Open and Distributed Data Monitoring

Dependencies

• if python < 2.5, then we need argparse package
• pika (RabbitMQ client library)

Build binary rpm package

python setup.py bdist_rpm

OpenShift

If you wish to run the aggregator inside an OpenShift pod, most of the setup work has already been done. Take a look at the README file in the openshift directory.

Metric Plugins

There are a few requirements for a valid plugin extension:

• A plugin needs to be named starting with metric_.
• Plugins must define 4 functions:
  • def metric_init(name, config_file, is_subscriber = False, loglevel=logging.DEBUG)
  • def metric_cleanup( is_subscriber = False)
  • def get_stats()
  • def save_stats( msg)

Plugin Function Behavior

• metric_init() returns True if it initialized properly.
• get_stats() returns the data that is to be published. (Currently all plugins use JSON-encoded text, but that's probably not actually necessary.)
• The other functions return nothing.
• get_stats() is called by the publisher.
• save_stats() is called by the subscriber.
• metric_init() and metric_cleanup() are called by both.

The data that get_stats() returns (on the publisher side) is passed to save_stats() (on the subscriber side). It's up to save_stats() to do something useful with that data.

Data Archiving

TBD