flowdumper - a grep(1)-like utility for raw flow files

   flowdumper [-h] [-v] [-s|S|r|R] [-a|n] [[-I expr] -e expr [-E expr]] [-c] [-B file] [-o output_file] [flow_file [...]]

but usually just:

   flowdumper [-s] -e expr flow_file [...]

flowdumper is a grep(1)-like utility for selecting and processing flows from cflowd or flow-tools raw flow files. The selection criteria are specified by using the CW-e option described below.

flowdumper's primary features are the ability to:

•

Print the content of raw flow files in one of two built-in formats or a format of the users own. The built-in long format is much like that produced by the flowdump command supplied with cflowd. The short, single-line format is suitable for subsequent post-processing by line-oriented filters like sed(1).

•

Act as a filter, reading raw flow input from either file(s) or standard input, and producing filtered raw flow output on standard output. This is similar to how grep(1) is often used on text files.

•

Select flows according to practically any criteria that can be expressed in perl syntax.

The flow variables and other symbols available for use in the CW-e expression are those made available by the Cflow module when used like this:

   use Cflow qw(:flowvars :tcpflags :icmptypes :icmpcodes);

See the Cflow perl documentation for full details on these values (i.e. perldoc Cflow.)

Most perl syntax is allowed in the expressions specified with the CW-e, CW-I, and CW-E options. See the perl man pages for full details on operators (man perlop) and functions (man perlfunc) available for use in those expressions.

If run with no arguments, filters standard input to standard output.

The options and their arguments, roughly in order of usefulness, are: shows the usage information mnemonic: 'h'elp print all flows implied if CW-e is not specified mnemonic: 'a'll evaluate this expression once per flow mnemonic: 'e'xpression print number of flows matched in input mnemonic: 'c'ount print flows in short (one-line) format, ignored with CW-n mnemonic: 's'hort print flows in the raw/binary flow file format ignored with CW-n mnemonic: 'r'aw repacks and print flows in the raw/binary flow file format requires CW-e, ignored with CW-n, useful with CW-p mnemonic: 'R'epack raw don't print matching flows mnemonic: like "perl CW-n or sed CW-n" send output to the specified file. A single printf(3) string conversion specifier can be used within the output_file value (such as /tmp/%s.txt) to make the output file name a function of the input file basename. mneomic: 'o'utput file print flows in the old short (one-line) format ignored with CW-n mnemonic: 'S'hort be verbose with messages mnemonic: 'v'erbose be very verbose with messages (implies "CW-v") mnemonic: 'V'ery verbose eval expression initially, before flow processing practically useless without CW-e mnemonic: 'I'nitial expression eval expression after flow processing is complete practically useless without CW-e mnemonic: 'E'ND expression Load the specified BGP dump file using Net::ParseRouteTable. In your optional expression, you can now refer to these variables:

   $dst_as_path_arrayref
   $dst_origin_as
   $dst_peer_as
   $src_as_path_arrayref
   $src_origin_as
   $src_peer_as

   10.42.69.0/24 -> 10.69.42.0/24
   ...

Print all flows, in a multi-line format, to a pager:

   $ flowdumper -a flows.* |less

Print all the UDP flows to another file using the raw binary flow format:

   $ flowdumper -re '17 == $protocol' flows.current > udp_flows.current

Print all TCP flows which have the SYN bit set in the TCP flags:

   $ flowdumper -se '6 == $protocol && ($TH_SYN & $tcp_flags)' flows.*

Print the first 10 flows to another file using the raw binary flow format:

   $ flowdumper -I '$n = 10' -re '$n-- or exit' flows.*0 > head.cflow

Print all flows with the start and end time using a two-line format:

   $ flowdumper -se 'print scalar(localtime($startime)), "\n"' flows.*

Print all flows with the specified source address using a short, single-line format:

   $ flowdumper -se '"10.42.42.42" eq $srcip' flows.*

Do the same thing in a quicker, but less obvious, way:

   $ flowdumper -I '
      use Socket;
      $addr = unpack("N", Socket::inet_aton("10.42.42.42"));
   ' -se '$addr == $srcaddr'  flows.*

(This latter method runs quicker because inet_aton(3) is only called once, instead of once per flow.)

