Release 2.99b.

[nsc-5.git] / README

             Domain Name Server Configuration Utilities -- NSC 2.99b

                    (c) 1997--2003 Martin Mares <mj@ucw.cz>

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~


------------------------------------------------------------------------------------
WARNING: This is a prerelease of the NSC 3.0. See NEWS for the list of changes.
Most importantly, the configuration files are NOT compatible with the old releases.
Also, it hasn't undergone much extensive testing yet, so there might be bugs around.
------------------------------------------------------------------------------------


   NSC is a set of shell and M4 scripts for easy maintenance of DNS zone files
and name server daemon configuration (currently available only for BIND 8.X,
but easily portable for other daemons). It has been designed to make administration
of a DNS server a piece of cake (unlike other utilities which resemble more
an English pudding :-) ), which includes automatic generation of reverse records
for all your hosts, handling of classless reverse delegations and support for IPv6
(AAAA and PTR in in6.arpa, not A6 and DNAME which seem to be dying out).

   NSC requires GNU m4 and a POSIX-compatible shell, some of the extra utilities
require Perl 5. I've tested everything on Linux (Debian Woody), but the whole
package should run on other unices as well.

   The whole package can be used and distributed according to the terms of the
GNU General Public License. See file COPYING in any of the GNU utility archives
(you should have one as you are expected to have at least GNU M4 ;-)).


0. Quick Howto for the Impatient
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
(everything will be explained in more detail in the subsequent sections)

        - Create a directory where all NSC files will reside (e.g., /etc/named)
          and copy everything from the NSC distribution here.

        - Symlink /etc/bind/named.conf (or /etc/named.conf or where the config file
          of your installation of BIND resides) to /etc/named/named.conf

        - Change directory to /etc/named

        - Edit cf/domains to suit your needs -- replace the example domains
          by your entries.

        - Create cf/<domain-name> for all domains (again, you can easily follow
          the example domains).

        - Run bin/nsconfig (Makefile and named.conf will be generated).

        - Run make.

        - Enjoy your new DNS setup. If everything goes OK, be happy. Else
          write a bug report :-)

        - Every time you modify the domain files

   An interesting companion to this package is the DNS Sleuth -- a DNS zone
consistency checker. It's a simple utility written in Perl with help of the
DNS module and it should be able to detect all common errors in DNS setup
(I have written it after much disappointment with the other checkers).
The Sleuth is available online on http://atrey.karlin.mff.cuni.cz/~mj/sleuth/,
follow the links to download the source.


1. Directory structure
~~~~~~~~~~~~~~~~~~~~~~
The NSC directory (/etc/named in the above example) contains the following
files and subdirectories:

        cf/                     - user-defined configuration files
        cf/domains              - the domain list (see Section 2)
        cf/config               - global settings (see Section 3)
        cf/<domain>             - each domain has its own config file
        bin/                    - commands (e.g., nsconfig)
        m4/                     - M4 scripts (used by the commands)
        zone/                   - primary zone files
        bak/                    - backups of zones we serve as a secondary NS for
        ver/                    - version files where NSC remembers version
                                  numbers of the primary zones

   How are different files created:

        - You create everything in cf/.
        - Then you run bin/nsconfig.
        - Makefile and named.conf gets created according to cf/domains.
        - You run make.
        - The Makefile creates primary zone files in zone/ and version files
          in ver/ and tells BIND to reload its configuration.
        - BIND downloads contents of secondary zones and puts them to bak/.


2. The Domain List File
~~~~~~~~~~~~~~~~~~~~~~~
The domain list contains configuration commands describing all domains handled
by your server and their parameters. In fact, it's a M4 script, but viewing it as
a config file is a good approximation (however, see Section 8 for some caveats).
Lines starting with a semicolon are treated as comments and ignored. Text outside
declarations is silently ignored.

You can specify:

PRIMARY(zone, [extra-files...])
                Define a zone (domain) we run a primary name server for.
                The contents of the zone are described in cf/<zone>
                and possibly in other specified cf files (all files are
                concatenated to produce a single configuration). See the next
                section for a look inside these files.

