Hello!
I want to know about some implementation issue about retrieving the running instance of namenode or fsnamesystem Because I want to get physical block information especially blockid and its physical path on the HDFS in the level of mapreduce framework, but it seems not possible without connecting namenode via HTTP protocol. For example, the issue of HDFS-4949 is implemented in the level of HDFS, not MapReduce framework, so retrieving block information is quite easy. However, in the level of MapReduce framework, it seems not easy as i expected and I think it is related in the design policy of HDFS and MapReduce. If someone tell me how to get running instance of NameNode in MapReduce framework? If it is not possible, then can you give some advices to make it happen? Thanks!
