2.2. Ports

This section provides information on the ports used by HDFS services.  

  • HTTP Ports: The NameNode and DataNode services have a web interface and therefore listening on an Hypertext Transfer (HTTP) port. This makes it possible for any client with network access to the nodes to go to the webpage and view specific information on the node regarding the HDFS service running on it. Each of the HDFS services can be configured to listen on a specific port.

    These ports are configured in the hdfs-site.xml file. Use the following table to determine ports for HDFS service, corresponding property in the hdfs-site.xml file, and default value of that port.

    Table 9.1. HDFS HTTP Ports
    HDFS service Configuration property in hdfs-site.xml file Default value
    NameNode <name>dfs.http.address</name> 50070
    DataNode <name>dfs.datanode.http.address</name> 50075

  • IPC Ports: Interprocess Communication (IPC) is the communication used between the HDFS services. IPC is a client server architecture. The following table lists the ports that the NameNode and DataNode use for Remote Procedure Call (RPC) calls.

    Table 9.2. HDFS IPC Ports
    HDFS service Configuration property Default value
    NameNode <name>fs.default.name</name> in the core-site.xml hdfs://hostname: 8020
    DataNode <name>dfs.datanode.address</name> in the hdfs-site.xml 50010
    DataNode <name>dfs.datanode.ipc.address</name> in the hdfs-site.xml 8010

    DataNode uses two different IPC ports. Port 50010 is for data transfer. When a client tries to get or put a file into HDFS the file stream transfer is completed through this port.

    HBase uses port 8010 port for short circuit feature. The 8010 port lets a dfsclient (located on the same machine as the particular block) access that file directly after making the request on the 8010 port of DataNode to release any holds on the block.

    Additionally, this port is also used for DataNodes to communicate with each other when needed.


loading table of contents...