hdfs-xrootd-fallback-proj
=========================
The HDFS XRootD Fallback system enables relaxed local Hadoop replication by
utilizing global redundancy provided by the XRootD Federation. This system
provides exception handling at the block level, a cache to locally store
repaired blocks retrieved from the Federation, and the ability to re-inject the
repaired blocks back into Hadoop.

The system currently is composed of 3 packages:
hdfs-xrootd-fallback
hdfs-xrootd-fbmon
hdfs-xrootd-healer

Installation instructions can be found here:
http://www.gled.org/cgi-bin/twiki/view/Main/HdfsXrootd


hdfs-xrootd-fallback
====================
The HDFS XRootD Fallback package is installed on every datanode in the Hadoop
cluster and accesses blocks on demand via XRootD Cache on failed read exceptions

hdfs-xrootd-fbmon
=================
The HDFS XRootD Fallback Monitor is a UDP listener that logs incomming messages
sent from the datanodes whenever a fallback is triggered.

hdfs-xrootd-healer
==================
The HDFS XRootD Healer is installed on the XRootD Cache node and periodically
compares corrupt blocks in Hadoop with blocks stored in the cache. It re-injects
the repaired blocks once they are fully cached.