Hey Eric, It works for hdfs (here are the scripts I used: https://github.com/elicollins/hadoop-dev)
Not long ago it worked for everything, looks like mr was recently broken. I think there's a jira for this. $ jt2 /home/eli/src/hadoop2/mapreduce/bin/mapred: line 22: /home/eli/src/hadoop2/mapreduce/bin/../libexec/mapred-config.sh: No such file or directory Thanks, Eli On Tue, Jun 21, 2011 at 2:41 PM, Eric Caspole <eric.casp...@amd.com> wrote: > Is it still possible to run hadoop directly out of a svn checkout and build > of trunk? A few weeks ago I was using the three variables > HADOOP_HDFS_HOME/HADOOP_**COMMON_HOME/HADOOP_MAPREDUCE_**HOME and it all > worked fine. It seems there has been a lot of changes in the scripts, and I > can't get it to work or figure out what else to set either in the shell env > or at the top of hadoop-env.sh. I have checked out trunk with a dir > structure like this: > > [trunk]$ pwd > /home/ecaspole/views/hadoop/**trunk > [trunk]$ ll > total 12 > drwxrwxr-x. 12 ecaspole ecaspole 4096 Jun 21 15:55 common > drwxrwxr-x. 10 ecaspole ecaspole 4096 Jun 21 13:20 hdfs > drwxrwxr-x. 11 ecaspole ecaspole 4096 Jun 21 16:19 mapreduce > > [ecaspole@wsp133572wss hdfs]$ env | grep HADOOP > HADOOP_HDFS_HOME=/home/**ecaspole/views/hadoop/trunk/**hdfs/ > HADOOP_COMMON_HOME=/home/**ecaspole/views/hadoop/trunk/**common > HADOOP_MAPREDUCE_HOME=/home/**ecaspole/views/hadoop/trunk/**mapreduce/ > > [hdfs]$ ./bin/start-dfs.sh > ./bin/start-dfs.sh: line 54: > /home/ecaspole/views/hadoop/**trunk/common/bin/../bin/hdfs: > No such file or directory > Starting namenodes on [] > localhost: starting namenode, logging to /home/ecaspole/views/hadoop/** > trunk/common/logs/ecaspole/**hadoop-ecaspole-namenode-** > wsp133572wss.amd.com.out > localhost: Hadoop common not found. > localhost: starting datanode, logging to /home/ecaspole/views/hadoop/** > trunk/common/logs/ecaspole/**hadoop-ecaspole-datanode-** > wsp133572wss.amd.com.out > localhost: Hadoop common not found. > Secondary namenodes are not configured. Cannot start secondary namenodes. > > Does anyone else actually run it this way? If so could you show what > variables you set and where so the components can find each other? > > Otherwise, what is the recommended way to run a build of trunk? > Thanks, > Eric > > >
