Uploaded image for project: 'Hadoop Common'
  1. Hadoop Common
  2. HADOOP-6469

Multiple RPC Servers using same object instance

VotersWatch issueWatchersCreate sub-taskLinkCloneUpdate Comment AuthorReplace String in CommentUpdate Comment VisibilityDelete Comments


    • New Feature
    • Status: Resolved
    • Major
    • Resolution: Not A Problem
    • None
    • None
    • None
    • None


      In HDFS-599 we want to be able to run multiple RPC servers on the namenode. One for DataNodes another for the rest of the protocols.
      The problem right now is that checking if the protocol is exposed is done on the handler side (the NameNode in our case).

      In our usecase we want to be able to tell if the request is coming on the "datanode port" or "the rest port" and based on that throw an exception when datanodes are accessing the client port and vice versa. Since the JobTracker has the same abstraction of Clients and Slaves those might later also be split on to different servers. So creating a solution just inside of the NameNode is not really an options, especially since we could not come up with one that would be clean and safe.

      The proposal is to have RPC servers have "tags" associated with them in these cases. Then in the NameNode 2 servers created will have tags "DataNodeServer" and "ClientAdminServer". Another change is instead of using VersionedProtocol to check the version and the protocol in the getProtocolVersion method, Classes that can be used in multiple contexts should implement another interface (MultiServer?) that will have a checkProtocol? method that receives the name of the protocol and the Tag associated with the Server that received the request. This way in the checkProtocol we can specify which Protocols are allowed for servers tagged with Tag.

      What do people think?


        Issue Links


          This comment will be Viewable by All Users Viewable by All Users


            dms Dmytro Molkov
            dms Dmytro Molkov
            0 Vote for this issue
            13 Start watching this issue



              Issue deployment