free hosting   image hosting   hosting reseller   online album   e-shop   famous people 
Free Website Templates
Free Installer

 


    Setting Up a Nameserver

     
    A nameserver is a server that resolves hostnames to IP addresses. Instead of having to
    type in "209.81.10.250", I can just type in "www.computing.com" to get to a
    site. BIND is nameserver software that runs on many types of machines, originally
    written by Paul Vixie and now maintained by the Internet Software Consortium (ISC).

    This was one of the trickiest things I've had to try to figure out so far. I didn't talk
    to anybody as to how to go about this, unlike PPP setup. But I finally found the
    Linux DNS HOWTO and used that. Hopefully this will save you some trouble. Here I
    have instructions for BIND 8, which is newer and more advanced than BIND 4 (which
    some distributions still use).

    Removing Old Nameserver (BIND 4)


    Installing New Nameserver (BIND 8)

    Configuration: /etc/named.conf

    Caching Nameserver
    Zone Files in /var/named
    Examples of Zone Files
    Domains
    Localhost
    Reverse Mapping
    Resources
    Removing Old Nameserver (BIND 4)

    In the event that your distribution already comes with BIND 8, then all you need to do
    is find out how the configuration works, and/or put in entries for machines that the
    Linux box will be the nameserver for. I think Slackware comes with BIND 8. I don't
    know about Debian. Most new distributions (including Red Hat 5.2) will already have
    BIND 8 so you'll probably have to look for the configuration file and how it has
    been pre-configured.

    I had to remove the old nameserver first, so it wouldn't get in the way of the new one.
    It's fine to remove it and replace it with a new one; there aren't any other
    packages that really depend on it.


    Using Red Hat 5.1, I typed rpm -e bind bind-utils caching-nameserver. That removed all
    the old packages. Be careful about this, especially about bind-utils, because
    bind-utils contains the tools such as dnsquery and nslookup that you might be using
    fairly often. Red Hat 5.2 already has BIND 8.

    Installing New Nameserver (BIND 8)
    First, download BIND from ftp://ftp.isc.org/. If you've got it on your system already,
    then you don't need to get BIND 8 unless you want to make a minor upgrade. The
    filename is something like bind-8.1.2-src.tar.gz, which says that it's BIND version
    8.1.2 in source format (which you have to compile on your system). I'll work with
    the source version since that's what I usually do anyway.


    After you have it on your system, type tar -zxvf bind-8.1.2.tar.gz. It will extract a
    directory called src/ and that's where you go into. Simply type "make" and have it
    compiled for you. If you need anything to be tweaked then read the INSTALL file
    (less INSTALL) and find what you need. After it finishes compiling, type make
    install.

    Configuration: /etc/named.conf



    Configuring the nameserver was probably the hardest part of the process that I had to go
    through. Hopefully, what I've written up about what I learned will save the reader
    some trouble.

    The main BIND 8 configuration file is in /etc/named.conf. The configuration syntax for
    the file is documented at http://www.isc.org/bind8/config.html. Here's a sample of a
    configuration file (as /etc/named.conf), with an explanation below.

    /*
    * A simple BIND 8 configuration
    */

    options {
    directory "/var/named";
    };

    zone "computing.com" in {
    type master;
    file "master/computing.com";
    };

    zone "0.0.127.in-addr.arpa" in {
    type master;
    file "zone/127.0.0";
    };

    zone "." in {
    type hint;
    file "named.cache";
    };
    In "options" I only had one option: where the directory for zone files were. Zone files
    are where information about domains is stored, and each file has information about a
    zone. It's a section to cover, I guess, so that's why they're called zones. I have
    /var/named/ as my named data directory.

    The "computing.com" section is pretty straightforward. It just indicates the
    location of the computing.com zone files and tells named that this server is
    a master nameserver for the computing.com zone.

    The "0.0.127.in-addr.arpa" zone is for mapping localhost to 127.0.0.1, basically. It has
    its own zone file.

