Traffic analysis system (in search of a better name)

This is a small project I did to track my traffic based on the traffic
of the web server - bytes sent, packets sent, tcp retransmits. The code
is very simple and should be easy to understand, but here are some docs,
anyway.

The DB structure has a few logical components:
- routers - for each router that handles the traffic from the web servers
there's an entry in the "router" table, with the network it serves.
! see the note about bgpupd.sh below

- traffic type - there's also an entry for each type of traffic in the 
"traffic" table. This can be set in the originator of the udp packets,
probably on a per-vhost or anything basis (in another incarnation of the
code ther was a check if the file was an video file and that was another
type of traffic).

- current routing table - in the "pfxtmp" table, for each router there's
its full routing table (prefix and ASpath).

- current traffic - in the "iptraf" table there's an entry for each
prefix/router/traffic type combination, that gets updated by one of the
cron scripts.

- historical data - in the "iptraf_arch" table there's a dump of the
data from the "iptraf" table each hour, with the rows with zero traffic
removed.

- country - there's a table "cntry" with all the prefixes for every country
that gets updated by another script, with data from http://ip.ludost.net/
(which is a geolocation db based on the RIR databases).

- there are two aggregate tables, "iptraf_cn" (per-country data)
and "iptraf_dt" (per-asn data).

NOTE: aspath field in the DB is only 255 chars, should be fixed.

The workflow looks like this:

- cron/countryupd.sh updates the table with the countries
- cron/bgpupd.sh updates the prefixes for the router. It needs a quagga
installation, and usually one quagga per router (most of the time this
means another machine). I didn't have the time to try making something
with views and having only one quagga to take the data from all routers.
  It uses a very (badly) stripped down version of the bgpdump tool from
RIPE.
- cron/processdwlog.sh moves the current log, sends HUP to the ipdwl
daemon (so it reopens the log), processes it into queries and sends
them in the db. It uses the readlog php script, which aggregates the
data per ip/router/traffic triplet in separate queries.

- on the other side, the module in the web server (nginx or something
else) for each request takes the information from the tcp_info struct
(to get the retransmits), creates an UDP packet, signs it with SHA1
and sends it to the ipdwl server.
- the ipdwl server daemon receives the packets and writes them to a log
file.


(brief) Installation instructions:

Install postgres, dev

Install ip4r http://pgfoundry.org/projects/ip4r/

create a database (ipr)

as postgres, load the ip4r type
psql -U postgres < ip4r.sql

create a database, load the ipr.schema
psql ipr < ipr.schema

install quagga
give someone access to the quaggavty and quagga groups to run bgpupd from
configure bgp neighbor

compile and install bgpdump in INSTDIR.

In all the sources grep for and replace: SOMESECRET, LOGDIR and INSTDIR.

replace IP.IP.IP.IP with the ip of the server that will listen