AnsweredAssumed Answered

Questions on Namespace->Container Mapping

Question asked by rini on Sep 23, 2012
Latest reply on Sep 23, 2012 by Ted Dunning
Hi All,

Few questions:

1) Where is the namespace --> container mapping maintained? It seems that CLDB only maintains the container --> server mapping. If I am a client and I want to read a file /user/me/foo.txt, how do I figure out the appropriate containers where this file's chunks are located? What will be the control flow?

2) If I have a directory /user/me/ and I create a file foo1 on node1, foo2 on node2, ..., fooN on nodeN, will the metadata for the directory /user/me be present on all the nodes where either a block or a replica of files foo1, foo2, ..., fooN resides? If yes, how do you manage file path traversal for security checks and consistency across name space additions/deletions in such a split scenario?

3) Am I right in understanding that MapR doesn't use local file sytem such as ext3 on the data nodes unlike Hadoop?

4) Any fragmentation issues with file/directory deletions in the containers?