David， You should set dfs.data.dir and dfs.name.dir for HDFS. These two directories is for the real data and meta data of file system.
Jeff Zhang On Mon, Dec 14, 2009 at 4:38 PM, David Stemmer <[email protected]>wrote: > > I'm trying to set up a basic Hadoop single-node cluster on a Red Hat > Enterprise Linux 5 system. The Hadoop version is 0.18.3-14.cloudera.CH0_3. > Unfortunately I am having problems getting Hadoop to read the configuration > properties from hadoop-site.xml. Specifically, when I try to run the > following command: > > ${HADOOP_HOME}/bin/hadoop namenode -format > > ...it seems that Hadoop is not using the storage directory that I have > configured for HDFS. I get the following output: > > --------------------------------------------------------------------- > --------------------------------------------------------------------- > > 09/12/14 02:24:14 INFO dfs.NameNode: STARTUP_MSG: > /************************************************************ > STARTUP_MSG: Starting NameNode > STARTUP_MSG: host = #####/###.##.#.## > STARTUP_MSG: args = [-format] > STARTUP_MSG: version = 0.18.3-14.cloudera.CH0_3 > STARTUP_MSG: build = -r HEAD; compiled by 'root' on Mon Jul 6 15:02:31 > EDT 2009 > ************************************************************/ > Re-format filesystem in /tmp/hadoop-myusername/dfs/name ? (Y or N) Y > 09/12/14 02:24:17 INFO fs.FSNamesystem: fsOwner=myusername,ugrad > 09/12/14 02:24:17 INFO fs.FSNamesystem: supergroup=supergroup > 09/12/14 02:24:17 INFO fs.FSNamesystem: isPermissionEnabled=true > 09/12/14 02:24:17 INFO dfs.Storage: Image file of size 82 saved in 0 > seconds. > 09/12/14 02:24:17 INFO dfs.Storage: Storage directory > /tmp/hadoop-myusername/dfs/name has been successfully formatted. > 09/12/14 02:24:17 INFO dfs.NameNode: SHUTDOWN_MSG: > /************************************************************ > SHUTDOWN_MSG: Shutting down NameNode at #####/###.##.#.## > ************************************************************/ > > --------------------------------------------------------------------- > > I do not want it to format /tmp/hadoop-myusername/dfs/name as the storage > directory. Here is what is in my hadoop-site.xml file. Notice that > hadoop.tmp.dir is set /home/u/fall06/myusername/hadoop_tmp: > > --------------------------------------------------------------------- > --------------------------------------------------------------------- > > <?xml version="1.0"?> > <?xml-stylesheet type="text/xsl" href="configuration.xsl"?> > > <!-- Put site-specific property overrides in this file. --> > > <configuration> > > <property> > <name>hadoop.tmp.dir</name> > <value>/home/u/fall06/myusername/hadoop_tmp</value> > </property> > > <property> > <name>fs.default.name</name> > <value>hdfs://localhost:50031</value> > </property> > > <property> > <name>mapred.job.tracker</name> > <value>localhost:50032</value> > </property> > > <property> > <name>dfs.replication</name> > <value>1</value> > </property> > </configuration> > > --------------------------------------------------------------------- > > > Here are the relevent lines of my .bash_profile where I set some of the > Hadoop environment variables: > > --------------------------------------------------------------------- > --------------------------------------------------------------------- > > export HADOOP_HOME=/home/u/fall06/myusername/Desktop/hadoop_project/hadoop > export > > HADOOP_CONF_DIR=/home/u/fall06/myusername/Desktop/hadoop_project/hadoop/conf > > --------------------------------------------------------------------- > > A few more details on the system setup: > > - I did not install Hadoop on the system myself; it was installed by a > system administrator. > - Because the default Hadoop directory is read/execute only, I copied the > Hadoop directory to a folder that I own and ran in from there. My > HADOOP_HOME enviromnent variable is set to this duplicate directory. > - hadoop_site.xml is in my ${HADOOP_HOME}/conf directory. > - I did try to echo what ${HADOOP_CONF_DIR} was being set to while > '{HADOOP_HOME}/bin/hadoop namenode -format' was running. The result was > this: > > --------------------------------------------------------------------- > --------------------------------------------------------------------- > /home/u/fall06/myusername/Desktop/hadoop_project/conf > --------------------------------------------------------------------- > > That's what I would expect, so I'm stumped. Does anyone have any ideas what > I might be doing wrong, or know of more tests I can run to figure out the > problem? Any and all advice is appreciated. > -- > View this message in context: > http://old.nabble.com/0.18.3-14.cloudera.CH0_3-RHEL5%3A-hadoop-site.xml-properties-not-being-read-when-formatting-the-Hadoop-namenode-tp26774461p26774461.html > Sent from the Hadoop core-user mailing list archive at Nabble.com. > >
