Hello! You are right Koji. The configuration should be done on each slave (datanode), or on each machine from where you run the "rmr" or "rm" command. Regards, Florin
--- On Thu, 6/9/11, Koji Noguchi <knogu...@yahoo-inc.com> wrote: From: Koji Noguchi <knogu...@yahoo-inc.com> Subject: Re: When rmr and rm strike To: "hdfs-user@hadoop.apache.org" <hdfs-user@hadoop.apache.org> Date: Thursday, June 9, 2011, 1:20 PM Re: When rmr and rm strike > What i have found is that the setting needs to be on the client side > configuration and not necessarily required on the namenode. > As long as the hadoop configuration on the node from which you run the -rmr > has this setting, the files will be moved into the .Trash directory. > Can anyone else confirm this behaviour? > Move to Trash is a complete client-side behavior. So, config on the client side is sufficient. However, purging the Trash periodically is done by the namenode. So unless namenode has the same config setup, those Trash remains forever. Koji On 6/9/11 7:52 AM, "sridhar basam" <s...@basam.org> wrote: On Thu, Jun 9, 2011 at 10:42 AM, Florin P <florinp...@yahoo.com> wrote: Hello! Your answers solved my problem. Below are the steps: 1. In the core-site.xml of the job-tracker add the following entry: <property> <name>fs.trash.interval</name> <value>1440</value> </property> 2. Shutdown job-tracker (job tracker machine hadoop/bin/stop-mapred.sh) 3. Shutdown namenode (namenode machine hadoop/bin/stop-dfs.sh) 4. Start namenode (namenode machine hadoop/bin/start-dfs.sh) 5. Start jobtracker (job tracker machine hadoop/bin/start-mapred.sh) What i have found is that the setting needs to be on the client side configuration and not necessarily required on the namenode. As long as the hadoop configuration on the node from which you run the -rmr has this setting, the files will be moved into the .Trash directory. Can anyone else confirm this behaviour? Observation. I've added the property in the core-site.xml of the namenode, but no effect. It worked only when I've added in the core-site.xml of the jobtracker. Were you by chance doing the -rmr on the jobtracker? Sridhar