I believe the format must have killed it all away. You might be lucky with the SNN location if you were running it, though - could you check in the SNN's directory for valid fsimage?
On Thu, Apr 28, 2011 at 12:53 PM, Adarsh Sharma <[email protected]> wrote: > I correct my posting that this problem arises due to running a script that > internally issue below commands as root user: > >> bin/hadoop namenode -format >>bin/start-all.sh > > No is it possible to start the previous cluster with the previous data or > not. > > Thanks > > > Adarsh Sharma wrote: >> >> Thanks harsh, >> >> My dfs.name.dir is /home/hadoop/project/hadoop-0.20.2/name >> & there is only one file fsimage inimage directory & current directory is >> empty. >> >> But there must be four files fsname, edits, image etc in current >> directory. How they got deleted even I don't issue format command yet. >> >> I think there must be a remedy to get the previous data. >> >> >> >> >> Harsh J wrote: >>> >>> Hello Adarsh, >>> >>> On Thu, Apr 28, 2011 at 11:02 AM, Adarsh Sharma >>> <[email protected]> wrote: >>> >>>> >>>> After correcting my mistake , when I try to run with the hadoop user, my >>>> Namenode fails with the below exception\ : >>>> 2011-04-28 10:53:49,608 ERROR >>>> org.apache.hadoop.hdfs.server.namenode.FSNamesystem: FSNamesystem >>>> initialization failed. >>>> java.io.IOException: NameNode is not formatted. >>>> at >>>> >>>> org.apache.hadoop.hdfs.server.namenode.FSImage.recoverTransitionRead(FSImage.java:317) >>>> >>> >>> Do you still have the valid contents in your good ${dfs.name.dir}? You >>> should be able to recover with that. >>> >>> >> >> > > -- Harsh J
