.my_conn.hp)) self.cur_conn = self.my_conn self.switched_flag = False continue使用thrift生成的代码中提供的方法有：
void enableTable(Bytes tableName)
void disableTable(Bytes tableName)
bool isTableEnabled(Bytes tableName)
void compact(Bytes tableNameOrRegionName)
void majorCompact(Bytes tableNameOrRegionName)
getTableNames()
getColumnDescriptors(Text tableName)
getTableRegions(Text tableName)
void createTable(Text tableName, columnFamilies)
void deleteTable(Text tableName)
get(Text tableName, Text row, Text column)
getVer(Text tableName, Text row, Text column, i32 numVersions)
getVerTs(Text tableName, Text row, Text column, i64 timestamp, i32 numVersions)
getRow(Text tableName, Text row)
getRowWithColumns(Text tableName, Text row, columns)
getRowTs(Text tableName, Text row, i64 timestamp)
getRowWithColumnsTs(Text tableName, Text row, columns, i64 timestamp)
getRows(Text tableName, rows)
getRowsWithColumns(Text tableName, rows, columns)
getRowsTs(Text tableName, rows, i64 timestamp)
getRowsWithColumnsTs(Text tableName, rows, columns, i64 timestamp)
void mutateRow(Text tableName, Text row, mutations)
void mutateRowTs(Text tableName, Text row, mutations, i64 timestamp)
void mutateRows(Text tableName, rowBatches)
void mutateRowsTs(Text tableName, rowBatches, i64 timestamp)
i64 atomicIncrement(Text tableName, Text row, Text column, i64 value)
void deleteAll(Text tableName, Text row, Text column)
void deleteAllTs(Text tableName, Text row, Text column, i64 timestamp)
void deleteAllRow(Text tableName, Text row)
void deleteAllRowTs(Text tableName, Text row, i64 timestamp)
ScannerID scannerOpenWithScan(Text tableName, TScan scan)
ScannerID scannerOpen(Text tableName, Text startRow, columns)
ScannerID scannerOpenWithStop(Text tableName, Text startRow, Text stopRow, columns)
ScannerID scannerOpenWithPrefix(Text tableName, Text startAndPrefix, columns)
ScannerID scannerOpenTs(Text tableName, Text startRow, columns, i64 timestamp)
ScannerID scannerOpenWithStopTs(Text tableName, Text startRow, Text stopRow, columns, i64 timestamp)
scannerGet(ScannerID id)
scannerGetList(ScannerID id, i32 nbRows)
void scannerClose(ScannerID id)
参考：http://yannramin.com/2008/07/19/using-facebook-thrift-with-python-and-hbase/
- Python 通过thrift接口连接Hbase读取存储数据
- python通过thrift连接Hbase
- 通过Python使用HBase Thrift接口 url
- [原创]Python通过Thrift连接HBase
- python利用thrift服务读取hbase数据
- java 通过thrift-0.9.1读取hbase表数据
- 使用Python通过Thrift接口访问HBase和Hive
- Python通过thrift连接Hbase（详细配置步骤）
- python通过thrift操作hbase
- pyspark Python 连接 HBase thrift
- Thrift安装配置及使用python通过thrift连接HBase测试
- C#通过thrift连接hbase操作步骤
- python 通过thrift 简单操作hbase
- python 通过thrift 简单操作hbase
- php/perl/python , 通过thrift 连接 hbase，进行条件过滤选择
- Python 使用thrift连接hbase || 远程连接hbase
- 利用ruby通过thrift接口批量插入hbase
- python通过thrift方式连接hive
- Openswan系列教程2：第一阶段数据包简析
- systemctl命令用法
- AIX 上添加ASM磁盘组遇 ORA-15075 disk(s) are not visible cluster-wide 错误
- 结构体(含位域)的sizeof
- Oracle_DAY01
- Python 通过thrift接口连接Hbase读取存储数据
- mysql中插入数据，跳过unique字段相同的数据
- Service 概述
- c语言指针的初始化赋值
- 联系人添加界面并添加相应数据
- 如何在互联网中找到利润高的项目
- JAVASCRIPT学习笔记(1)
- android异常处理
- Rose 2003 破解版在Win7上的步骤详解
