[ 
https://issues.apache.org/jira/browse/HADOOP-2270?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12545045
 ] 

Karam Singh commented on HADOOP-2270:
-------------------------------------

Please read  dfs.trash.root, dfs.trash.interval as  fs.trash.root, 
fs.trash.interval

> Title: DFS submit client params overrides final params on cluster 
> ------------------------------------------------------------------
>
>                 Key: HADOOP-2270
>                 URL: https://issues.apache.org/jira/browse/HADOOP-2270
>             Project: Hadoop
>          Issue Type: Bug
>          Components: conf
>    Affects Versions: 0.15.1
>            Reporter: Karam Singh
>
> hdfs client params over-rides the params set as final on hdfs cluster nodes. 
> default valuesv of cleint side hadoop-site.xml values override the final 
> prameters of hdfs hadoop-site.xml .
> oberved the following cases -:
> 1. dfs.trash.root=/recycle, dfs.trash.interval=10 and dfs.replication=2 
> marked final under hadoop-site.xml on hdfs cluster.
>    When fsShel command "hadoop dfs -put local_dir dest" fired from submission 
> host
>    Files will still get replicated 3 times (default) instead of final 
> dfs.replication=2.
>    Similarly when "hadoop dfs -rmr dfs_dir OR hadoop dfs -rm file_path " 
> fired from submit client the file/driectory diectly got deleted without being 
> moved to /recycle.
>    Here hadoop-site.xml on submit client does not specify dfs.trash.root, 
> dfs.trash.interval and dfs.replication.
>   
>    Same is the case when we submit mapred JOB from client and job.xml 
> dispalys default values which overrides the lsuter values.
> 2. dfs.trash.root=/recycle, dfs.trash.interval=10 and dfs.replication=2 
> marked final under hadoop-site.xml on hdfs cluster.
>    And 
>    dfs.trash.root=/rubbish, dfs.trash.interval=2 and dfs.replication=5 under 
> hadoop-site.xml on submit client.
>    When fsShel command "hadoop dfs -put local_dir dest" fired from submit 
> client
>    Files will  get replicated 5 times instead of final dfs.replication=2.
>    Similarly when "hadoop dfs -rmr dfs_dir OR hadoop dfs -rm file_path " 
> fired from submit client the file/driectory diectly will be moved to /rubbish 
> instead of /recycle.
>   
>    Same is the case when we submit mapred job from client, job.xml displays 
> following values -:
>    dfs.trash.root=/rubbish, dfs.trash.interval=2 and dfs.replication=5

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

Reply via email to