    The "." zone indicates a caching nameserver; that is, someone can actually use your
    machine to resolve hostnames (including you). I've heard that is is efficient
    especially when using PPP connections, but I don't know for sure. Read the "Caching
    Nameserver" section to read up on how to create one.

    Caching Nameserver

    First you need to get a "named.cache" file. I'm not sure if you can name it anything
    else, but let's just use that filename. In /var/named/ (or wherever you put your
    nameserver's data files), type dig @a.root-servers.net > named.cache. This will ask
    for the addresses of the main DNS servers of the Internet and direct them to a file.
    I'm guessing that the purpose of this is to give your machine an idea of which
    machines on the Internet to ask about hosts.


    Periodically, like once a month, update the named.cache file by running that command
    once in a while. You can use a cron job for that. If you don't know what I'm talking
    about here, don't worry about it. Just be sure to update it using dig once in a
    while, that's all you have to do.

    You have /etc/named.conf point to wherever your named.cache file is under the "." zone.

    Zone Files in /var/named/
    In /var/named/, I created directories for every type of zone I had. The directories I
    have in there are: master, slave/, and zone. With the domain name system, there is a
    server for each domain that is the main server (the master). I suppose that the
    slave server is there in case the main (master) server is down. For each domain
    there should be at least 2 servers, one master and one slave. That's just the way it
    goes.

    While interning at Penguin Computing I set up both the master and slave DNS servers. The
    master's information should go in the master directory. You should be able to figure
    out where the slave's information goes. The information they store is the same, but
    since one machine is the main one that keeps the information (master) and the other
    simply follows the master's information (slave), you need to stay organized and make
    sure you're configuring the right machine for its right place in the nameserver
    system.


    Note that the slave nameserver for one domain can also be the master nameserver for
    another domain. There just can't be two masters for a single domain, though I think
    there can be several slaves.


    Examples of Zone Files
    To figure something like this out, I was looking hard for examples. And examples really
    help, so hopefully you won't be too confused by my examples. Hey, I try.

    Domains
    The information for each domain is put in a single file. This file contains valuable
    information for each domain, such as machines that are under that domain (like, for
    the computing.com domain, the nameservers would have the information for what
    IP address pasta.computing.com gets and the one that
    antarctica.computing.com gets). Here's an example of a domain's records:

    @ IN SOA computing.com. root.computing.com.
    (
    1998082403 ; serial
    4H ; refresh, seconds
    2H ; retry, seconds
    1W ; expire, seconds
    1D ) ; minimum, seconds
    NS pasta.computing.com.
    NS rice.computing.com.
    MX 10 computing.com. ; Primary Mail Exchanger

    localhost A 127.0.0.1
    router A 140.174.204.2

    computing.com. A 209.81.10.250
    ns A 209.81.10.250
    www A 209.81.10.250
    ftp CNAME computing.com.
    mail CNAME computing.com.
    news CNAME computing.com.
    pasta CNAME computing.com.
    slashdot CNAME computing.com.
    rice CNAME antarctica.computing.com.
    antarctica A 209.81.10.252
    antarctic CNAME antarctica.computing.com.
    www.antarctic CNAME antarctica.computing.com.
    www.antarctica CNAME antarctica.computing.com.
    zork A 209.81.10.253
    tux A 209.81.10.146
    xfce A 209.81.10.252

    @ TXT "Penguin Computing"
    @ HINFO Linux 2.0.34
    There's a pretty weird syntax to be used for these zone files. I never would have
    figured it out on my own had I not read the Linux DNS HOWTO document. Basically, it
    specifies information about all the machines in the domain, and it contains
    information about the domain itself, such as the type of machine the server is
    running on.

    I'll start explaining what all the stuff does. In the first line, it's saying that this
    file specifies the zones for the computing.com domain, and to send anything
    about the domain to root@computing.com. Since the "@" character has special
    significance in these zone files, the username (root) and machine name
    (computing.com) have to be separated by a dot. I guess BIND just knows how to
    split it up. That's how you fill in stuff for your domain as well.

