On May 19, 2008, at 1:37 AM, Fabrizio detto Mario wrote:
How does Hadoop manage the failure of the JobTracker (Master Node)?
For example, Google Map/Reduce version "aborts the MapReduce
computation if
the master fails".
As NameNode and SecondaryNameNode, exists a SecondaryJobTracker?
Currently, no.
However, one can use Hadoop-On-Demand (under src/contrib/hod - http://
hadoop.apache.org/core/docs/current/hod.html) to dynamically
provision a Map-Reduce cluster (usually per-job or per-set-of-jobs)
to minimize your failure granularity.
Arun