File:  [NetBSD Developer Wiki] / wikisrc / guide / dns.mdwn
Revision 1.3: download - view: text, annotated - select for diffs
Thu Mar 21 11:11:20 2013 UTC (12 months, 4 weeks ago) by jdf
Branches: MAIN
CVS tags: HEAD
Add TOCs to all chapters.

**Contents**

[[!toc levels=3]]

# The Domain Name System

Use of the Domain Name System has been discussed in previous chapters, without
going into detail on the setup of the server providing the service. This chapter
describes setting up a simple, small domain with one Domain Name System (DNS)
nameserver on a NetBSD system. It includes a brief explanation and overview of
the DNS; further information can be obtained from the DNS Resources Directory
(DNSRD) at [http://www.dns.net/dnsrd/](http://www.dns.net/dnsrd/).

## DNS Background and Concepts

The DNS is a widely used *naming service* on the Internet and other TCP/IP
networks. The network protocols, data and file formats, and other aspects of the
DNS are Internet Standards, specified in a number of RFC documents, and
described by a number of other reference and tutorial works. The DNS has a
distributed, client-server architecture. There are reference implementations for
the server and client, but these are not part of the standard. There are a
number of additional implementations available for many platforms.

### Naming Services

Naming services are used to provide a mapping between textual names and
configuration data of some form. A *nameserver* maintains this mapping, and
clients request the nameserver to *resolve* a name into its attached data.

The reader should have a good understanding of basic hosts to IP address mapping
and IP address class specifications, see
[[Name Service Concepts|guide/net-intro#nsconcepts]].

In the case of the DNS, the configuration data bound to a name is in the form of
standard *Resource Records* (RRs). These textual names conform to certain
structural conventions.

### The DNS namespace

The DNS presents a hierarchical name space, much like a UNIX filesystem,
pictured as an inverted tree with the *root* at the top.

    TOP-LEVEL                                .org
                                               |
    MID-LEVEL                             .diverge.org
                         ______________________|________________________
                        |                      |                        |
    BOTTOM-LEVEL strider.diverge.org   samwise.diverge.org   wormtongue.diverge.org

The system can also be logically divided even further if one wishes at different
points. The example shown above shows three nodes on the diverge.org domain, but
we could even divide diverge.org into subdomains such as
"strider.net1.diverge.org", "samwise.net2.diverge.org" and
"wormtongue.net2.diverge.org"; in this case, 2 nodes reside in
"net2.diverge.org" and one in "net1.diverge.org".

There are directories of names, some of which may be sub-directories of further
names. These directories are sometimes called *zones*. There is provision for
symbolic links, redirecting requests for information on one name to the records
bound to another name. Each name recognised by the DNS is called a *Domain
Name*, whether it represents information about a specific host, or a directory
of subordinate Domain Names (or both, or something else).

Unlike most filesystem naming schemes, however, Domain Names are written with
the innermost name on the left, and progressively higher-level domains to the
right, all the way up to the root directory if necessary. The separator used
when writing Domain Names is a period, ".".

Like filesystem pathnames, Domain Names can be written in an absolute or
relative manner, though there are some differences in detail. For instance,
there is no way to indirectly refer to the parent domain like with the UNIX `..`
directory. Many (but not all) resolvers offer a search path facility, so that
partially-specified names can be resolved relative to additional listed
sub-domains other than the client's own domain. Names that are completely
specified all the way to the root are called *Fully Qualified Domain Names* or
*FQDN*s. A defining characteristic of an FQDN is that it is written with a
terminating period. The same name, without the terminating period, may be
considered relative to some other sub-domain. It is rare for this to occur
without malicious intent, but in part because of this possibility, FQDNs are
required as configuration parameters in some circumstances.

On the Internet, there are some established conventions for the names of the
first few levels of the tree, at which point the hierarchy reaches the level of
an individual organisation. This organisation is responsible for establishing
and maintaining conventions further down the tree, within its own domain.

### Resource Records

Resource Records for a domain are stored in a standardised format in an ASCII
text file, often called a *zone file*. The following Resource Records are
commonly used (a number of others are defined but not often used, or no longer
used). In some cases, there may be multiple RR types associated with a name, and
even multiple records of the same type.

#### Common DNS Resource Records

 * *A: Address* -- This record contains the numerical IP address associated with
   the name.

