Without a checkpoint of the namenode before the fsck -move you will not be able to recover the files to their original state within HDFS.
-----Original Message----- From: Nikos Papailiou [mailto:np...@cslab.ntua.gr] Sent: Friday, July 13, 2012 7:12 AM To: hdfs-user@hadoop.apache.org Subject: Recover datablocks Hello all, After a disk problem in one of our cluster's datanodes, there were some corrupted files in the local file system. In the begining we did a hadoop fsck -move that moved corrupted files to /lost+found. We have now recovered the corrupted files in the local filesystem(datanode's directory). Is there a way to recover the files from lost+found folder? hadoop version: 0.20.2-cdh3u3 There exists no checkpoint of the namenode before moving the files to /lost+found Thanks, Nikos
