metrics-graphite-pickle

An extension to Coda Hale's excellent Metrics library. It provides GraphitePickleReporter, which allows your application to stream metric values to a Graphite server using graphite's pickle receiver. This is recommended if sending large amounts of data:

GraphitePickleReporter.enable(1, TimeUnit.MINUTES, "graphite.example.com", 2004);

Setting up Maven

To add metrics-graphite-pickle to your project:

<dependency>
  <groupId>com.brightcove.metrics</groupId>
  <artifactId>metrics-graphite-pickle</artifactId>
  <version>1.2.0</version>
</dependency>

Change Log

1.2.0

• Stop using jython for metrics pickling. Hand-coded minimal pickling of the primitive data types. Jython unpickling remains in use during tests.

1.1.0

• Modify the GraphitePickleReporter to use executor.scheduleAtFixedRate rather than executor.scheduleWithFixedDelay (will be unnecessary after upgrading to metrics 3.0.0)
• Upgrade to metrics-core-2.2.0
• Upgrade to metrics-graphite-2.2.0
• Upgrade to jython-standalone-2.5.3
• Upgrade to slf4j 1.7.2

1.0.5

• Fixed a bug where a null prefix caused the last word in a metric name to be lost in some cases (e.g. 'mean')
• Upgrade to metrics-graphite-2.1.3

1.0.4

• Move the creation of MetricsPickler back into the run method but use a synchronized method to ensure it is only created once. This improves startup time by a few seconds in some cases.

1.0.3

• Moved the creation of the jython script engine out of the reporter's "run" method. This should fix a memory leak caused by ThreadLocal entries that never got cleaned up.
• Upgrade to metrics-graphite 2.1.2

1.0.2

• Changed encoding to ISO-8859-1 to match the python default

1.0.1

• Fixed a bug where a batchSize passed to enable was ignored.
• Recreates the socket on each batch to prevent reaching the 100K byte limit (and a broken pipe)

1.0.0

Initial release.