On most modern networks, including the Internet, users locate other computers by name. This frees users from the daunting task of remembering the numerical network address of network resources. The most effective way to configure a network to allow such name-based connections is to set up a Domain Name Service (DNS) or a nameserver, which resolves hostnames on the network to numerical addresses and vice versa.
This chapter reviews the nameserver included in Red Hat Enterprise Linux, the Berkeley Internet Name Domain (BIND) DNS server, with an emphasis on the structure of its configuration files and how it may be administered both locally and remotely.
For instructions on configuring BIND using the graphical Domain Name Service Configuration Tool (redhat-config-bind), refer to the chapter called BIND Configuration in the Red Hat Enterprise Linux System Administration Guide.
If using the Domain Name Service Configuration Tool, do not manually edit any BIND configuration files as all changes are overwritten the next time the Domain Name Service Configuration Tool is used.
When hosts on a network connect to one another via a hostname, also called a fully qualified domain name (FQDN), DNS is used to associate the names of machines to the IP address for the host.
Use of DNS and FQDNs also has advantages for system administrators, allowing the flexibility to change the IP address for a host without effecting name-based queries to the machine. Conversely, administrators can shuffle which machines handle a name-based query.
DNS is normally implemented using centralized servers that are authoritative for some domains and refer to other DNS servers for other domains.
When a client host requests information from a nameserver, it usually connects to port 53. The nameserver then attempts to resolve the FQDN based on its resolver library, which may contain authoritative information about the host requested or cached data from an earlier query. If the nameserver does not already have the answer in its resolver library, it queries other nameservers, called root nameservers, to determine which nameservers are authoritative for the FQDN in question. Then, with that information, it queries the authoritative nameservers to determine the IP address of the requested host. If performing a reverse lookup, the same procedure is used, except the query is made with an unknown IP address rather than a name.
On the Internet, the FQDN of a host can be broken down into different sections. These sections are organized into a hierarchy (much like a tree), with a main trunk, primary branches, secondary branches, and so forth. Consider the following FQDN:
When looking at how an FQDN is resolved to find the IP address that relates to a particular system, read the name from right to left, with each level of the hierarchy divided by periods (.). In this example, com defines the top level domain for this FQDN. The name example is a sub-domain under com, while sales is a sub-domain under example. The name furthest to the left, bob, identifies a specific machine hostname.
Except for the hostname, each section is a called a zone, which defines a specific namespace. A namespace controls the naming of the sub-domains to its left. While this example only contains two sub-domains, a FQDN must contain at least one sub-domain but may include many more, depending upon how the namespace is organized.
Zones are defined on authoritative nameservers through the use of zone files, which describe the namespace of that zone, the mail servers to be used for a particular domain or sub-domain, and more. Zone files are stored on primary nameservers (also called master nameservers), which are truly authoritative and where changes are made to the files, and secondary nameservers (also called slave nameservers), which receive their zone files from the primary nameservers. Any nameserver can be a primary and secondary nameserver for different zones at the same time, and they may also be considered authoritative for multiple zones. It all depends on how the nameserver is configured.
There are four primary nameserver configuration types:
master — Stores original and authoritative zone records for a namespace, and answers queries about the namespace from other nameservers.
slave — Answers queries from other nameservers concerning namespaces for which it is considered an authority. However, slave nameservers get their namespace information from master nameservers.
caching-only — Offers name to IP resolution services but is not authoritative for any zones. Answers for all resolutions are cached in memory for a fixed period of time, which is specified by the retrieved zone record.
forwarding — Forwards requests to a specific list of nameservers for name resolution. If none of the specified nameservers can perform the resolution, the resolution fails.
A nameserver may be one or more of these types. For example, a nameserver can be a master for some zones, a slave for others, and only offer forwarding resolutions for others.
BIND performs name resolution services through the /usr/sbin/named daemon. BIND also includes an administration utility called /usr/sbin/rndc. More information about rndc can be found in Section 12.4 Using rndc.
BIND stores its configuration files in the following locations:
/etc/named.conf — The configuration file for the named daemon.
/var/named/ directory — The named working directory which stores zone, statistic, and cache files.
The next few sections review the BIND configuration files in more detail.