My preference has been to roll my own. Since a large part of effective log analysis is pattern matching, Perl's native support for regular expressions make it a natural choice. Additionally, there a number of Perl libraries and code examples publicly available that can provide a sound basis for analytical tools.

logresolve.pl

A parallelized forking IP to host name conversion tool (Apache)

Logfile::Apache

A basic statistics summarizer (CPAN)

Data::Xtab

A count tabulator (CPAN)

Net::Country

Conveniently maps top level domain country codes to names (CPAN)

GIFgraph

Takes numeric data structures and output graphs (CPAN)

DBI

A generic interface to RDBMS (CPAN)