logtools

Table of Contents generated with DocToc

logtools

A log files analysis / filtering framework.

logtools encompasses of a few easy-to-use, easy to configure command-line tools, typically used in conjunction with Apache logs (but not only).

The idea is to standardize log parsing and filtering using a simple, coherent configuration methodology and UNIX/POSIX-like command-line interface (STDIN input streaming, command piping etc.), so as to create a consistent environment for creating reports, charts and other such log mining artifacts that are typically employed in a Website context.

logtools can be used either programmatically from within a python program, or as a set of atomic command-line utilities. See below for description of the command-line tools.

For help on using the programmatic interface, I'm working on creating API documentation using Sphinx. For now, the source code itself is fairly well commented.

This software is distributed under the Apache 2.0 license.

Installation

To install this package and associated console scripts, unpack the distributable tar file, or check out the project directory, and then run:

python setup.py install

You will need sudo/root rights to install this in the global system python installation. Alternatively, you can install logtools into a python virtualenv. See http://pypi.python.org/pypi/virtualenv

To run unit tests, issue this command:

python setup.py test

You can also issue the following to get code-coverage report (needs the coverage python package):

python setup.py nosetests

If for some reason setuptools does not install 'nose' by default, you can easily install it if you have setuptools installed by running:

easy_install nose

Console Scripts

filterbots Used to filter bots based on an ip blacklist and/or a useragent blacklist file(s). The actual regular expression mask used for matching is also user-specified, so this can be used with any arbitrary log format (See examples below). Blacklist files can specify both exact match as well as more complex matching types (prefix/suffix match, regexp match. See examples below).
geoip Simple helper utility for using the Maxmind GeoIP library to tag log lines by the IP's country. The regular expression mask used for matching the IP in the log line is user-specified. This tool requires the Maxmind GeoIP library and python bindings. See http://www.maxmind.com/app/country
logparse Use the logtools.parsers module to intelligibly parse the log, emitting/filtering user-selectable field(s). This can be used for inspecting logs whos format is harder to parse than simply cut-ing on whitespace, e.g CLF, JSON and so forth.
logmerge Merge multiple (individually sorted) input logstreams and stream them out in (combined) sorted order. This is useful for combining logs from multiple traffic-serving machines (e.g in a load-balanced environment) into a single stream which is globally ordered.
logjoin Perform a join on some field between input log stream and an additional, arbitrary source of data. This uses a pluggable driver (similar to logparse) allowing all kinds of joins, e.g between logfile and a database, filesystem objects etc. See examples below.
logsample Produce a random sample of lines from an input log stream. This uses Reservoir Sampling to efficiently produce a random sampling over an arbitrary large input stream. Both uniformly random as well as weighted random variants are available. See examples below.
logfilter Parse a log file and perform generic blacklist/whitelist-based row filtering against a specific field in the log row. Can use simple delimited field formats, or use a parser (see logtools.parsers) for dealing with more complex formats/encodings, e.g JSON
logtail Tail a logfile, based on more complex expressions rather than the number of lines N to limit to. Currently supported is tailing via a date format/expression, e.g filter only to lines in which the date expression is equal or greater than a given start date.
qps Compute QPS averages given a log file using a datetime/timestamp field and a sliding window interval specified by the user. Can be very handy to quickly assess current QPS of some arbitrary service based on real-time logfiles (e.g Apache access_log, Tomcat catalina.out).
aggregate Convenient shortcut for aggregating values over a given field and sorting/counting by a value's frequency. See example below.
logplot Render a plot based on some fields/values from input log. This tool supports a pluggable backend interface, and currently includes an implementation for plotting using the Google Charts API as well as matplotlib. See examples below.
urlparse Parse URLs and extract specific fields from them (domain, path, query parameters), also decoding various URL encoding formats .
sumstat Display various summary statistics from a logfile/dataset given as a histogram (e.g (,) rows). This includes average, min/max, percentiles etc.
percentiles A quick and simple utility to compute percentiles from input numeric values.
flattenjson Extract a key from a JSON blob pointing to a list of items (e.g dictionaries) and output the dictionaries as individual JSON lines. Useful for bridging into tools such as json2csv

Configuration

All tools' command-line parameters can assume a default value using parameter interpolation from /etc/logtools.cfg and/or ~/.logtoolsrc, if these exist. This allows for convenient operation in the usual case where these rarely change, or when they assume one of a small set of different configurations (e.g usage profiles).

The basic configuration file format is of the form:

 [script_name]
 optname: optval

For example:

 [geoip]
 ip_re: ^(.*?) -

 [filterbots]
 bots_ua: /home/www/conf/bots_useragents.txt
 bots_ips: /home/www/conf/bots_hosts.txt
 ip_ua_re: ^(?P<ip>.*?) -(?:.*?"){5}(?P<ua>.*?)"

Available parameters per each command-line tool can be gleaned by running the tool with the --help flag.

Different configuration 'profiles' can also be specified. This is useful for cases where you have a set of common, distinct configurations that you'd like to keep around. All the tools admit a -P/--profile <profile_name> flag that will try to load default parameter values from a [<profile_name>] section in the aforementioned .ini files. e.g:

filterbots --profile fbots_accesslog

will look up default parameter values from the section [fbots_accesslogs] in ~/.logtoolsrc or /etc/logtools.cfg if that exists.

Usage Examples

See RECIPES

Credits

logtools was created by Adam Ever-Hadani

Contact

Adam Ever-Hadani

License

logtools is available under the Apache license 2.0. See the LICENSE file for more info.

adamhadani/logtools

adamhadani

Reviews

Repository Details