利用Zookeeper对HDFS中Namenode单点失败的改进方法  

A method to avoid single failure of Namenode in HDFS using Zookeeper

作　　者：鲁阳[1] 郑岩[1] 

机构地区：[1]北京邮电大学计算机学院,北京100876

出　　处：《软件》2012年第12期192-196,共5页Software

摘　　要：在大数据时代分布式处理的已经成为潮流,而Hadoop是一种应用十分广泛的分布式处理框架。但在Hadoop的使用中,Namenode的单点失败问题一直困扰着框架的使用者。本文提出了一种利用Zookeeper对Namenode进行冗余备份协同工作的处理方法,避免了Namenode单点失败造成的服务不可用与文件丢失问题。In the age of big data, it has became a trend to use distributed processing to solve problems, and Hadoop is a widely used framework for distributed processing. But in the use of Hadoop, the failure of Namenode always feezes the users. In this paper, we use an open source framework, Zookeeper, to make Namenode redundant, and avoid the single point failure of Namenode which leads to the unavailable state of HDFS service.

关 键 词：分布式系统 HADOOP HDFS Namenode失败 Zookeeper 

分 类 号：TP338.8[自动化与计算机技术—计算机系统结构]