chronicreplay

Replays Logfiles (especially Apache logfiles) against a given target host considering the original time delays between requests

How to run

Download the latest release, unzip it and run it:

wget https://github.com/paulwellnerbou/chronicreplay/releases/download/v1.5.0/chronicreplay-1.5.0.zip
unzip chronicreplay*.zip
./chronicreplay-*/bin/chronicreplay --host=host.to.replay.against --logfile=/path/to/your/logfile --logparser=combined

Where "combined" is the name of the logfile reader exptecting Apache's combined log format (see Apache's documentation and CombinedLogFormatLogLineReader's JavaDoc).

Standard log level is set as environment variable, named LOG_LEVEL, default value is DEBUG. To change the log level, that environment variable needs to be altered. Available relevant levels are INFO and WARN. To disable logging completely, use WARN (or higher).

LOG_LEVEL=warn ./chronicreplay-*/bin/chronicreplay --host=host.to.replay.against --logfile=/path/to/your/logfile --logparser=combined

How to build

./gradlew build

How to run from source directory with gradle

./gradlew run -Pargs='--host=http://localhost --logfile=src/test/resources/combined-log-example.log --logparser=combined'

How to build a binary distribution

./gradlew installDist

The distribution will be put into build/install. For more options about packaging and distribution, see http://www.gradle.org/docs/current/userguide/application_plugin.html.

To run the application, you will find the start scripts for Windows and Unix/Linux in build/install/chronicreplay/bin.

How to write a custom log line parser

The selection of the logfile reader (implementation of LogLineReader) is done using Java's ServiceLoader. See http://docs.oracle.com/javase/7/docs/api/java/util/ServiceLoader.html for more information.

The steps you have to follow:

• Create your own implementation of LogLineParser
• Add the classname to src/main/resources/META-INF/services/de.wellnerbou.chronic.logparser.LogLineParser
• Use --logparser=ID-OF-YOUR-IMPLEMENTATION starting the program

You don't have to modifiy the original services file for this purpose, you can create a separate project which depends on chronicreplay and add your services file in this project. Or, the other way round, you add a (runtime) dependency to your custom LogLineReader implementation project.

For a basic example, have a look at de.wellnerbou.chronic.logparser.CommonLogFormatLogLineParser.

How to get duration comparisons in your chronicreplayer's logfiles

Apache does not log the duration of the requests in the default log formats. To achieve this, you have to add '%D' to your LogFormat. See (http://httpd.apache.org/docs/current/mod/mod_log_config.html) for more information of customizing your LogFormat.

To parse the microseconds you will get then, you can use the code in CombinedWithDurationLogLineParser or you can extend from this class, overwriting the extractDuration method or/and the parseLine method.

Plotting the results with gnuplot

The CSV files created contain all data you will need to plot. The file is tab separated, which is the default data format for gnuplot.

In most cases you are mainly interested in the results containing the same status code. That's why the responses with the same status as the original ones are logged additionally in another file, ending with .sameStatus.csv.

Plotting the durations over time with gnuplot:

gnuplot> set timefmt "%s"
gnuplot> set format x "%H:%M:%S"
gnuplot> set xdata time
gnuplot> plot "chronicreplay-host_2015-08-21_08-48-42.csv" using 1:6

Overlaying results from different servers in one plot:

gnuplot> set timefmt "%s"
gnuplot> set format x "%H:%M:%S"
gnuplot> set xdata time
gnuplot> plot "chronicreplay-host1_2015-08-21_08-48-42.csv" using 1:6, "chronicreplay-host2_2015-08-21_08-48-42.csv" using 1:6