Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with
or
.
Download ZIP
a java statistics and monitoring lib I put together by pillaging other open source libs and adding some stuff too.
Java Groovy Scala
branch: master

Fetching latest commit…

Cannot retrieve the latest commit at this time

Failed to load latest commit information.
pillage-core/src
pillage-groovy
pillage-scala
.gitignore
CHANGELOG
README.md
build.gradle
gradle-app.setting
init.gradle
jmeter.log
settings.gradle cleaned up groovy and scala projects. increased version to 0.3.4

README.md

Pillage

A super lightweight stats and monitoring lib for the jvm. It provides counters, metrics, timers, labels, gauges and jvm stats. The goals are to be lightweight, fast, make good use of resources and be flexible enough to easily hack on pieces when needed.

The library was put together by pillaging a number of other open source libs ( hence the name ) for their good parts and ideas then mixing them together with some new code and ideas.

Libraries pillaged thus far:

Components

Counters

Counters just count things. They can be incremented, read and reset.

Metrics

Metrics are tracked with a Distribution. The the only current implementation is based on the Histogram in the Ostrich scala lib. It provides max, min, mean within 5% and percentiles.

Labels

Labels are just string values. This also came from Ostrich and can be used to set flags instead of holding statistical value.

Gauges

Gauges are methods that can be functions that are executed when the stat is collected to get an reading at that instance.

StatsSummary

A StatsSummary contains Counters, Metrics and Labels for a set amount of time. A summary is an immutable object and will throw exceptions if you try to mess with it.

StatsContainer

A StatsContainer is the client interface to interact with Pillage. Through this interface you can increment counters, add metrics, set labels, acquire summaries, etc.

StatsCollector

The StatsCollector interface is used to collect stats from the container for purposes of reporting or aggregation. A collector will cache a set of stats from a StatsSummary and when the collect() method is called, it will acquire a new StatsSummary object and calculate the diff between the cached version and the new version and return that diff as a new StatsSummary object. The collect() method is how you can see change over time which is good for graphing.

The stats collector provides the option to include jvm stats which is off by default. If this option is turned on then stats from the MBeans provided with java will be included the StatsSummarys. This is a good way to get heap usage, threads counts, classes loaded/unloaded, gc counts etc without having to mess with JMX.

A stats collector can have listeners attached called StatsReporters that will receive the delta StatsSummary when collect() is called.

StatsReporter

A StatsReporter has just one method report that takes in a StatsSummary. There is currently one implementation, GangliaStatsReporter which will send the stats to a Ganlgia server over UDP.

PeriodicStatsCollector

This is a utility singleton that can be used to schedule regular calls to collect() for a given StatsCollector.

Wrapper Projects

There are two wrapper projects that provide some groovy and scala syntax sugar. They allow timing with closures and registering gauges with out having to use the Gauge interface.


TODO

  • create some other reporters ( logging, graphite )
  • make calls to report() non blocking. A call to report will most likely involve IO so if it hangs I don't want it to cause a backup or other bad things.
  • add gauges through reflection. This is a really cool feature in Ostrich but since java doesn't have closures it is a bit more difficult to implement.
  • standardize the stats naming, I think it is a bit all over the place right now.
  • enable PeriodicStatsCollector to be restarted. Right now if the thread dies the app will have to be restarted to work on reporting.
  • try alternate metric backings. Histogram is cool but I would like to collect some other things like stddev, skew etc.
  • Profiled annotation like Perf4js
  • get some production usage and iterate!
Something went wrong with that request. Please try again.