SECONDARY(zone, primary)
                Define a zone we run a secondary name server for.
                "primary" is an IP address of the primary name server.

REVERSE(network, primary-files...)
                Define a reverse zone for the given network. The network name
                consists of several numbers separated by dots, just like an IP
                address does, but the network usually has only 3 components.
                Each reverse zone has its own config file cf/<network> which
                can of course specify the contents of the zone.

                However, there is a more convenient method to generate the PTR
                records directly from the A records: just specify the REVERSE
                directive in cf/<network> and then include all the config files
                for the primary zones containing hosts from this network. The
                automatic concatenation of multiple primary-files comes very
                handy for that.

                In fact, REVERSE(network, p-f...) is almost an equivalent of
                PRIMARY(REV(network), p-f...) where REV(network) is a macro
                translating network numbers to names of the corresponding
                reverse zones [e.g., REV(1.2.3) equals 3.2.1.in-addr.arpa].
                The only difference is that although the domain name is translated
                by REV, the config file is still named according to the network.
                You can also use the REV macro explicitly, which can be handy
                for example in SECONDARY declarations.

CONFIG(...)     - insert user data to named.conf

MAKEFILE(...)   - insert user data to Makefile


3. The Domain Files
~~~~~~~~~~~~~~~~~~~
The domain files contain descriptions of all DNS records for the given
domain, starting with the SOA record. Again, these are M4 scripts and the
declarations are macro calls. Lines starting with a semicolon are treated
as comments and just copied to the generated zone file. Text outside
declarations is copied to the zone file as well, so you can spice up the NSC
output with your own records.

All host or domain names are either names relative to the current domain
with no dots inside or absolute names (in this case, NSC automatically
ensures that the trailing dot is present in the resource records). Relative
names with dots are not supported, but they are rare and you can always write
them as absolute anyway.

Your menu:

SOA(domain-name)
                Generate a SOA record for the domain. This must be the first
                declaration in the config file. The parameters of the SOA
                are taken from configuration variables (see below). The
                serial number is calculated from the version number remembered
                in the version file, following the usual practice of encoding
                current date and a sequence number within the current day
                in the serial number, which is guaranteed to be strictly
                increasing unless you perform more than 99 updates in a single
                day (in which case NSC stops and tells you to tweak the serial
                number manually).

                The SOA record otherwise acts like a sub-domain (D) declaration,
                therefore it can be followed by other records like NS (mandatory)
                or MX.

H(host)
                Start declaration of a host. Doesn't generate anything, only
                remembers the host's name.

ADDR(addr...)
                Specify addresses for the current host. In the normal mode, it
                creates A records, in the reverse mode, PTR records.

H(host, addr...)
                A shortcut for H(host) ADDR(addr...) -- in many cases everything
                you need for a single host.

DADDR(addr...)
                Like ADDR, but suppresses PTR records. (This one is useful if you
                have a single IP address used for zillions of names and you want
                to avoid having zillions of PTR records for the same address.)

DH(host, addr...)
                A shortcut for H(host) DADDR(addr...)

D(domain)
                Start declaration of a sub-domain. Technically the same as H(domain),
                but this one should be more intuitive.

GLUE(ns, addr...)
                Specify a glue record for a name server contained within a sub-domain
                it's a primary for. Currently it's an equivalent of DH(ns, addr...).

NS(ns...)
                Specify a list of name server names for the current domain
                (started by either a SOA or D declaration). Generates NS records.

MX(mx...)
                Specify a list of mail exchangers for the current host or domain.
                Each mail exchanger should be preceded by a priority. Generates
                MX records.

HI(hw,os)
                Specify a HINFO record for the current host. Very rare in the
                today's Internet.

ALIAS(alias...)
                Specify a list of aliases for the current host or domain.
                Generates a series of CNAME records pointing from the aliases
                to the current host/domain.

CNAME(src, dest)
                Generate a CNAME record -- "src" points to "dest".

PTR(src, dest)
                Generate a PTR record -- "src" points to "dest". It's a common
                record in reverse zones (and although it's legal in forward
                zones as well, such use is very rare), however it's more convenient
                to have your PTR's generated by the REVERSE directive. But if you
                need anything special, here is the tool.

REVBLOCK(subdomain, min, max)
                Generate a series of CNAME records numbered from `min' to `max'
                and pointing to the same name in the given sub-domain, finally
                declaring the sub-domain as well, so you can continue with its
                NS records.

                Example: REVBLOCK(a, 16, 18) NS(ns.xyzzy.org) yields

                        16      CNAME   16.a
                        17      CNAME   17.a
                        18      CNAME   18.a
                        a       NS      ns.xyzzy.org.

                This is a very common construct for classless reverse delegations,
                see Section 6 for more details.

REVERSE(network)
                Switch to reverse mode. From this point on, all output is suppressed
                except for ADDR declarations belonging to the specified network which
                are automatically converted to PTR records.

                With help of this feature, defining reverse zones can be as easy as:

                        ; Reverse zone for 10.0.0.0/24 a.k.a. 0.0.10.in-addr.arpa.
                        SOA(REV(10.0.0))
                        NS(ns1.example.com, ns2.example.com)
                        REVERSE(10.0.0)
                        ; Include all primary zones containing ADDR's from this range,
                        ; which can be accomplished by a multi-file REVERSE declaration
                        ; in cf/domains.


4. Configuration variables
~~~~~~~~~~~~~~~~~~~~~~~~~~
There is a fair amount of configuration variables (which are in reality normal
M4 macros). Each variable has a hard-wired default value which can be overridden
in cf/config by re-defining the variable. Also, all other config files can specify
their local definitions, but you need to be careful to change the variable before
it is used for the first time.

