NameServer Documentation

The NameServer maintains the global name space and allows processes to find other processes by name. This means that there is only one fixed port in the system - the NameServer port (12345). The NameServer should always be running on each of the robots and cass. Note that there will probably be multiple NameServers running and that they share information for robustness.

Table of contents:

  1. Invoking NameServer
  2. Diagnostics - NameServerShow
  3. NameServer FAQ
  4. Design of NameServer

Invoking NameServer

 
NameServer [-d debug_level] [-p port] [-q findport] [-o number] [-x]

Diagnostics - NameServerShow

 The program NameServerShow (which should be installed somewhere public) lists all of the entries in the NameServer list. NameServerShow contacts the closest/quickest NameServer and lists all of the entries that it has. The format is a list of sets of entries that belong to the NameServers on each host. Each set starts with the hostname and is followed by the entries. The format for each entry is the name of the process followed by a colon, a set of methods that the process supports (INET, UNIX, QNX), the hostname, the IP address, the inet socket, the unix socket and the pid of the process.

cass:~> NameServerShow 
cass.nada.kth.se:
  NameServer: INET UNIX cass.nada.kth.se 192.168.5.31 12345 12345 25379

NameServer FAQ

1. Why can't my process register itself with the NameServer?
Processes send a UDP broadcast packet on the local subnet to locate the NameServer. There must be a NameServer running on the same subnet (and with reasonable communication performance) for things to work. If the results are unreliable, then there is probably too much communication delay (there is a timeout when locating the NameServer). Check things by pinging between the host that you are running the process on and the NameServer host.
2. Why can't my process find another process using the NameServer?
First check that the process that you are trying to locate has been registered with the NameServer (using NameServerShow), then read the answer to question 1.

Design of NameServer

 The NameServer maintains the global name space and allows processes to find other processes by name. There will probably be multiple NameServers running and that they must share information for robustness. In addition, the NameServer should continue to operate even if some of it's peers disappear. This has motivated a two part design for the NameServer with all of the critical tasks done in the primary process and secondary tasks in the background process (which may block sometimes).

Tasks for the primary process:

  1. Maintain a list of names and their addresses
  2. Supply addresses to client processes (allow name lookup)

Tasks for the secondary process:

  1. Periodically try to find other NameServers
  2. Periodically check each of the entries that the primary process has to see if they are valid (delete them if not).
  3. Periodically check if other NameServers have disappeared (and delete their entries from the primary process if so).

Note that some of the tasks for the secondary process can block (specifically, trying to connect to each of the listed processes to see if they still exist can take considerable time if the computer is down).

The NameServer also provides a mechanism for instantiating processes that do not currently exist but can be supplied by some unknown entity (actually the StateManager).