When ns_helper runs in a network, the naming server (the part of the operating system that locates objects) has two sources of information about entry directory names: the node's local root directory, which is a subset of ns_helper's master root directory, and the master root directory itself.
When the naming server tries to locate an object, it first looks in the node's local cache of entry directory names. If the name is not in the node's cache, the naming server refers to ns_helper's master root directory for information about the entry name. Also, if the naming server cannot locate an object using the locally cached information, it refers to the master root directory for more recent information about the name. Whenever the naming server gets information from the master root directory, it adds (updates) that information to the local node's cache.
When you use ns_helper on the network, you do not need to use the ctnode command to maintain a node's root directory except in the following unusual cases:
• When you change the name of a node (disked or diskless) in the ns_helper database, the old name is not deleted from the local caches. When you refer to the node by its new name, the nam-ing server updates the local cache. This means that both its old and new names are then in the cache. Use uctnode at each node on the network to remove the old entry from all local caches.
• If you interchange the names of two nodes in the ns_helper database, you must use uctnode locally at each node on the network to remove the old entries. When the naming server refers to the new name, it updates the local cache with current information.
The following subsections describes tools, considerations, and procedures for managing root directories on networks that use ns_helper.
o
The edns Utilityo
o
The edns utility enables you to manage·the ns_helper and its database. Table 2-2 briefly describes the edns commands. The 'DOMAIN System Command Reference describes the edns commands in greater detail. Online help is available by typing
$ help edns and
$ help edns commands
Command
Deletes an ns_helper node from the ns_helper replica lists.
Lists the differences between two ns_helper databases, including both the master root directories and replica lists.
Displays the DOMAIN address and status information for the de-fault ns_helper.
Initializes an ns_helper database with data from all nodes that are currently responding on the network.
Lists master root directory information.
Lists the addresses of all ns_helper nodes on the network; can
Synchronizing Clocks on Replicated Databases
The ns_helper processes keep only the most recent information about an entry in their databases.
o
o o
o
o
Network Availability and edns
Two edns operations, initialize and update, are particularly sensitive to network and node availability because they request information from all nodes on the network. If a node fails to respond, it may not be cataloged in the root directory. Therefore, it is a good idea to initialize the first ns_helper at a time when network traffic is light and all nodes are connected to the network. At these times, most nodes can respond to requests for information about themselves from ns_helper.
edns and Diskless Nodes
When edns initializes a database, it always assigns the default name diskless_$nnnnn
to a diskless node, where nnnnn represents the diskless node's hexadecimal ID. The value is right justified and is preceded by the number of zeros required to form a six-digit number. For example, if the node ID is 3d3, the edns representation of that node is
diskless _ $0003d3
Edns associates a DID that it generates for a diskless node with the name it gives to the diskless node. This information about the diskless node appears in the master root directory and can be copied to a node's local cache. If you use ns_helper in a single DOMAIN network, you should know that the DID and other information that edns generates for the diskless node serves as a place marker for information that is used in a DOMAIN internet; it has no effect in a single network.
As a general rule, you should give diskless nodes non-default names. These names can be mnemonic and enable you to refer to the nodes in commands that take node specifications without using the node's hexadecimal ID. However, remember that the diskless node's name is not the name of an entry directory and cannot be used in a pathname argument.
To name a new diskless node on an existing network, simply use the ctnode -root command described in the next section, or use Procedure 2-10. If the node was on the network when ns_helper was initialized, it already has a default name; in this case, use Procedure 2-12 or use uctnode -root to uncatalog the node before recataloging it.
The command ld / / -In lists the names of all diskless nodes in the local copy of the root directory.
For example:
$ Id -In /I
joes diskless
calamity diskless $003476
bridgel - nodisk
comserver The following commands tell you if a node is diskless:
$ netstat -n node_spec
$ lusr -n node_spec
$ bldt -n node_spec
$ 1cnode -me For example:
$ lusr -n //comserver
barebones
b_jones