 * *CNAME: Canonical Name* -- This record contains the Canonical Name (an FQDN
   with an associated A record) of the host name to which this record is bound.
   This record type is used to provide name aliasing, by providing a link to
   another name with which other appropriate RR's are associated. If a name has
   a CNAME record bound to it, it is an alias, and no other RR's are permitted
   to be bound to the same name.

   It is common for these records to be used to point to hosts providing a
   particular service, such as an FTP or HTTP server. If the service must be
   moved to another host, the alias can be changed, and the same name will reach
   the new host.

 * *PTR: Pointer* -- This record contains a textual name. These records are
   bound to names built in a special way from numerical IP addresses, and are
   used to provide a reverse mapping from an IP address to a textual name. This
   is described in more detail in [[Reverse Resolution|guide/dns#bg-reverse]].

 * *NS: Name Server* -- This record type is used to *delegate* a sub-tree of the
   Domain Name space to another nameserver. The record contains the FQDN of a
   DNS nameserver with information on the sub-domain, and is bound to the name
   of the sub-domain. In this manner, the hierarchical structure of the DNS is
   established. Delegation is described in more detail in
   [[Delegation|guide/dns#bg-delegation]].

 * *MX: Mail eXchange* -- This record contains the FQDN for a host that will
   accept SMTP electronic mail for the named domain, together with a priority
   value used to select an MX host when relaying mail. It is used to indicate
   other servers that are willing to receive and spool mail for the domain if
   the primary MX is unreachable for a time. It is also used to direct email to
   a central server, if desired, rather than to each and every individual
   workstation.

 * *HINFO: Host Information* -- Contains two strings, intended for use to
   describe the host hardware and operating system platform. There are defined
   strings to use for some systems, but their use is not enforced. Some sites,
   because of security considerations, do not publicise this information.

 * *TXT: Text* -- A free-form text field, sometimes used as a comment field,
   sometimes overlaid with site-specific additional meaning to be interpreted by
   local conventions.

 * *SOA: Start of Authority* -- This record is required to appear for each zone
   file. It lists the primary nameserver and the email address of the person
   responsible for the domain, together with default values for a number of
   fields associated with maintaining consistency across multiple servers and
   caching of the results of DNS queries.

### Delegation

Using NS records, authority for portions of the DNS namespace below a certain
point in the tree can be delegated, and further sub-parts below that delegated
again. It is at this point that the distinction between a domain and a zone
becomes important. Any name in the DNS is called a domain, and the term applies
to that name and to any subordinate names below that one in the tree. The
boundaries of a zone are narrower, and are defined by delegations. A zone starts
with a delegation (or at the root), and encompasses all names in the domain
below that point, excluding names below any subsequent delegations.

This distinction is important for implementation - a zone is a single
administrative entity (with a single SOA record), and all data for the zone is
referred to by a single file, called a *zone file*. A zone file may contain more
than one period-separated level of the namespace tree, if desired, by including
periods in the names in that zone file. In order to simplify administration and
prevent overly-large zone files, it is quite legal for a DNS server to delegate
to itself, splitting the domain into several zones kept on the same server.

### Delegation to multiple servers

For redundancy, it is common (and often administratively required) that there be
more than one nameserver providing information on a zone. It is also common that
at least one of these servers be located at some distance (in terms of network
topology) from the others, so that knowledge of that zone does not become
unavailable in case of connectivity failure. Each nameserver will be listed in
an NS record bound to the name of the zone, stored in the parent zone on the
server responsible for the parent domain. In this way, those searching the name
hierarchy from the top down can contact any one of the servers to continue
narrowing their search. This is occasionally called *walking the tree*.

There are a number of nameservers on the Internet which are called *root
nameservers*. These servers provide information on the very top levels of the
domain namespace tree. These servers are special in that their addresses must be
pre-configured into nameservers as a place to start finding other servers.
Isolated networks that cannot access these servers may need to provide their own
root nameservers.

### Secondaries, Caching, and the SOA record

In order to maintain consistency between these servers, one is usually
configured as the *primary* server, and all administrative changes are made on
this server. The other servers are configured as *secondaries*, and transfer the
contents of the zone from the primary. This operational model is not required,
and if external considerations require it, multiple primaries can be used
instead, but consistency must then be maintained by other means. DNS servers
that store Resource Records for a zone, whether they be primary or secondary
servers, are said to be *authoritative* for the zone. A DNS server can be
authoritative for several zones.

