Overview

This code package contains the source for RainMon. In brief, it is a system for decomposing, summarizing, and predicting timeseries systems timeseries monitoring data. For more detail on the technique, please see the KDD 2012 paper. The description below is a quick guide to some of the key parts of the code.

Installation

For a guide on how to install these requirements, please see the Setup Guide

Base requirements (all Python-based):

• Python (tested on 2.7)
• http://www.python.org/getit/releases/2.7.2/
• scipy/numpy
• see http://www.scipy.org/Installing_SciPy
• If you don't already have these, consider using the Enthought Python distribution: http://www.enthought.com/products/epd.php
• matplotlib (for producing figures)
• see http://matplotlib.sourceforge.net/users/installing.html

Required for interface:

• Django
• The database portion is not required (using SQLite is fine)
• https://docs.djangoproject.com/en/dev/topics/install/?from=olddocs
• RabbitMQ (though other AMQP brokers compatible with Celery should work)
• http://www.rabbitmq.com/download.html
• Note: requires Erlang (the installation will prompt for it)
• Leave the server running as a daemon
• Celery and django-celery for distributed task scheduling:
• easy_install celery
• easy_install django-celery

Configuration

config.json

First, copy config.template.json to config.json file in this directory, and edit the following keys:

• "rrddir": "...path to directory with .rrd files"
• "tsdbhost": "... host/IP that default timeseries database is located on ..."
• "tsdbport": "... port of default timeseries database (4242 for OpenTSDB, 8000 for RRD)"

RainMon currently supports two timeseries databases: OpenTSDB, a scalable distributed database, and an RRD server.

OpenTSDB

For detail on how to set up OpenTSDB, please see the OpenTSDB "getting started" guide at http://opentsdb.net/getting-started.html. The Haodop dataset is available only from within the CMU campus network (or on VPN).

RRDtool

The rrd folder contains a server for RRDtool-format data.

Running

First, cd to the code folder.

From the command line

A simple example of how to only run the analysis pipeline without the UI is located in boot.py. Look inside generate(): it shows what needs to be specified (name to save, nodes, attributes, and time range).

Running the UI

To run the UI:

• make celerystart
• starts up a task node on the current machine. For the single-node configuration, just run this on the same node as the Django server.
• make serverstart
• starts the Django server

Then, navigate with a web browser to http://localhost:8000/, where localhost is the IP of the machine you ran the make commands from.

Disclaimer

This is "research" code - we fully expect that there are bugs.

For licensing information that applies to the project, see LICENSE.