NIS+ Tables and Information

NIS+ stores a wide variety of network information in tables. This section describes the structure of those tables and provides a brief overview of how they can be set up.

NIS+ Table Structure

NIS+ tables provide several features not found in simple text files or maps. They have a column-entry structure, they accept search paths, they can be linked together, and they can be set up in several different ways. NIS+ provides preconfigured system tables, and you can also create your own tables. The following table lists the preconfigured NIS+ tables.

These tables store a wide variety of information, ranging from user names to Internet services. Most of this information is generated during a setup or configuration procedure. For instance, an entry in the passwd table is created when a user account is set up. An entry in the hosts table is created when a workstation is added to the network. And an entry in the networks table is created when a new network is set up.

Since this information is generated from such a wide field of operations, much of it is beyond the scope of this manual. However, as a convenience, the appendix of this book summarizes the information contained in each column of the tables, providing details only when necessary to keep things from getting confusing, such as when distinguishing groups from NIS+ groups and netgroups. For thorough explanations of the information, consult AIX Version 4.3 System Management Guide: Communications and Networks.

The Cred table, because it contains only information related to NIS+ security, is described in Administering NIS+ Credentials.

Columns and Entries

Although NIS+ tables store different types of information, they all have the same underlying structure; they are each made up of rows and columns (the rows are called entries or entry objects).

A client can access information by a key, or by any column that is searchable. For example, to find the network address of a workstation named baseball, a client could look through the hostname column until it found baseball.

it then would move along the baseball entry to find its network address.

Because a client can access table information at any level, NIS+ provides security mechanisms for all three levels. For instance, an administrator could assign read rights to everyone for a table at the object level, modify rights to the owner at the column level, and modify rights to the group at the entry level. Details about table security are provided in Administering NIS+ Access Rights.

Search Paths

A table contains information only about its local domain. For instance, tables in the wiz.com. domain contain information only about the users, clients, and services of the wiz.com. domain. The tables in the sales.wiz.com. domain store information only about the users, clients, and services of the sales.wiz.com. domain. And so on.

If a client in one domain tries to find information that is stored in another domain, it has to provide a fully qualified name. As described in NIS+ Name Expansion, if the NIS_PATH environment variable is set up properly, the NIS+ service will do this automatically.

Every NIS+ table can also specify a search path that a server will follow when looking for information. The search path is an ordered list of NIS+ tables, separated by colons:

The table names in the search path do not have to be fully qualified; they can be expanded just like names entered at the command line. When a server cannot find information in its local table, it returns the table's search path to the client. The client uses that path to look for the information in every table named in the search path, in order, until it finds the information or runs out of names.

Here is an example that demonstrates the benefit of search paths. Assume a domain hierarchy as shown in the following figure.

The hosts tables of the lower three domains have the contents similar to that described in the following example. Assume now that a user logged onto a client in the sales.wiz.com. domain wants to log in remotely to another client. If that user does not provide a fully qualified name, it can only remotely log on to five workstations: vermont, maine, cherry, apple, and the mailhost.

Now assume that the search path of the hosts table in the sales.wiz.com. domain listed the hosts tables from the test.wiz.com. and eng.wiz.com. domains:

hosts.org_dir.test.wiz.com.:hosts.org_dir.eng.wiz.com.

Now a user in the sales.wiz.com. domain can enter something like rlogin oklahoma, and the NIS+ server will find it. It will first look for oklahoma in the local domain, but when it does not find a match, it will look in the test.wiz.com. domain. How does the client know how to find the test.wiz.com. domain? The information is stored in its directory cache. If it is not stored in its directory cache, the client will obtain the information by following the process described in this chapter.

There is a slight drawback, though, to specifying a search path. If the user were to enter an incorrect name, such as rlogin potatoe, the server would need to look through three tables--instead of just one--before returning an error message. If you set up search paths throughout the namespace, an operation may end up searching through the tables in 10 domains instead of just 2 or 3. Another drawback is a performance loss from having many clients contact more than one set of servers when they need to access NIS+ tables.

You should also be aware that since "mailhost" is often used as an alias, when trying to find information about a specific mailhost, you should use its fully qualified name (for example, mailhost.sales.wiz.com.), or NIS+ will return all the mailhosts it finds in all the domains it searches through.

You can specify a table's search path by using the -p option to the nistbladm command.

Ways to Set Up Tables

Setting up NIS+ tables involves three or four tasks:

1. Creating the org_dir directory
2. Creating the system tables
3. Creating nonsystem tables (optional)
4. Populating the tables with information

NIS+ system tables are stored under an org_dir directory. So, before you can create any tables:

• Use the nisserver script. The nisserver script creates the appropriate directories and a full set of system tables. Running the nisserver script is the recommended method.
• Use the nismkdir command. The nismkdir command simply creates the directory.
• Use the /usr/lib/nis/nissetup utility. The nissetup utility creates the org_dir and groups_dir directories and a full set of system tables.

The nisserver script and the nissetup and nismkdir utilities are described in their reference descriptions.

A benefit of the nissetup utility is its capability to assign the proper access rights to the tables of a domain whose servers are running in NIS-compatibility mode. When entered with the -Y flag, it assigns read permissions to the nobody class of the objects it creates, allowing NIS clients, who are unauthenticated, to get information from the domain's NIS+ tables.

The NIS+ system tables and the type of information they store are described in the appendix. To create them, you could use one of the three ways mentioned above. The nistbladm utility also creates and modifies NIS+ tables. You could conceivably create all the tables in a namespace with the nistbladm command, but you would have to type much more and you would have to know the correct column names and access rights. A much, much easier way is to use the nisserver script.

To create a nonsystem table--that is, a table that has not been preconfigured by NIS+--use the nistbladm command.

You can populate NIS+ tables in three ways: from NIS maps, from ASCII files (such as /etc files), and manually.

If you are upgrading from the NIS service, you already have most of your network information stored in NIS maps. You do not have to re-enter this information manually into NIS+ tables. You can transfer it automatically with the nispopulate script or the nisaddent utility.

If you are not using another network information service, but maintain network data in a set of /etc files, you do not have to re-enter this information, either. You can transfer it automatically, also using the nispopulate script or the nisaddent utility.

If you are setting up a network for the first time, you may not have much network information stored anywhere. In that case, you'll need to first get the information and then enter it manually into the NIS+ tables. You can do this with the nistbladm command. You can also do it by entering all the information for a particular table into an input file--which is essentially the same as an /etc file--and then transferring the contents of the file with the nispopulate script or the nisaddent utility.

How Tables Are Updated

When a domain is set up, its servers receive their first versions of the domain's NIS+ tables. These versions are stored on disk, but when a server begins operating, it loads them into memory. When a server receives an update to a table, it immediately updates its memory-based version of the table. When it receives a request for information, it uses the memory-based copy for its reply.

Of course, the server also needs to store its updates on disk. Since updating disk-based tables takes time, all NIS+ servers keep log files for their tables. The log files are designed to temporarily store changes made to the table, until they can be updated on disk. They use the table name as the prefix and append .log. For example:

hosts.org_dir.log
bootparams.org_dir.log
password.org_dir.log

You should update disk-based copies of a table on a daily basis so that the log files do not grow too large and take up too much disk space. This process is called checkpointing. To do this, use the nisping -C command.