HI , I didn't try Hadoop rebalancer 。Because I remember rebalancer only considers disk load, and won't consider that data blocks which rack 。 I can try 。Thank you for your reply 。
------------------ 原始邮件 ------------------ 发件人: "Yehia Elshater";<y.z.elsha...@gmail.com>; 发送时间: 2014年7月14日(星期一) 下午4:52 收件人: "user"<user@hadoop.apache.org>; 主题: Re: Block should be additionally replicated on 1 more rack(s) Hi, Did you try Hadoop rebalancer ? http://hadoop.apache.org/docs/r1.0.4/hdfs_user_guide.html#Rebalancer On 14 July 2014 04:10, 风雨无阻 <232341...@qq.com> wrote: HI all: After the cluster configuration rack awareness,run " hadoop fsck / " A lot of the following error occurred: Replica placement policy is violated for blk_-1267324897180563985_11130670. Block should be additionally replicated on 1 more rack(s). Online said "The reason is that three copies on the same rack" . The solution is now: hadoop dfs -setrep 4 /user/hive/warehouse/tbl_add_av_errorlog_android/dt=2013-08-24/000004_0 sleep N hadoop dfs -setrep 3 /user/hive/warehouse/tbl_add_av_errorlog_android/dt=2013-08-24/000004_0 But the speed is very slow。 What is a better good way to make HDFS become healthy? Thanks, Ma Jian