When nameservers receive responses to queries, they can *cache* the results.
This has a significant beneficial impact on the speed of queries, the query load
on high-level nameservers, and network utilisation. It is also a major
contributor to the memory usage of the nameserver process.

There are a number of parameters that are important to maintaining consistency
amongst the secondaries and caches. The values for these parameters for a
particular domain zone file are stored in the SOA record. These fields are:

#### Fields of the SOA Record

 * *Serial* -- A serial number for the zone file. This should be incremented any
   time the data in the domain is changed. When a secondary wants to check if
   its data is up-to-date, it checks the serial number on the primary's SOA
   record.

 * *Refresh* -- A time, in seconds, specifying how often the secondary should
   check the serial number on the primary, and start a new transfer if the
   primary has newer data.

 * *Retry* -- If a secondary fails to connect to the primary when the refresh
   time has elapsed (for example, if the host is down), this value specifies, in
   seconds, how often the connection should be retried.

 * *Expire* -- If the retries fail to reach the primary within this number of
   seconds, the secondary destroys its copies of the zone data file(s), and
   stops answering requests for the domain. This stops very old and potentially
   inaccurate data from remaining in circulation.

 * *TTL* -- This field specifies a time, in seconds, that the resource records
   in this zone should remain valid in the caches of other nameservers. If the
   data is volatile, this value should be short. TTL is a commonly-used acronym,
   that stands for "Time To Live".

### Name Resolution

DNS clients are configured with the addresses of DNS servers. Usually, these are
servers which are authoritative for the domain of which they are a member. All
requests for name resolution start with a request to one of these local servers.
DNS queries can be of two forms:

 * A *recursive* query asks the nameserver to resolve a name completely, and
   return the result. If the request cannot be satisfied directly, the
   nameserver looks in its configuration and caches for a server higher up the
   domain tree which may have more information. In the worst case, this will be
   a list of pre-configured servers for the root domain. These addresses are
   returned in a response called a *referral*. The local nameserver must then
   send its request to one of these servers.

 * Normally, this will be an *iterative* query, which asks the second nameserver
   to either respond with an authoritative reply, or with the addresses of
   nameservers (NS records) listed in its tables or caches as authoritative for
   the relevant zone. The local nameserver then makes iterative queries, walking
   the tree downwards until an authoritative answer is found (either positive or
   negative) and returned to the client.

In some configurations, such as when firewalls prevent direct IP communications
between DNS clients and external nameservers, or when a site is connected to the
rest of the world via a slow link, a nameserver can be configured with
information about a *forwarder*. This is an external nameserver to which the
local nameserver should make requests as a client would, asking the external
nameserver to perform the full recursive name lookup, and return the result in a
single query (which can then be cached), rather than reply with referrals.

### Reverse Resolution

The DNS provides resolution from a textual name to a resource record, such as an
A record with an IP address. It does not provide a means, other than exhaustive
search, to match in the opposite direction; there is no mechanism to ask which
name is bound to a particular RR.

For many RR types, this is of no real consequence, however it is often useful to
identify by name the host which owns a particular IP address. Rather than
complicate the design and implementation of the DNS database engine by providing
matching functions in both directions, the DNS utilises the existing mechanisms
and creates a special namespace, populated with PTR records, for IP address to
name resolution. Resolving in this manner is often called *reverse resolution*,
despite the inaccurate implications of the term.

The manner in which this is achieved is as follows:

 * A normal domain name is reserved and defined to be for the purpose of mapping
   IP addresses. The domain name used is `in-addr.arpa.` which shows the
   historical origins of the Internet in the US Government's Defence Advanced
   Research Projects Agency's funding program.

 * This domain is then subdivided and delegated according to the structure of IP
   addresses. IP addresses are often written in *decimal dotted quad notation*,
   where each octet of the 4-octet long address is written in decimal, separated
   by dots. IP address ranges are usually delegated with more and more of the
   left-most parts of the address in common as the delegation gets smaller.
   Thus, to allow delegation of the reverse lookup domain to be done easily,
   this is turned around when used with the hierarchical DNS namespace, which
   places higher level domains on the right of the name.

 * Each byte of the IP address is written, as an ASCII text representation of
   the number expressed in decimal, with the octets in reverse order, separated
   by dots and appended with the in-addr.arpa. domain name. For example, to
   determine the hostname of a network device with IP address 11.22.33.44, this
   algorithm would produce the string `44.33.22.11.in-addr.arpa.` which is a
   legal, structured Domain Name. A normal nameservice query would then be sent
   to the nameserver asking for a PTR record bound to the generated name.

