Imported Upstream version 2.23.05

[hcoop/zz_old/debian/webalizer.git] / DNS.README

The Webalizer - A log file analysis program  -- DNS information

The webalizer has the ability to perform reverse DNS lookups,  and
fully supports both IPv4 and IPv6 addressing schemes.  This document
attempts to explain how it works, and some things that you should be
aware of when using the DNS lookup features.

Note: The Reverse DNS feature may be enabled or disabled at compile
      time.  DNS lookup code is enabled by default.  You can run The
      Webalizer using the '-vV' command line options to determine what
      options are enabled in the version you are using.


How it works
------------

DNS lookups are made against a DNS cache file containing IP addresses
and resolved names.  If the IP address is not found in the cache file,
it will be left as an IP address.  In order for this to happen, a
cache file MUST be specified when the Webalizer is run, either using
the '-D' command line switch, or a "DNSCache" configuration file
keyword.  If no cache file is specified, no attempts to perform DNS
lookups will be done. The cache file can be made three different ways.

1) You can have the Webalizer pre-process the specified log file at
   run-time, creating the cache file before processing the log file
   normally.  This is done by setting the number of DNS Children
   processes to run, either by using the '-N' command line switch or
   the "DNSChildren" configuration keyword.  This will cause the
   Webalizer to spawn the specified number of processes which will
   be used to do reverse DNS lookups.. generally, a larger number
   of processes will result in faster resolution of the log, however
   if set too high may cause overall system degradation.  A setting
   of between 5 and 20 should be acceptable, and there is a maximum
   limit of 100.   If used, a cache filename MUST be specified also,
   using either the '-D' command line switch, or the "DNSCache"
   configuration keyword.  Using this method, normal processing will
   continue only after all IP addresses have been processed, and the
   cache file is created/updated.

2) You can pre-process the log file as a standalone process, creating
   the cache file that will be used later by the Webalizer.  This is
   done by running the Webalizer with a name of 'webazolver' (ie: the
   name 'webazolver' is a symbolic link to 'webalizer') and specifying
   the cache filename (either with '-D' or DNSCache).   If the number
   of child processes is not given, the default of 5 will be used. In
   this mode, the log will be read and processed, creating a DNS cache
   file or updating an existing one, and the program will then exit
   without any further processing.

3) You can use The Webalizer (DNS) Cache file Manager program 'wcmgr'
   to create and manipulate a cache file.  A blank cache file can be
   created which would be later populated, or data for the cache file
   can be imported using tab delimited text files.  See the wcmgr(1)
   man page for usage information.


Run-time DNS cache file creation/update
---------------------------------------

The creation/update of a DNS cache file at run-time occurs as follows:

1) The log file is read, creating a list of all IP addresses that are
   not already cached (or cached but expired) and need to be resolved.
   Addresses are expired based on the TTL value specified using the
   'CacheTTL' configuration option or after 7 days (default) if no TTL
   is specified.

2) The specified number of children processes are forked, and are used
   to perform DNS lookups.

3) Each IP address is given, one at a time, to the next available child
   process until all IP addresses have been processed.  Each child will
   update the cache file when a result is returned.  This may be either
   a resolved name or a failed lookup, in which case the address will be
   left unresolved.  Unresolved addresses are not normally cached, but
   can be, if enabled using the 'CacheIPs' configuration file keyword.

4) Once all IP addresses have been processed and the cache file updated,
   the Webalizer will process the log normally.  Each record it finds
   that has an unresolved IP address will be looked up in the cache file
   to see if a hostname is available (ie: was previously found).

Because there may be a significant amount of time between the initial
unresolved IP list and normal processing, the Webalizer should not be
run against live log files (ie: a log file that is actively being written
to by a server), otherwise there may be additional records present that
were not resolved.


Stand-Alone DNS cache file creation/update
------------------------------------------

The creation/update of the DNS cache file, when run in stand-alone mode,
occurs as follows:

1) The log file is read, creating a list of all IP addresses that are
   not already cached (or cached but expired) and need to be resolved.

