hadoop集群dfs.data.dir目录权限问题导致节点无法启动

 

安装使用Hive的时候发现文件无法从本地拷贝到hadoop集群系统，老是报错，错误内容为：

java.io.IOException: File  … could only be replicated to 0 nodes, instead of 1。查找namenode的日志，也是报这个错，而且非常多，即便在启动的时候也是报类似的错误。这个学习环境为单namenode+2个datanode，使用dfsadmin -report报无数据节点，如下：

[hadoop@namenode hadoop]$ hadoop dfsadmin -report   Configured Capacity: 0 (0 KB) Present Capacity: 0 (0 KB) DFS Remaining: 0 (0 KB) DFS Used: 0 (0 KB) DFS Used%: ?% Under replicated blocks: 0 Blocks with corrupt replicas: 0 Missing blocks: 0 ————————————————- Datanodes available: 0 (0 total, 0 dead)

 

 

 

重新格式化namenode，并重新启动，故障依旧，清除历史日志，从datanode节点的日志发现了错误“Incompatible namespaceIDs “，删除了datanode的dfs.data.dir目录，并重建，然后重新格式化namenode、启动，发现dfsadmin -report还是没有数据节点加入。再查日志，发现错误不同了：All directories in dfs.data.dir are invalid。没有生成目录结构，正纳闷呢，发现日志中前面有警告：

Invalid directory in dfs.data.dir: Incorrect permissio
n for /hadoop/hadoop-data, expected: rwxr-xr-x, while actual: rwxrwxr-x。

    根据警告内容，分别在两个datanode上修改该目录权限：

[hadoop@namenode logs]$ ssh datanode01.hadoop Last login: Wed Mar 14 01:58:39 2012 from namenode.hadoop [hadoop@datanode01 ~]$ chmod g-w /hadoop/hadoop-data/ [hadoop@datanode01 ~]$ exit [hadoop@namenode logs]$ ssh datanode02.hadoop Last login: Wed Mar 14 01:59:00 2012 from datanode01.hadoop [hadoop@datanode02 ~]$ chmod g-w /hadoop/hadoop-data/ [hadoop@datanode02 ~]$ exit

启动后，故障解除：

[hadoop@namenode hadoop]$ hadoop dfsadmin -report Configured Capacity: 158030774272 (147.18 GB) Present Capacity: 141718949918 (131.99 GB) DFS Remaining: 141718892544 (131.99 GB) DFS Used: 57374 (56.03 KB) DFS Used%: 0% Under replicated blocks: 0 Blocks with corrupt replicas: 0 Missing blocks: 0 ————————————————- Datanodes available: 2 (2 total, 0 dead) Name: 172.21.126.102:50010 Decommission Status : Normal Configured Capacity: 79015387136 (73.59 GB) DFS Used: 28687 (28.01 KB) Non DFS Used: 8155709425 (7.6 GB) DFS Remaining: 70859649024(65.99 GB) DFS Used%: 0% DFS Remaining%: 89.68% Last contact: Wed Mar 14 01:40:41 CST 2012 Name: 172.21.126.103:50010 Decommission Status : Normal Configured Capacity: 79015387136 (73.59 GB) DFS Used: 28687 (28.01 KB) Non DFS Used: 8156114929 (7.6 GB) DFS Remaining: 70859243520(65.99 GB) DFS Used%: 0% DFS Remaining%: 89.68% Last contact: Wed Mar 14 01:40:41 CST 2012

 

运行hive测试，也能正常使用了：

hive> create table dummy(value STRING); OK Time taken: 10.642 seconds hive> load data local inpath ‘/tmp/dummy.txt’ overwrite into table dummy; Copying data from file:/tmp/dummy.txt Copying file: file:/tmp/dummy.txt Loading data to table default.dummy Deleted hdfs://namenode.hadoop/user/hive/warehouse/dummy OK Time taken: 0.439 seconds hive> select * from dummy; OK X Time taken: 0.254 seconds hive>

 

没想到数据目录的权限检查如此严格，多了用户组的写权限能造成集群系统的无法启动，在这记录，以备查。

 

转自：http://hellodatabase.com/?p=345