Print all flows with a source address within the specifed network/subnet:

   $ flowdumper \
   -I 'use Socket;
       $mask = unpack("N", Socket::inet_aton("10.42.0.0"));
       $width = 16' \
   -se '$mask == ((0xffffffff << (32-$width)) & $srcaddr)' flows.*

Print all flows where either the source or the destination address, but not both, is within the specified set of networks or subnets:

   $ flowdumper \
   -I 'use Net::Patricia;
       $pt = Net::Patricia->new;
       map { $pt->add_string($_, 1) } qw( 10.42.0.0/16
                                          10.69.0.0/16 )' \
   -se '1 == ($pt->match_integer($srcaddr) +
              $pt->match_integer($dstaddr))' flows.*

Count the total number of talkers (unique source host addresses) by piping them to sort(1) and wc(1) to count them:

   $ flowdumper \
   -I 'use Net::Patricia;
       $pt = Net::Patricia->new;
       map { $pt->add_string($_, 1) } qw( 10.42.0.0/16
                                          10.69.0.0/16 )' \
   -ne '$pt->match_integer($srcaddr) and print "$srcip\n"' flows.* \
   |sort -u |wc -l

Count the total number of talkers (unique source host addresses) that are within a the specified networks or subnets:

   $ flowdumper \
   -I 'use Net::Patricia;
       $pt = new Net::Patricia;
       map { $pt->add_string($_, 1) } qw( 10.42.0.0/16
                                          10.69.0.0/16 );
       $talkers = new Net::Patricia' \
   -ne '$pt->match_integer($srcaddr) &&
        ($talkers->match_integer($srcaddr) or
         $talkers->add_string($srcip, 1))' \
   -E 'printf("%d\n", $talkers->climb( sub { 1 } ))' flows.*

(For large numbers of flows, this latter method is quicker because it populates a Net::Patricia trie with the unique addresses and counts the resulting nodes rather than having to print them to standard output and then having to sort them to determine how many are unique.)

Select the TCP flows and ENCODE the IP addresses according to the prefix encodings specified in prefix_encodings.txt:

   $ flowdumper -p prefix_encodings.txt -se '6 == $protocol && ENCODE'

Produce a new raw flow file with the IP addresses ENCODEd according to the prefix encodings specified in prefix_encodings.txt:

   $ flowdumper -p prefix_encodings.txt -Re 'ENCODE' flows > flows.enc

Produce a set of raw flow files that have the CW$src_as and CW$dst_as origin AS values filled in based upon a lookup in externally-specified routing table (in the file router.bgp) and have the IP address info replaces with zeroes (for anonymity):

   $ ssh router "show route protocol bgp terse" > router.bgp # Juniper

   $ flowdumper \
   -B router.bgp \
   -e '$src_as = $src_origin_as,
       $dst_as = $dst_origin_as,
       (($exporter = 0),
        ($srcaddr  = 0),
        ($src_mask = 0),
        ($dstaddr  = 0),
        ($dst_mask = 0),
        ($nexthop  = 0), 1)' \
   -R \
   -o /tmp/%s.cflow_enc \
   flows*

This utility was inspired by Daniel McRobb's flowdump utility which is supplied with cflowd. flowdumper was originally written as merely a sample of what can be done with the Cflow perl module, but has since been developed into a more complete tool.

When using the CW-B option, routing table entries that contain AS sets at the end of the AS path are quietly discarded. (It's not so quiet if you also specified CW-V.) It was necessary to discard these, because I did not consider AS sets when designing the API and therefore have no way to communicate more than one origin AS value per for a single source or destination IP address.

There are perhaps some pathological combinations of options that currently do not produce usage error messages, but should.

Since the expression syntax is that of perl itself, there are lots of useless expressions that will happily be accepted without complaint. This is particular troublesome when trying to track down typos, for instance, with the flow variable names.

This script probably has the same bugs as the Cflow module, since it's based upon it.

Dave Plonka <plonka@doit.wisc.edu>

Copyright (C) 1998-2002 Dave Plonka. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version.

perl(1), Socket, Net::Netmask, Net::Patricia, Cflow.