1

I have installed hadoop 1.2.1 on linux with single node cluster configuration. It was running fine and the jps command was displaying the information of all 5 jobs

  • JobTracker
  • NameNode
  • TaskTracker
  • SecondaryNameNode
  • jps
  • DataNode.`

Now, when I start the hadoop using command bin/start-all.sh, hadoop starts all 5 jobs but within few seconds namenode shuts down itself.

Any ideas how can I solve this issue?

I have checked the namenode log file and it shows the following error:

 ERROR org.apache.hadoop.hdfs.server.namenode.NameNode: java.io.IOException: Edit log corruption detected: corruption length = 98362 > toleration length = 0; the corruption is intolerable.
Kevin Brown-Silva
  • 40,873
  • 40
  • 203
  • 237
Maya
  • 161
  • 1
  • 3
  • 8

1 Answers1

1

This is been asked many times and answered as well, searching with the exception message would give you the results.
Before asking questions in Stackoverflow, please check for same kind of question is asked earlier by search option at top right corner.
coming to the problem statement,
It was most probably due to the hadoop.tmp.dir where your namenode stores the edit logs and and check point data.
After every reboot of your machine, tmp folder will be cleared by many services which causing the problem while trying to access by namenode again.
so only the length is 0 after you reboot it.
in core-site.xml change the property hadoop.tmp.dir directory to other directory.
Reference is : here
Hope it helps!

Community
  • 1
  • 1
Mr.Chowdary
  • 3,389
  • 9
  • 42
  • 66
  • I have already tried your suggested solutions. The only thing remaining to try is to format the namenode. I have alot of data on it and I was looking for the solution that can still save my data. – Maya Jan 13 '15 at 15:38
  • You no longer can get the data from the `temp` folder after rebooting the machine.if your checkpoint director is not `temp` then you can get your data back from the fsImage in that folder. – Mr.Chowdary Jan 14 '15 at 02:02