To change the setting, use

                define(`variable', `value')

As usually, even this config file is a M4 script. Comments can be started by
semicolons, text outside macros is ignored.

The following variables are available:

NAMED_RESTART_CMD       Shell command for restarting the name server daemon
                        (default: ndc restart)

ROOT                    Root directory of the whole package (default: /etc/named)
CFDIR                   Directory with config files (default: cf)
ZONEDIR                 Directory with zone files (default: zone)
BAKDIR                  Directory with backup files (default: bak)
VERSDIR                 Directory with version files (default: var)
ROOTCACHE               File with the cache of root name servers

REFRESH                 SOA record parameters
RETRY
EXPIRE
MINTTL
NSNAME                  Origin server (default: hostname of your machine)
MAINTNAME               Domain maintainer name (default: root@NSNAME)

BIND_OPTIONS            Extra options to put to the options { ... } section of named.conf

For the timing parameters, the following shortcuts are available:

HOURS(n)                Convert hours to seconds
MINUTES(n)              Convert minutes to seconds
DAYS(n)                 Convert days to seconds

For the BIND_OPTIONS, we offer:

FORWARD(ip...)          Try to ask the given name servers first to see if they
                        have the reply cached.
SLAVE(ip...)            Pass all non-local requests to the given name servers.


5. Makefile targets
~~~~~~~~~~~~~~~~~~~
The Makefile generated by NSC offers the following targets:

        all (default)           - update all zone files and reload the daemon
        clean                   - clean all generated zone files and backups
        clobber                 - clean + delete Makefile and named.conf
                                  (wise to do after major reconfigurations)
        distclean               - clobber + delete all version files (use only
                                  if you really know what you are doing as the
                                  serial number information in newly generated
                                  files might be inconsistent then).


6. Classless reverse delegations
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
NSC also supports classless delegations for reverse zones using the mechanism
described in RFC 2317, i.e. by putting CNAME records to the reverse zone which
point to records of the same name in a sub-domain which you can delegate directly.

For example if you want to delegate 64-127 in 0.0.10.in-addr.arpa to ns.example.net,
you create a 64/26 sub-domain (26 is the network prefix length) and add the following
records to 0.0.10.in-addr.arpa:

        64      CNAME   64.64/26.0.0.10.in-addr.arpa.
        65      CNAME   65.64/26.0.0.10.in-addr.arpa.
        ...
        127     CNAME   127.64/26.0.0.10.in-addr.arpa.

        64/26   NS      ns.example.net.

Then you configure ns.example.net to be a primary name server for the zone
64/26.0.0.10.in-addr.arpa and put the PTR records there:

        64      PTR     sixty-four.example.net.
        65      PTR     sixty-five.example.net.
        ...
        127     PTR     two-to-seven-minus-one.example.net.

NSC offers special primitives for configuring such delegations, but not limited
to the sub-domain name syntax shown above (which is recommended by the RFC, but it's
far from being the only one used in the real world, other possibilities being for