 * The PTR record, if found, will contain the FQDN of a host.

One consequence of this is that it is possible for mismatch to occur. Resolving
a name into an A record, and then resolving the name built from the address in
that A record to a PTR record, may not result in a PTR record which contains the
original name. There is no restriction within the DNS that the "reverse" mapping
must coincide with the "forward" mapping. This is a useful feature in some
circumstances, particularly when it is required that more than one name has an A
record bound to it which contains the same IP address.

While there is no such restriction within the DNS, some application server
programs or network libraries will reject connections from hosts that do not
satisfy the following test:

 * the state information included with an incoming connection includes the IP
   address of the source of the request.

 * a PTR lookup is done to obtain an FQDN of the host making the connection

 * an A lookup is then done on the returned name, and the connection rejected if
   the source IP address is not listed amongst the A records that get returned.

This is done as a security precaution, to help detect and prevent malicious
sites impersonating other sites by configuring their own PTR records to return
the names of hosts belonging to another organisation.

## The DNS Files

Now let's look at actually setting up a small DNS enabled network. We will
continue to use the examples mentioned in [Chapter 24, *Setting up TCP/IP on
NetBSD in practice*](chap-net-practice.html "Chapter 24. Setting up TCP/IP on
NetBSD in practice"), i.e. we assume that:

 * Our IP networking is working correctly
 * We have IPNAT working correctly
 * Currently all hosts use the ISP for DNS

Our Name Server will be the `strider` host which also runs IPNAT, and our two
clients use "strider" as a gateway. It is not really relevant as to what type of
interface is on "strider", but for argument's sake we will say a 56k dial up
connection.

So, before going any further, let's look at our `/etc/hosts` file on "strider"
before we have made the alterations to use DNS.

**Example strider's `/etc/hosts` file**

    127.0.0.1       localhost
    192.168.1.1     strider
    192.168.1.2     samwise sam
    192.168.1.3     wormtongue worm

This is not exactly a huge network, but it is worth noting that the same rules
apply for larger networks as we discuss in the context of this section.

The other assumption we want to make is that the domain we want to set up is
`diverge.org`, and that the domain is only known on our internal network, and
not worldwide. Proper registration of the nameserver's IP address as primary
would be needed in addition to a static IP. These are mostly administrative
issues which are left out here.

The NetBSD operating system provides a set of config files for you to use for
setting up DNS. Along with a default `/etc/named.conf`, the following files are
stored in the `/etc/namedb` directory:

 * `localhost`
 * `127`
 * `loopback.v6`
 * `root.cache`

You will see modified versions of these files in this section, and I strongly
suggest making a backup copy of the original files for reference purposes.

*Note*: The examples in this chapter refer to BIND major version 8, however, it
should be noted that format of the name database and other config files are
almost 100% compatible between version. The only difference I noticed was that
the `$TTL` information was not required.

### /etc/named.conf

The first file we want to look at is `/etc/named.conf`. This file is the config
file for bind (hence the catchy name). Setting up system like the one we are
doing is relatively simple. First, here is what mine looks like:

    options {
            directory "/etc/namedb";
            allow-transfer { 192.168.1.0/24; };
            allow-query { 192.168.1.0/24; };
            listen-on port 53 { 192.168.1.1; };
    };
    
    zone "localhost" {
       type master;
       notify no;
       file "localhost";
    };
    
    zone "127.IN-ADDR.ARPA" {
       type master;
       notify no;
       file "127";
    };
    
    zone "0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.ip6.int" {
       type master;
       file "loopback.v6";
    };
    
    zone "diverge.org" {
       type master;
       notify no;
       file "diverge.org";
    };
    
    zone "1.168.192.in-addr.arpa" {
       type master;
       notify no;
       file "1.168.192";
    };
    
    zone "." in {
       type hint;
       file "root.cache";
    };

Note that in my `named.conf` the root (".") section is last, that is because
there is another domain called diverge.org on the internet (I happen to own it)
so I want the resolver to look out on the internet last. This is not normally
the case on most systems.

Another very important thing to remember here is that if you have an internal
setup, in other words no live internet connection and/or no need to do root
server lookups, comment out the root (".") zone. It may cause lookup problems if
a particular client decides it wants to reference a domain on the internet,
which our server couldn't resolve itself.

