https://github.com/ccnmtl/hound

Alert off graphite metrics
https://github.com/ccnmtl/hound

Last synced: 5 months ago
JSON representation

Alert off graphite metrics

• Host: GitHub
• URL: https://github.com/ccnmtl/hound
• Owner: ccnmtl
• Created: 2014-04-08T20:10:33.000Z (over 11 years ago)
• Default Branch: master
• Last Pushed: 2025-07-09T16:03:22.000Z (6 months ago)
• Last Synced: 2025-07-10T01:08:29.526Z (6 months ago)
• Language: Go
• Size: 861 KB
• Stars: 6
• Watchers: 5
• Forks: 2
• Open Issues: 8
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
[![Build Status](https://travis-ci.org/ccnmtl/hound.svg?branch=master)](https://travis-ci.org/ccnmtl/hound)

[![Coverage Status](https://coveralls.io/repos/github/ccnmtl/hound/badge.svg?branch=master)](https://coveralls.io/github/ccnmtl/hound?branch=master)

## Hound

This is a simple service that watches a number of Graphite metrics and

sends alert emails when they cross a threshold.

It automatically backs off on failing metrics. You'll get an email

when the metric first fails, another 5 minutes later, another 30

minutes after that, one hour after that, 2 hours after that, 4 hours,

8 hours, then every 24 hours thereafter. Finally, you will get an

email when the metric has recovered.

### Docker image

There is a docker image for running hound and postfix. See

,

 and

 for more information.

### Without docker

#### Dependencies

1. Obviously enough, hound needs a running graphite server, accessible via

   network.

2. In addition, an SMTP host is necessary (without authentication or

   encryption) to send the emails out.

### Configuration

There are a couple example configs in the `examples/` directory.

* `CheckInterval` is how many minutes to wait between checks

* `GlobalThrottle` is the maximum number of alerts that Hound will send in a

  cycle. Ie, if there's a major network outage and all the metrics start

  failing, you want to stop it once you've figured that out. Once this

  threshold is passed, Hound sends just one more message saying how many

  metrics are failing.

The rest of the values in this file should be self-explanatory.

The alerts configuration is set in `config.json` (by default - it is passed as

an argument to `hound` in `run_nohup.sh`).

Each Alert has:

* `Name`: obvious.

* `Type`: defaults to 'Alert'. You can also set it to 'Notice'. The

  convention is that an alert means "something is broken and a human

  needs to fix it NOW" while a Notice is just "you should know about

  this" and it's either not directly actionable or not urgent. Eg,

  high CPU load is probably a Notice, while high request error rates

  would warrant an alert.

* `Metric`: the actual Graphite metric being checked. This can be as

  complicated as you like and use the full suite of Graphite

  functions.

* `Threshold`: fairly obvious. Format it as a float. Treat it as ">="

  or "<=". Ie, it will trigger if the metric matches the threshold.

* `Direction`: "above" or "below". Specified whether a failure is when

  the metric crosses above or below the threshold, respectively.