監視

Purpose

Kanshi watches your Postgres database and reports metrics to the log stream. You can then drain this log to an application (or provide your own reporter class) to send these metrics to a service, like Librato or Splunk.

Installation

Kanshi will work as a gem or as a standalone app.

Add to existing app on Heroku

  1. Add Kanshi to your Gemfile:

       gem 'kanshi', :require => false

  2. Add Kanshi to your Procfile:

       kanshi: bundle exec kanshi

  3. Run bundle install, commit, and deploy.

  4. heroku scale kanshi=1

Create as separate Heroku app

  1. Create a new app.
  2. Follow steps 1-3 from above.
  3. Add environment variables for each database URL you want to monitor.
  4. heroku scale kanshi=1

Custom install

You can create your own version of bin/kanshi to run Kanshi in any way of your choosing, e.g. monitor specific databases or provide your own reporter for output. An easy way to do this would be to create a Rake task to run Kanshi.

For more information, see the Configuration section below and the source code for bin/kanshi.

Configuration

If using the kanshi binary, configuration is optional but is done through environment variables:

  • Kanshi will monitor all postgres:// URLs in your environment.
  • KANSHI_SAMPLE_DELAY will change how often Kanshi monitors your database in seconds. Default: 300
  • KANSHI_PREFIX will change the prefix used on metrics. Default: kanshi

Environment variables have no effect on the Kanshi library code, so if executing Kanshi manually, configuration is done via the options hash passed to Kanshi.run:

  • :databases: A hash of database names to URLs (=> String)
  • :delay: The delay between samples (Fixnum). Default: 300.
  • :reporter: A Class that responds to #report to recieve metrics (Class). Default: Kanshi::ScrollsReporter.

Name

Kanshi (監視) is Japanese for surveillance.