2) The specified number of children processes are forked, and are used
   to perform DNS lookups. If the number of processes was not specified,
   the default of 5 will be used.

3) Each IP address is given, one at a time, to the next available child
   process until all IP addresses have been processed.  Each child will
   update the cache file when a result is returned.

4) Once all IP addresses have been processed and the cache file updated,
   the program will terminate without any further processing.


Larger sites may prefer to use a stand-alone process to create the DNS
cache file, and then run the Webalizer against the cache file.  This
allows a single cache file to be used for many virtual hosts, and reduces
the processing needed if many sites are being processed.  The Webalizer
can be used in stand alone mode by running it as 'webazolver'.  When
run in this fashion, it will only create the cache file and then exit
without any further processing.  A cache filename MUST be specified,
however unlike when running the Webalizer normally, the number of child
processes does not have to be given (will default to 5).  All normal
configuration and command line options are recognized, however, many
of them will simply be ignored.. this allows the use of a standard
configuration file for both normal use and stand alone use.


Examples:
---------

webalizer -c test.conf -N 10 -D dns_cache.db /var/log/my_www_log

   This will use the configuration file 'test.conf' to obtain normal
   configuration options such as hostname and output directory.. it
   will then either create or update the file 'dns_cache.db' in the
   default output directory (using 10 child processes) based on the
   IP addresses it finds in the log /var/lib/my_www_log, and then
   process that log file normally.


webalizer -o out -D dns_cache.db /var/log/my_www_log

  This will process the log file /var/log/my_www_log, resolving IP
  addresses from the cache file 'dns_cache.db' found in the default
  output directory "out".  The cache file must be present as it will
  not be created with this command.


for i in /var/log/*/access_log; do
  webazolver -N 20 -D /var/lib/dns_cache.db $i
done

  The above is an example of how to run through multiple log files
  creating a single DNS cache file.. this might be typically used on
  a larger site that has many virtual hosts, all keeping their log
  files in a separate directory.  It will process each access_log it
  finds in /var/log/* and create a cache file (var/lib/dns_cache.db).
  This cache file can then be used to process the logs normally with
  with the Webalizer in a read-only fashion (see next example).


for i in /etc/webalizer/*.conf; do webalizer -c $i -D /etc/cache.db; done

  This will process each configuration file found in /etc/webalizer,
  using the DNS cache file /etc/cache.db.  This will also typically be
  used on a larger site with multiple hosts..  Each configuration file
  will specify a site specific log file, hostname, output directory, etc.
  The cache file used will typically be created using a command similar
  to the one previous to this example.


Cache File Maintenance
----------------------

The Webalizer DNS cache files generally require very little or no
special attention.  There are times though when some maintenance
is required, such as occasional purging of very old cache entries.
The Webalizer never removes a record once it's inserted into the
cache.  If a record expires based on its timestamp, the next time
that address is seen in a log, its name is looked up again and the
timestamp is updated.  However, there will always be addresses that
are never seen again, which will cause the cache files to continue
to grow in size over time.  On extremely busy sites or sites that
attract many one time visitors, the cache file may grow extremely
large, yet only contain a small amount of valid entries.  Using
The Webalizer (DNS) Cache file Manager ('wcmgr'), cache files can
be purged, removing expired entries and shrinking the file size.
A TTL (time to live) value can be specified, so the length of time
an entry remains in the cache can be varied depending on individual
site requirements.  In addition to purging cache files, 'wcmgr' can
also be used to list cache file contents, import/export cache data,
lookup/add/delete individual entries and gather overall statistics
regarding the cache file (number of records, number expired, etc..).

To purge a cache file using 'wcmgr', an example command would be:

wcmgr -p31 /path/to/dns.cache

This would purge the 'dns.cache' cache file of any records that are
over 31 days old, and would reclaim the space that those records
were using in the file.  If you would like to see the records that
get purged, adding the command line option '-v' (verbose) will cause
the program to print each entry and its age as they are removed.
You can also use the 'wcmgr' to display statistics on cache files
to aid in determining when a cache file should be purged.  See the
'wcmgr' man page (wcmgr.1) for additional information on the various
options available.