    The line with the comment "serial" shows the serial number of that domain. The syntax is
    YYYYMMDDRR; that is, a four digit year, two digit month in numerical form, two digit
    day format, and a two digit revision number. In this example (1998082403), it shows
    that the zone file was last modified on August 24, 1998. It's the third revision for
    that day. When you're changing anything in the file, make sure to increase the
    revision number by one if the previous change was on the same day. If I were to
    change the IP of one of the hosts, I would make the last two numbers, currently 03,
    to 04.

    The next few lines show times for certain functions such as refreshing, retrying, and
    expiring the information. I'm not absolutely sure, but my best guess is that H
    stands for hour, D stands for day, and W stands for week.

    The "NS" line indicates all the nameservers for that particular domain, including the
    one this information is on. This information has to match what has been registered
    with InterNIC. For the hostnames of the nameservers, remember to add a dot at the
    end. If you don't, it will add the hostname to the current domain. For example, if
    you forgot the dot at the end of pasta.computing.com, you would end up with
    the nameserver being pasta.computing.com.computing.com, which is
    obviously not what it's supposed to be. Watch out for this.

    The MX file is the Mail eXchange record, so that mail can get through to the domain.
    There should also be an entry in /etc/sendmail.cw to allow messages coming in from
    that domain (assuming you're using Sendmail, the default on many Linux systems, for
    mail transfer).

    The next couple of lines point to the local loopback (127.0.0.1), which all Linux
    systems should have even if they aren't connected to a network. The "router" line
    points to the IP address of where the machine's Internet connection is. I'm not sure
    if it's really necessary but I was playing it safe back then and trying to copy the
    example from the DNS HOWTO as closely as possible.

    The rest of the entries use either A (address) or CNAME (Canonical Name) to point
    hostnames to IP addresses. Note that hostnames can be mapped to other hostnames, or
    they can be mapped to IP addresses. Use A to map a name to an IP address, and CNAME
    to map a hostname to another hostname (which must be mapped to another IP address).

    Localhost

    The file for mapping localhost is pretty simple. Not much explanation needed. Of course,
    if you want to copy and paste, be sure you make the proper changes.

    @ IN SOA computing.com root.computing.com (
    1998072401 ; Serial number
    3H ; Refresh
    1H ; Retry
    604800 ; Expire
    86400) ; Minimum TTL

    NS pasta.computing.com.
    NS rice.computing.com.

    1 PTR localhost.
    Reverse Mapping
    This file looks similar to the zone file for the domains, but it provides the opposite
    function. It points IP addresses to hostnames (as opposed to vice versa), because
    many servers on the Internet do this thing called reverse lookup on the IP address
    of your hostname to make sure that you're not doing anything sneaky.

    This is for the zone "209.81.10" specified in the sample configuration file. Note that
    my example is not complete, nor does it work in reality, because Penguin Computing
    doesn't own the whole block of "209.81.10.*". But this is how you'd fill in a file
    to resolve your IP addresses to hostnames if you owned the entire block of IP
    addresses.

    @ IN SOA computing.com. root.computing.com. (
    1998072002 ; Serial
    4H ; Refresh
    2H ; Retry
    604800 ; Expire
    86400) ; Minimum TTL
    NS pasta.computing.com.
    NS rice.computing.com.
    ;
    ; Servers
    ;
    250 PTR pasta.computing.com.
    250 PTR computing.com.
    250 PTR ftp.computing.com.
    250 PTR www.computing.com.
    250 PTR mail.computing.com.
    251 PTR rice.computing.com.

    ;
    ; Workstations
    ;
    252 PTR antarctica.computing.com.
    252 PTR antarctic.computing.com.
    If you were to fill in an actual zone file like this, it's necessary to fill in all the
    entries in your block of IP addresses, from 1 to 255. For something like that you
    may want to assign the task to anyone who looks bored.

    So what should you do if you only own a domain but not the block of IP addresses that
    it's part of? Ask the people who are in charge of that block of IP addresses to map
    your IP addresses to their respective hostnames for you.

     More on http://linuxperf.nl.linux.org/