Monday 20 April 2015


Everytime, you format the namenode it generates its namespace id.  Now, if datanodes didn't get the correct name space id, then we may end of in error. This is because as namespace id changed, datanodes are not able to connect to namenode.

The situation is, when we run "jps" command on datanode, we will be able to find "TaskTracker" but no "Datanode".

Oh, we broke our cluster. Now, here is the solution to bring it back online.

Solution 1:

grep for the logs in data node, you will be able to find the below error message.

Now, copy the new space id, as highlighted in the above snapshot.

Navigate to the "VERSION" file in location:

Change the "namespaceID" parameter  to "565898590" as below:

This has to be done on all DATA-NODES in your cluster.  Then start your cluster.

Solution 2:

If the cluster is relatively new. The quick solution is to remove the directory
"HADOOP_PREFIX/tmp" and restart the cluster.

Kool :) 

No comments:

Post a Comment

Note: only a member of this blog may post a comment.