If you have decided to take the 'fsck -move' route, then you need to exit safemode first:
# Exit safemode first. After this, NN will not prevent writes/updates. sudo -u hdfs hadoop dfsadmin -safemode leave # Move the files to /lost+found (newdir) next. sudo -u hdfs hadoop fsck -move On Wed, May 16, 2012 at 1:48 AM, Terry Healy <the...@bnl.gov> wrote: > Running 1.0.2 on cluster with 10 datanodes. After running stop-all.sh > and start-all.sh following the addition of a new datanode, the NN stays > in SafeMode. > > hadoop fsck / reports several MISSING and CORRUPT blocks, but I have > not been able to continue after trying both the -move and -delete > options. When I run the -move option, it reports: > > org.apache.hadoop.ipc.RemoteException: > org.apache.hadoop.hdfs.server.namenode.SafeModeException: Cannot create > directory /lost+found. Name node is in safe mode. > > In what directory is it trying to create the /lost+found? > > What can I do to purge the errors and get HDFS running again? > > Thanks, > > Terry -- Harsh J
