You only need a
dns = 126.96.36.199entry in your ipnet configuration ( see Network configuration ) and that server will be used to answer dns queries.
If you're using ip/ipconfig to configure your network via dhcp, dns entry will be automatically setup.
INTERNAL DNS SERVER
(normally used as a dns-cache to avoid waste on bandwidth with www.google.com queries :) ndb/dns -s)
You need to use root dns servers directly, then you can answer all the names to your clients directly (they doesn't need to go out there to ask for them, so you saves bandwidth). You will need an entry like:
dom= ns=A.ROOT-SERVERS.NET ns=B.ROOT-SERVERS.NET dom=A.ROOT-SERVERS.NET ip=188.8.131.52 dom=B.ROOT-SERVERS.NET ip=184.108.40.206see /lib/ndb/local.complicated for a complete example.
PRIMARY DNS SERVER
(or master -> ndb/dns -s)
You will need an entry like:
dom=cs.bell-labs.com soa= refresh=3600 ttl=3600 ns=plan9.bell-labs.com ns=ns1.cs.bell-labs.com ns=ns2.cs.bell-labs.com email@example.com mx=mail.research.bell-labs.com pref=20 mx=plan9.bell-labs.com pref=10 dnsslave=nslocum.cs.bell-labs.com dnsslave=vex.cs.bell-labs.com
in your /lib/ndb/local file. You can see examples in /lib/ndb/local.complicated that comes with standard distribution.
If you are familiar with DNS, you will find that entry self-explanatory. If not:
- ns - the name of your dns server, this should be defined with a sys entry (see Network configuration). You can put all the ns entries that serve your domain (master and slaves)
- mb - the email of the administrator
- mx adn pref - the mail exchanger (mx) and the preference (pref) it has. The lower number is the one with more preference when choosing a mail server.
- dnsslave - the name of your slave dns server (normally other server). As far as i know you can't transfer zones to a Plan9 dns server automatically, you can distribute your ndb file across your plan9 servers via fs or use external tools to query a remote server and write the appropiate ndb file. Steve Simon created such tool, see on /n/sources/contrib/steve/zonefresh.tar.bz2.
- ttl and refresh - the defaults should be enought for normal setups
If you have your zone entry ready, take a look at ndb file again to see if all the names you used in the zone definition can be resolved using ndb information (with entries like dom=plan9.bell-labs.com ip=220.127.116.11 or sys entries as described in Network configuration ).
SLAVE DNS SERVER
I never done that in plan9, but i suppose is like setting up a primary one. There is no options and no automatic transfer zone from a master to a Plan9 dns slave server.
COMBINATIONS (AND NDB/DNS COMMAND)
It is possible to use a mixture of those configurations. You only need to make all the entries you need in your ndb file and use the correct option to start your dns server:
- ndb/dns -r -> this will answer queries using a remote dns server (or can use root servers entry)
- ndb/dns -s -> this will answer queries about your zones
- ndb/dns -sr -> this will answer queries about your zones, and will answer queries about other names as well (can use root servers entry)
- ndb/dns -snr -> this will do the same of he (-sr) and will update the record of the dnsslaves when changed (you need this if use dnsslaves)
- /rc/bin/service/tcp53 -> you will need enable this listener if you want to transfer zones from a Plan9 dns server to unix slave servers (normally move !tcp53 to tcp53 and restart the listeners)
The ndb/dns command normally is executed via /rc/bin/cpurc, so take a look in this file for the ndb/dns line.