Looks like a pretty big mess, upon closer examination it is revealed that many
of the lines in each section are somewhat redundant. So we should only have to
explain them a few times.

Lets go through the sections of `named.conf`:

#### options

This section defines some global parameters, most noticeable is the location of
the DNS tables, on this particular system, they will be put in `/etc/namedb` as
indicated by the "directory" option.

Following are the rest of the params:

 * `allow-transfer` -- This option lists which remote DNS servers acting as
   secondaries are allowed to do zone transfers, i.e. are allowed to read all
   DNS data at once. For privacy reasons, this should be restricted to secondary
   DNS servers only.

 * `allow-query` -- This option defines hosts from what network may query this
   name server at all. Restricting queries only to the local network
   (192.168.1.0/24) prevents queries arriving on the DNS server's external
   interface, and prevent possible privacy issues.

 * `listen-on port` -- This option defined the port and associated IP addresses
   this server will run
   [named(8)](http://netbsd.gw.com/cgi-bin/man-cgi?named+8+NetBSD-5.0.1+i386)
   on. Again, the "external" interface is not listened here, to prevent queries
   getting received from "outside".

The rest of the `named.conf` file consists of `zone`s. A zone is an area that
can have items to resolve attached, e.g. a domain can have hostnames attached to
resolve into IP addresses, and a reverse-zone can have IP addresses attached
that get resolved back into hostnames. Each zone has a file associated with it,
and a table within that file for resolving that particular zone. As is readily
apparent, their format in `named.conf` is strikingly similar, so I will
highlight just one of their records:

#### zone diverge.org

 * `type` -- The type of a zone is usually of type "master" in all cases except
   for the root zone `.` and for zones that a secondary (backup) service is
   provided - the type obviously is "secondary" in the latter case.

 * `notify` -- Do you want to send out notifications to secondaries when your
   zone changes? Obviously not in this setup, so this is set to "no".

 * `file` -- This option sets the filename in our `/etc/namedb` directory where
   records about this particular zone may be found. For the "diverge.org" zone,
   the file `/etc/namedb/diverge.org` is used.

### /etc/namedb/localhost

For the most part, the zone files look quite similar, however, each one does
have some unique properties. Here is what the `localhost` file looks like:

     1|$TTL    3600
     2|@              IN SOA  strider.diverge.org. root.diverge.org. (
     3|                        1       ; Serial
     4|                        8H      ; Refresh
     5|                        2H      ; Retry
     6|                        1W      ; Expire
     7|                        1D)     ; Minimum TTL
     8|               IN NS   localhost.
     9|localhost.     IN      A       127.0.0.1
    10|               IN      AAAA    ::1

Line by line:

 * *Line 1*: This is the Time To Live for lookups, which defines how long other
   DNS servers will cache that value before discarding it. This value is
   generally the same in all the files.

 * *Line 2*: This line is generally the same in all zone files except
   `root.cache`. It defines a so-called "Start Of Authority" (SOA) header, which
   contains some basic information about a zone. Of specific interest on this
   line are "strider.diverge.org." and "root.diverge.org." (note the trailing
   dots!). Obviously one is the name of this server and the other is the contact
   for this DNS server, in most cases root seems a little ambiguous, it is
   preferred that a regular email account be used for the contact information,
   with the "@" replaced by a "." (for example, mine would be
   "jrf.diverge.org.").

 * *Line 3*: This line is the serial number identifying the "version" of the
   zone's data set (file). The serial number should be incremented each time
   there is a change to the file, the usual format is to either start with a
   value of "1" and increase it for every change, or use a value of "YYYYMMDDNN"
   to encode year (YYYY), month (MM), day (DD) and change within one day (NN) in
   the serial number.

 * *Line 4*: This is the refresh rate of the server, in this file it is set to
   once every 8 hours.

 * *Line 5*: The retry rate.

 * *Line 6*: Lookup expiry.

 * *Line 7*: The minimum Time To Live.

 * *Line 8*: This is the Nameserver line, which uses a "NS" resource record to
   show that "localhost" is the only DNS server handing out data for this zone
   (which is "@", which indicates the zone name used in the `named.conf` file,
   i.e. "diverge.org") is, well, "localhost".

 * *Line 9*: This is the localhost entry, which uses an "A" resource record to
   indicate that the name "localhost" should be resolved into the IP-address
   127.0.0.1 for IPv4 queries (which specifically ask for the "A" record).

