This program was ported to Go. That version is faster, now has more features and has (also) endured billions and billions of production requests.
pgeodns is an authoritative DNS server that can give different replies to each client, taking into account the country of origin of the client and do weighted responses so some records are returned more than others.
It's used to give IPs of "nearby" servers among the almost 2000 servers registered in the NTP Pool. The responses are also weighted by the available bandwidth for each IP (as configued by the server admins).
It's also used by apache.org for svn.apache.org to send European users to their European SVN mirror and North American ones to the US based server. They are providing their configuration as a minimal example: https://svn.apache.org/repos/infra/infrastructure/trunk/dns/zones/pgeodns.conf https://svn.apache.org/repos/infra/infrastructure/trunk/dns/zones/geo.apache.org.json
perl Makefile.PL # will warn if any dependencies are missing make make test # optional make install
You'll need the following modules installed, all available from CPAN: Net::DNS, Geo::IP, List::Util, JSON. It's optional, but if you install JSON::XS loading large zone data files will be ever so slightly faster.
pgeodns needs two configuration files; one simple text file to define the zones served and some options, and then for each zone a JSON formatted data file with the zone data.
JSON is relatively easy to read and write for humans, and extremely easy for computers to use, practically in any language.
The pgeodns.conf file should look like the following. Only one or more "base" lines are required.
# global options base some.zone.example.com data/some.file.json # options for this zone # base another.example.com data/some.file.json # options for this zone
See t/example.com.json for a small example for now.
- --config=[ configuration file ]
Name of the configuration file to load; defaults to pgeodns.conf in the current directory.
- --interface=[ ip | host ]
IP or hostname to listen on (for example 192.168.10.10)
You can specify a comma separated list of IPs. The first IP will be used as the "server id" when returning status information (see "Special Queries" below), so if using anycast be sure to put the local/internal IP first or you won't be able to tell the nodes apart.
- --user=[ username | userid ]
Username or ID to change to after binding to the port.
Provide lots of details for each incoming and outgoing packet.
Load the config and exit. Exits with 0 as the return value if all is well.
- --port=[ 53 ]
Specify which port to listen on. Defaults to 53. Only use this in development or if you are behind a NAT/SNAT device that forwards queries to a different port.
This will enable a query to shutdown.$domain to make pgeodns exit. Obviously not a good idea in production, but can be handy in a development/testing environment.
The options allowed in the 'base' configuration file are
- ns name.server.tld
name.server.tld as a nameserver for this zone (or globally). If
you specify one or more NS'es for a zone, it'll override the global
configuration. You can also specify the ns records in the JSON data,
but doing it in the configuration gives some flexibility for re-using
the data under different namespaces.
- serial 123
Set the serial number of the zone; generally this is better done in the JSON data.
- ttl 300
Set the default time-to-live for the zone in seconds.
- include filename
Include another filename.
To ease monitoring pgeodns supports some special queries, all
If your application is sensitive to revealing this sort of information you will need to disable this in the code.
They work with both 'IN' (internet) and 'CH' (chaos) class queries; in the future we might only support CH.
Returns a text status with the "id"
Returns a text status with the "version".
dig +short -t txt version.pgeodns @a.ntpns.org "184.108.40.206, v1.41"
Returns a JSON formatted data structure with query count, uptime etc.
For historical reasons the special queries work on both the 'pgeodns' top level domain and on any other configured base domain; combined with working with the internet class this has the side effect of making it easy to see which server is responding to your queries:
$ dig +short -t txt -c in status.pool.ntp.org "220.127.116.11, upt: 8901928, q: 993862438, 111.65/qps"