I am thinking about that, let datanodes carry additional information about which file that it belong to, sequence number in the order of blocks, and how many blocks of the file.
Is it make sense? 2008/9/9 Raghu Angadi <[EMAIL PROTECTED]> > > There is certainly value in it. But it can not be done with the data > datanodes currently have. > > The usual way to protect NameNode metadata in Hadoop is to write the > metadata to two different locations. > > Raghu. > > > 叶双明 wrote: > >> Hi all. >> >> We all know the importance of NameNode for a cluster, when the NameNode >> whith all matadata of the DFS is breaked down, the whole of the >> clustes'data is gone. >> >> So, if we can retrive DFS metadata from datanodes, It should be a >> additional >> great robustness for hadoop. >> >> Does any one thinking It's valuable for it? >> >> > -- Sorry for my english!! 明
