Diamond is a python daemon that collects system metrics and publishes them to Graphite. It is capable of collecting cpu, memory, network, i/o, load and disk metrics. Additionally, it features an API for implementing custom collectors for gathering metrics from almost any source.

Dependencies

• CentOS or Ubuntu
• Python 2.4+
• python-configobj
• pysnmp
• pyasn1
• ant (for packaging)

To install diamond, use ant:

ant install

For testing, diamond can also be started directly via ant without installing:

ant run

The run task will invoke diamond in debug mode for testing.

Ant can also build packages for CentOS and Ubuntu.

ant package

sudo dpkg -i build/diamond-2.0.1-0.deb

The package task will detect Ubuntu or CentOS and build .debs or .rpms.

ant tar

The tar task will build a tarball if thats your thing.

If you've installed diamond via a package, the configuration file is /etc/diamond/diamond.cfg. By default, diamond will push to a graphite server host "graphite". You should probably change this to point to your own graphite server.

Other configuration should not be necessary.

By default diamond publishes metrics using the following form:

systems.<hostname>.<metrics>.<metric>

You can override the "systems" portion of the metric path by changing the "path_prefix" setting in the configuration file.

• NetworkCollector
• CPUCollector
• MemoryCollector
• LoadAverageCollector
• IOCollector
• VMStatCollector
• DiskSpaceCollector
• TCPStatsCollector

Diamond collectors run within the diamond process and collect metrics that can be published to a graphite server.

Collectors are subclasses of diamond.collector.Collector. In their simplest form, they need to implement a single method called "collect".

import diamond.collector

class ExampleCollector(diamond.collector.Collector):
   
    def collect(self):
    """
    Overrides the Collector.collect method
    """
        # Set Metric Name
        metric_name = "my.example.metric"

        # Set Metric Value
        metric_value = 42

        # Publish Metric
        self.publish(metric_name, metric_value)

To run this collector in test mode you can invoke the diamond server with the -r option and specify the collector path.

python src/diamond/server.py -f -v -r examples/examplecollector.py

Diamond supports dynamic addition of collectors. Its configured to scan for new collectors on a regular interval (configured in diamond.cfg). If diamond detects a new collector, or that a collectors module has changed (based on the file's mtime), it will be reloaded.

Diamond looks for collectors in /usr/lib/diamond/collectors/ (on Ubuntu). By default diamond will invoke the collect method every 60 seconds.

Diamond collectors that require a separate configuration file should place a .cfg file in /etc/diamond/. The configuration file name should match the name of the diamond collector class. For example, a collector called examplecollector.ExampleCollector could have its configuration file placed in /etc/diamond/ExampleCollector.cfg.

Maintainer: Andy Kipp