Progress
Installation and Configuration Guide
Version 9
for UNIX


Using NameServer Replication

UDP broadcasting supports NameServer replication by allowing a client or Unified Broker request to be received by multiple NameServers listening on the same UDP port and configured on different machines within the same subnet. Because every host on a subnet receives every broadcast request, one or more of these hosts can support a NameServer that receives and handles the same messages. This provides fault tolerance for both a client connection request and a Unified Broker registration request.

Thus, to configure and use replicated NameServers, you must:

There is one broadcast address for each subnet. Using this address and the specified UDP port number, a client or Unified Broker sends a single request that is recognized by every NameServer listening on that port in the subnet.

Figure 8–6 shows a client, a Unified Broker, and two replicated NameServers. The NameServer configurations shown for NameServer NS1 (above the dotted line) appear as they might in the ubroker.properties file for each host.

Figure 8–6: NameServer Replication

In Figure 8–6, one NameServer is located on a machine with the IP address 172.20.0.7 and another is located on a machine with the IP address 172.20.16.12. Both NameServers listen on UDP port 5162. The UDP broadcast address for these NameServers is 172.20.255.255. The Unified Broker is configured to register with a controlling NameServer remote from the Unified Broker machine using the UDP broadcast address 172.20.255.255 as the hostName. When the Unified Broker registers with its controlling NameServer using the UDP broadcast, it registers with both replicated NameServers. Similarly, when the client broadcasts its connection request using 172.20.255.255 as the NameServer host name, both replicated NameServers receive the request. The client uses the Unified Broker connection returned by the first NameServer that responds.

Note that if the NameServer at IP address 172.20.0.7 moves to a different host on the subnet, for example, with IP address 172.20.16.5, neither the client application nor the Unified Broker configuration has to change.

Implementing NameServer Replication

To configure and use NameServer replication, follow these general steps:

  1. Install the NameServer on each host within a single subnet where you want to replicate a NameServer configuration.
  2. Configure each replicated NameServer to listen on the same UDP port number.
  3. Determine the UDP broadcast address for the subnet where the NameServer hosts reside. For more information, see the "Determining the Broadcast Address" section.
  4. Configure each Unified Broker instance (AppServer, WebSpeed Transaction Server, or DataServer) to use a controlling NameServer as follows:
    • Location — Remote
    • Host name — The UDP broadcast address that you determined in Step 3
    • Port number — The UDP port number that you specified in Step 2
  5. Provide connection parameters to the client (AppServer, DataServer, or WebSpeed) that specify the required Application Service name, the broadcast address in Step 3, and the UDP port number that you specified in Step 2.
Determining the Broadcast Address

You can determine the broadcast address of a UNIX machine by using the netstat and ifconfig commands, as in this example:

$ netstat -i 
Name  Mtu  Net/Dest   Address    Ipkts   Ierrs Opkts  Oerrs Collis Queue 
lo0   8232 loopback    localhost   771334  0     771334 0    0     0  
le0   1500 bali        bali       15069970 286170 10019158 1     302211 0 
$ ifconfig le0
le0: flags=863<UP,BROADCAST,NOTRAILERS,RUNNING,MULTICAST> mtu 1500
       inet 172.20.0.7 netmask ffff0000 broadcast 172.20.255.255 

This shows that the IP address for bali is 172.20.0.7, and its broadcast address is 172.20.255.255.

On NT, follow these steps to determine the broadcast address:

  1. Enter the ipconfig command in the console, as shown in this example:
  2. C:\>ipconfig
    
    Windows NT IP Configuration
    
    Ethernet adapter CE2XPS1:
    
    IP Address. . . . . . . : 172.18.103.44
    Subnet Mask . . . . . . : 255.255.0.0
    Default Gateway . . . . . : 172.18.0.19 
    

  3. For each bit in the Subnet Mask that has a value of 0, convert the corresponding bit in the IP Address to 1.
  4. Note that the IP Address and Subnet Mask are composed of four dot-separated decimal numbers, and each decimal number represents an 8-bit binary number. Also note that the decimal number 255 is 11111111 in binary.

    In this example, the last two decimal digits of the Subnet Mask are zeros. Since the corresponding bits in the IP Address must be converted to 1, the last two decimal numbers of the IP Address should be 255. Therefore the broadcast address is 172.18.255.255. (For more information on determining broadcast addresses, consult with you network administrator.)


Copyright © 2004 Progress Software Corporation
www.progress.com
Voice: (781) 280-4000
Fax: (781) 280-4095