example 64-127, 64+64 etc.).

The CNAME block can be generated by the REVBLOCK(subdomain-name, low-addr, high-addr)
directive in the configuration of the whole reverse zone. The example above would
be written as:

        REVBLOCK(64/26, 64, 127)

The sub-zone can be created automatically like any another reverse zone, you only
need to use the three-parameter form of the REVERSE directive to specify the
address range in order to filter out possible hosts falling outside your range.

CAVEAT: The slashes in zone names are automatically translated to @'s when forming
file names.

Again for the example above, you need to put the following to cf/domains:

        REVERSE(10.0.0.64/26, <list-of-domains-to-gather-the-addresses-from>)

And to cf/64@26.0.0.10:

        SOA(REV(10.0.0.64/26))
        NS(<list-of-name-servers>)
        REVERSE(10.0.0, 64, 127)

NOTE: It's usually helpful to configure the primary name server for the parent
domain (i.e., the one where you configure the delegation and create the CNAME's)
as a secondary for the sub-zone as well, so if it replies with the CNAME, it will
include the PTR record pointed to by the CNAME in the additional section of its
reply, eliminating the need for an extra query.


7. Support for IPv6
~~~~~~~~~~~~~~~~~~~
NSC also supports IPv6 in a pretty straightforward form: wherever you can write
an IPv4 address, you can use an IPv6 address as well. Incomplete IP addresses
or ranges used for specifying address blocks for reverse delegations are replaced
by network prefixes of the standard form <address>/<prefix-length>.

Example:

        H(ianus, 1.2.3.4, fec0::1234:5678:9abc:def0)

specifies a dual-stack host with both an A record and an AAAA record.

CAVEAT: The backward-compatible IPv6 address syntax with ":v.w.x.y" at the end
is not supported. All other syntaxes and quirks hopefully are.


8. Interaction with M4
~~~~~~~~~~~~~~~~~~~~~~
All config files are fully-fledged M4 scripts, so you can use any M4 features
you need, the most helpful one being definition of your own macros by

        define(`macro_name', `expansion')

However, there is a couple of things you need to care about:

  o  The comment character is redefined to `;'. I.e., wherever a semicolon
     occurs, the rest of the line is a comment which is copied verbatim
     to the output file (if the output is not suppressed like in case
     of the cf/domains file).

  o  Names starting with 'nsc_' or spelled in all caps are reserved
     for the NSC itself and unless documented, messing with them can
     bring surprising results. If you need to use such a name in your
     zone file (maybe you like to shout in your host names :-) ),
     quote it with ` and '.

  o  Don't use commas, quotes nor parentheses in your record names.


9. Other utilities
~~~~~~~~~~~~~~~~~~
convert         A simple Perl script for conversion of zone files to NSC
                domain files. Requires the Net::DNS module (available from
                CPAN at ftp.cpan.org; present in recent versions of Perl).
                Keep in mind that the script is very simple and its craft
                is of a very limited kind, so check its output carefully.

chkdel          A simple Perl script for checking of domain delegations --
                it checks all PRIMARY and SECONDARY records in cf/domains
                against NS records. Requires the Net::DNS module and also
                some tweaking of parameters at the top of the script.

 LocalWords:  MINTTL NSNAME hostname MAINTNAME