 * *Line 10*: This line is the IPv6 entry, which returns ::1 when someone asks
   for an IPv6-address (by specifically asking for the AAAA record) of
   "localhost.".

### /etc/namedb/zone.127.0.0

This is the reverse lookup file (or zone) to resolve the special IP address
127.0.0.1 back to "localhost":

     1| $TTL    3600
     2| @              IN SOA  strider.diverge.org. root.diverge.org. (
     3|                        1       ; Serial
     4|                        8H      ; Refresh
     5|                        2H      ; Retry
     6|                        1W      ; Expire
     7|                        1D)     ; Minimum TTL
     8|                 IN NS   localhost.
     9| 1.0.0           IN PTR  localhost.

In this file, all of the lines are the same as the localhost zonefile with
exception of line 9, this is the reverse lookup (PTR) record. The zone used here
is "@" again, which got set to the value given in `named.conf`, i.e.
"127.in-addr.arpa". This is a special "domain" which is used to do
reverse-lookup of IP addresses back into hostnames. For it to work, the four
bytes of the IPv4 address are reserved, and the domain "in-addr.arpa" attached,
so to resolve the IP address "127.0.0.1", the PTR record of
"1.0.0.127.in-addr.arpa" is queried, which is what is defined in that line.

### /etc/namedb/diverge.org

This zone file is populated by records for all of our hosts. Here is what it
looks like:

     1| $TTL    3600
     2| @              IN SOA  strider.diverge.org. root.diverge.org. (
     3|                         1       ; serial
     4|                         8H      ; refresh
     5|                         2H      ; retry
     6|                         1W      ; expire
     7|                         1D )    ; minimum seconds
     8|                 IN NS   strider.diverge.org.
     9|                 IN MX   10 strider.diverge.org.   ; primary mail server
    10|                 IN MX   20 samwise.diverge.org.   ; secondary mail server
    11| strider         IN A     192.168.1.1
    12| samwise         IN A     192.168.1.2
    13| www             IN CNAME samwise.diverge.org.
    14| worm            IN A     192.168.1.3

There is a lot of new stuff here, so lets just look over each line that is new
here:

 * *Line 9*: This line shows our mail exchanger (MX), in this case it is
   "strider". The number that precedes "strider.diverge.org." is the priority
   number, the lower the number their higher the priority. The way we are setup
   here is if "strider" cannot handle the mail, then "samwise" will.

 * *Line 11*: CNAME stands for canonical name, or an alias for an existing
   hostname, which must have an A record. So we have aliased `www.diverge.org`
   to `samwise.diverge.org`.

The rest of the records are simply mappings of IP address to a full name (A
records).

### /etc/namedb/1.168.192

This zone file is the reverse file for all of the host records, to map their IP
numbers we use on our private network back into hostnames. The format is similar
to that of the "localhost" version with the obvious exception being the
addresses are different via the different zone given in the `named.conf` file,
i.e. "0.168.192.in-addr.arpa" here:

     1|$TTL    3600
     2|@              IN SOA  strider.diverge.org. root.diverge.org. (
     3|                     1       ; serial
     4|                     8H      ; refresh
     5|                     2H      ; retry
     6|                     1W      ; expire
     7|                     1D )    ; minimum seconds
     8|               IN NS   strider.diverge.org.
     9|1              IN PTR  strider.diverge.org.
    10|2              IN PTR  samwise.diverge.org.
    11|3              IN PTR  worm.diverge.org.

### /etc/namedb/root.cache

This file contains a list of root name servers for your server to query when it
gets requests outside of its own domain that it cannot answer itself. Here are
first few lines of a root zone file:

    ;
    ;       This file holds the information on root name servers needed to
    ;       initialize cache of Internet domain name servers
    ;       (e.g. reference this file in the "cache  .  <file>"
    ;       configuration file of BIND domain name servers).
    ;
    ;       This file is made available by InterNIC
    ;       under anonymous FTP as
    ;           file                /domain/db.cache
    ;           on server           FTP.INTERNIC.NET
    ;       -OR-                    RS.INTERNIC.NET
    ;
    ;       last update:    Jan 29, 2004
    ;       related version of root zone:   2004012900
    ;
    ;
    ; formerly NS.INTERNIC.NET
    ;
    .                        3600000  IN  NS    A.ROOT-SERVERS.NET.
    A.ROOT-SERVERS.NET.      3600000      A     198.41.0.4
    ;
    ; formerly NS1.ISI.EDU
    ;
    .                        3600000      NS    B.ROOT-SERVERS.NET.
    B.ROOT-SERVERS.NET.      3600000      A     192.228.79.201
    ;
    ; formerly C.PSI.NET
    ;
    .                        3600000      NS    C.ROOT-SERVERS.NET.
    C.ROOT-SERVERS.NET.      3600000      A     192.33.4.12
    ;
    ...