Stupid Cache Tricks
-------------------

The DNS cache files used by The Webalizer allow for efficient IP address
to name translations.  Resolved names are normally generated by using an
existing DNS name server to query the address, either locally or over
the Internet.  However, using The Webalizer (DNS) Cache file Manager,
almost any IP address to Name translation can be included in the cache.
One such example would be for mapping local network addresses to real
names, even though those addresses may not have real DNS entries on the
network (or may be 'local' addresses prohibited from use on the Internet).
A simple tab delimited text file can be created and imported into a cache
for use by The Webalizer,  which will then be used to convert the local
IP addresses to real names.  Additional configuration options for The
Webalizer can then be used as would be normally.  For example, consider
a small business with 10 computers and a DSL router to the Internet.
Each machine on the local network would use a private IP address that
would not be resolved using an external (public) DNS server, so would
always be reported by The Webalizer as 'unknown/unresolved'.  A simple
cache file could be created to map those unresolved addresses into more
meaningful names, which could then be further processed by the Webalizer.
An example might look something like:

# Local machines
192.168.123.254 0       0       gw.widgetsareus.lan
192.168.123.253 0       0       mail.widgetsareus.lan
192.168.123.250 0       0       sales.widgetsareus.lan
192.168.123.240 0       0       service.widgetsareus.lan
192.168.123.237 0       0       mgr.widgetsareus.lan
192.168.123.235 0       0       support1.widgetsareus.lan
192.168.123.234 0       0       support2.widgetsareus.lan
192.168.123.232 0       0       pres.widgetsareus.lan
192.168.123.230 0       0       vp.widgetsareus.lan
192.168.123.225 0       0       reception.widgetsareus.lan
192.168.123.224 0       0       finance.widgetsareus.lan
127.0.0.1       0       1       127.0.0.1


There are a couple of things here that should be noted.  The first
is that the timestamps (first zero on each line above) are set to
zero.  This tells The Webalizer that these cached entries are to
be considered 'permanent', and should never be expired (infinite
TTL or time to live).  The second thing to note is that the resolved
names are using a non-standard TLD (top level domain) of '.lan'.
The Webalizer will map this special TLD to mean "Local Network" in
its reports, which allows local traffic to be grouped separately
from normal Internet traffic.  Lastly, you may notice that the
last line of the file contains an entry with the same IP address
where a name should be.  This entry will prevent the Webalizer
from ever trying to lookup 127.0.0.1,  which is the 'localhost'
address, when it is found in a log.  The second number after the IP
address (1) tells the Webalizer that it is an unresolved entry, not
a resolved hostname (ie: has no name).  Entries such as this one can
be used to reduce DNS lookups on addresses that are known not to
resolve.


Considerations
--------------

Processing of live log files is discouraged, as the chances of log records
being written between the time of DNS resolution and normal processing will
cause problems.

If you are using STDIN for the input stream (log file) and have run-time
DNS cache file creation/update enabled.. the program will exit after the
cache file has been created/updated and no output will be produced.  If
you must use STDIN for the input log, you will need to process the stream
twice, once to create/update the cache file, and again to produce the
reports.  The reason for this is that stream inputs from STDIN cannot
be 'rewound' to the beginning like files can, so must be given twice.

Cached DNS addresses have a default TTL (time to live) of 7 days.  This
may now be changed using the CacheTTL config file keyword to any value
from 1 to 100 (days).  You may also now specify if unresolved addresses
should be stored in the DNS cache.  Normally, unresolved IP addresses
are NOT saved in the cache and are looked up each time the program is
run.

There is an absolute maximum of 100 child processes that may be created,
however the actual number of children should be significantly less than
the maximum.. typical usage should be between 5 and 20.

Special thanks to Henning P. Schmiedehausen <hps@tanstaafl.de> for the
original dns-resolver code he submitted,  which was the basis for this
implementation.  Also thanks to Jose Carlos Medeiros for the inital IPv6
support code.