This file can be obtained from ISC at <http://www.isc.org/> and usually comes
with a distribution of BIND. A `root.cache` file is included in the NetBSD
operating system's "etc" set.

This section has described the most important files and settings for a DNS
server. Please see the BIND documentation in `/usr/src/dist/bind/doc/bog` and
[named.conf(5)](http://netbsd.gw.com/cgi-bin/man-cgi?named.conf+5+NetBSD-5.0.1+i386)
for more information.

## Using DNS

In this section we will look at how to get DNS going and setup "strider" to use
its own DNS services.

Setting up named to start automatically is quite simple. In `/etc/rc.conf`
simply set `named=yes`. Additional options can be specified in `named_flags`,
for example, I like to use `-g nogroup -u nobody`, so a non-root account runs
the "named" process.

In addition to being able to startup "named" at boot time, it can also be
controlled with the `ndc` command. In a nutshell the `ndc` command can stop,
start or restart the named server process. It can also do a great many other
things. Before use, it has to be setup to communicate with the "named" process,
see the [ndc(8)](http://netbsd.gw.com/cgi-bin/man-cgi?ndc+8+NetBSD-5.0.1+i386)
and
[named.conf(5)](http://netbsd.gw.com/cgi-bin/man-cgi?named.conf+5+NetBSD-5.0.1+i386)
man pages for more details on setting up communication channels between "ndc"
and the "named" process.

Next we want to point "strider" to itself for lookups. We have two simple steps,
first, decide on our resolution order. On a network this small, it is likely
that each host has a copy of the hosts table, so we can get away with using
`/etc/hosts` first, and then DNS. However, on larger networks it is much easier
to use DNS. Either way, the file where order of name services used for
resolution is determined is `/etc/nsswitch.conf` (see
[[`nsswitch.conf`|guide/net-practice#ex-nsswitch]]. Here is part of a typical
`nsswitch.conf`:

    ...
    group_compat:   nis
    hosts:          files dns
    netgroup:       files [notfound=return] nis
    ...

The line we are interested in is the "hosts" line. "files" means the system uses
the `/etc/hosts` file first to determine ip to name translation, and if it can't
find an entry, it will try DNS.

The next file to look at is `/etc/resolv.conf`, which is used to configure DNS
lookups ("resolution") on the client side. The format is pretty self explanatory
but we will go over it anyway:

    domain diverge.org
    search diverge.org
    nameserver 192.168.1.1

In a nutshell this file is telling the resolver that this machine belongs to the
"diverge.org" domain, which means that lookups that contain only a hostname
without a "." gets this domain appended to build a FQDN. If that lookup doesn't
succeed, the domains in the "search" line are tried next. Finally, the
"nameserver" line gives the IP addresses of one or more DNS servers that should
be used to resolve DNS queries.

To test our nameserver we can use several commands, for example:

    # host sam
    sam.diverge.org has address 192.168.1.2

As can be seen, the domain was appended automatically here, using the value from
`/etc/resolv.conf`. Here is another example, the output of running
`host www.yahoo.com`:

    $ host www.yahoo.com
    www.yahoo.com is an alias for www.yahoo.akadns.net.
    www.yahoo.akadns.net has address 68.142.226.38
    www.yahoo.akadns.net has address 68.142.226.39
    www.yahoo.akadns.net has address 68.142.226.46
    www.yahoo.akadns.net has address 68.142.226.50
    www.yahoo.akadns.net has address 68.142.226.51
    www.yahoo.akadns.net has address 68.142.226.54
    www.yahoo.akadns.net has address 68.142.226.55
    www.yahoo.akadns.net has address 68.142.226.32

Other commands for debugging DNS besides
[host(1)](http://netbsd.gw.com/cgi-bin/man-cgi?host+1+NetBSD-5.0.1+i386) are
[nslookup(8)](http://netbsd.gw.com/cgi-bin/man-cgi?nslookup+8+NetBSD-5.0.1+i386)
and
[dig(1)](http://netbsd.gw.com/cgi-bin/man-cgi?dig+1+NetBSD-5.0.1+i386). Note
that
[ping(8)](http://netbsd.gw.com/cgi-bin/man-cgi?ping+8+NetBSD-5.0.1+i386)
is *not* useful for debugging DNS, as it will use whatever is configured in
`/etc/nsswitch.conf` to do the name-lookup.

At this point the server is configured properly. The procedure for setting up
the client hosts are easier, you only need to setup `/etc/nsswitch.conf` and
`/etc/resolv.conf` to the same values as on the server.

## Setting up a caching only name server

A caching only name server has no local zones; all the queries it receives are
forwarded to the root servers and the replies are accumulated in the local
cache. The next time the query is performed the answer will be faster because
the data is already in the server's cache. Since this type of server doesn't
handle local zones, to resolve the names of the local hosts it will still be
necessary to use the already known `/etc/hosts` file.

Since NetBSD supplies defaults for all the files needed by a caching only
server, it only needs to be enabled and started and is immediately ready for
use! To enable named, put `named=yes` into `/etc/rc.conf`, and tell the system
to use it adding the following line to the `/etc/resolv.conf` file:

    # cat /etc/resolv.conf
    nameserver 127.0.0.1

Now we can start named:

    # sh /etc/rc.d/named restart

### Testing the server

Now that the server is running we can test it using the
[nslookup(8)](http://netbsd.gw.com/cgi-bin/man-cgi?nslookup+8+NetBSD-5.0.1+i386)
program:

    $ nslookup
    Default server: localhost
    Address: 127.0.0.1
    
    >

Let's try to resolve a host name, for example "www.NetBSD.org":

    > www.NetBSD.org
    Server:  localhost
    Address:  127.0.0.1
    
    Name:    www.NetBSD.org
    Address:  204.152.190.12

If you repeat the query a second time, the result is slightly different:

    > www.NetBSD.org
    Server:  localhost
    Address:  127.0.0.1
    
    Non-authoritative answer:
    Name:    www.NetBSD.org
    Address:  204.152.190.12

As you've probably noticed, the address is the same, but the message
`Non-authoritative answer` has appeared. This message indicates that the answer
is not coming from an authoritative server for the domain NetBSD.org but from
the cache of our own server.

The results of this first test confirm that the server is working correctly.

We can also try the
[host(1)](http://netbsd.gw.com/cgi-bin/man-cgi?host+1+NetBSD-5.0.1+i386) and
[dig(1)](http://netbsd.gw.com/cgi-bin/man-cgi?dig+1+NetBSD-5.0.1+i386) commands,
which give the following result.

    $ host www.NetBSD.org
    www.NetBSD.org has address 204.152.190.12
    $
    $ dig www.NetBSD.org
    
    ; <<>> DiG 8.3 <<>> www.NetBSD.org
    ;; res options: init recurs defnam dnsrch
    ;; got answer:
    ;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 19409
    ;; flags: qr rd ra; QUERY: 1, ANSWER: 1, AUTHORITY: 5, ADDITIONAL: 0
    ;; QUERY SECTION:
    ;;      www.NetBSD.org, type = A, class = IN
    
    ;; ANSWER SECTION:
    www.NetBSD.org.         23h32m54s IN A  204.152.190.12
    
    ;; AUTHORITY SECTION:
    NetBSD.org.             23h32m54s IN NS  uucp-gw-1.pa.dec.com.
    NetBSD.org.             23h32m54s IN NS  uucp-gw-2.pa.dec.com.
    NetBSD.org.             23h32m54s IN NS  ns.NetBSD.org.
    NetBSD.org.             23h32m54s IN NS  adns1.berkeley.edu.
    NetBSD.org.             23h32m54s IN NS  adns2.berkeley.edu.
    
    ;; Total query time: 14 msec
    ;; FROM: miyu to SERVER: 127.0.0.1
    ;; WHEN: Thu Nov 25 22:59:36 2004
    ;; MSG SIZE  sent: 32  rcvd: 175

As you can see
[dig(1)](http://netbsd.gw.com/cgi-bin/man-cgi?dig+1+NetBSD-5.0.1+i386) gives
quite a bit of output, the expected answer can be found in the "ANSWER SECTION".
The other data given may be of interest when debugging DNS problems.


CVSweb for NetBSD wikisrc <wikimaster@NetBSD.org> software: FreeBSD-CVSweb