A Graphite Dashboard, massively inspired by the excellent GDash made by ripienaar

It was rewritten in Python.

Differences with Gdash :

• Configuration is in full YAML instead of a DSL.
• Ability to display multiple dashboards on a single page for comparison
• Zoom in/out
• Graph inclusion allows globbing

This dashboard is a Flask application. Have a look at this documentation if you want to deploy it in a production environment.

If you just want to play with it, simply install dependencies and launch the run script:

# pip install pyyaml
# pip install flask
# ./run.py

If you want to run tests:

# pip install nose
# nosetests

Main configuration example is included in config/leonardo.yaml.example, you should rename it to leonardo.yaml and adjust the main options here.

Two config options are mandatory in leonardo.yaml:

* graphite: put your Graphite server url there
* templatedir: directory where your dashboards are.

and additional options:

* The title to show at the top of your Dashboard
* How many columns of graphs to create, 3 by default.
* How often dashboard page is refreshed, 30 sec by default.
* The width of the graphs, 450 by default
* The height of the graphs, 220 by default
* Optional interval quick filters

You can have multiple top level categories of dashboard. Just create directories in the templatedir for each top level category.

In each top level category create a sub directory with a short name for each new dashboard.

You need a file called dash.yaml for each dashboard, here is a sample:

name: Server-1
description: Main System Metrics for server-1

Then create descriptions in files like cpu.graph in the same directory, in pure YAML. here is a sample:

title: Combined CPU Usage
vtitle: percent
area: stacked
description: The combined CPU usage

fields:
    iowait: 
        color: red
        alias: IO Wait
        data: sumSeries(collectd.server-1.cpu.*.cpu.wait.value)
        cacti_style: si

    system:
        color: orange
        alias: System
        data: sumSeries(collectd.server-1.cpu.*.cpu.system.value)
        cacti_style: si

    user:
        color: 4169E1
        alias: User
        data: sumSeries(collectd.server.cpu.*.cpu.user.value)
        cacti_style: si

The directory layout is such that you can have many groupins of dashboards each with many dashboards underneath it, an example layout of your templates dir would be:

    graph_templates
    `-- virtualization
        |-- dom0
        |   |-- dash.yaml
        |   |-- iowait.graph
        |   |-- load.graph
        |   |-- system.graph
        |   |-- threads.graph
        |   `-- user.graph
        `-- kvm1
            |-- dash.yaml
            |-- disk_read.graph
            |-- disk_write.graph
            |-- ssd_read.graph
            `-- ssd_write.graph

Here we have a group of dashboards called 'virtualization' with 2 dashboards inside it each with numerous graphs.

You can create as many groups as you want each with many dashboards inside.

You can reuse your dashboards and adjust the time interval by using the following url structure:

http://gdash.example.com/System/Server-1/?from=-8d&until=-7d

This will display the email dashboard with a time interval same day last week.

Quick interval filters shown in interface are configurable in leonardo.yaml options sections. Eg:

options:
  interval_filters:
    - label: 30 minutes
      from: -30min
      to: now
    - label: 24 hours
      from: -1day
    - label: 1 week
      from: -7day
    - label: 1 month
      from: -30day
    - label: 1 year
      from: -1year

If you configure time intervals in the config file you can click on any graph in the main dashboard view and get a view with different time intervals of the same graph

options:
    intervals:
        - [ "-1hour", "1 hour" ]
        - [ "-2hour", "2 hour" ]
        - [ "-1day", "1 day" ]
        - [ "-1month", "1 month" ]
        - [ "-1year", "1 year" ]

With this in place in the leonardo.yaml clicking on a graph will show the 5 intervals defined above of that graph

You can reuse your dashboards for big displays against a wall in your NOC or office by using the following url structure:

http://gdash.example.com/System/Server-1/?full=yes

This is how it looks :

You can include the graphs from other dashboard with the include property in dash.yaml. You can also use globbing :

include_graphs:
- templates/os.basic
- templates/os.nfs
- System/Server-*/cpu.graph

If you got a set of common properties that you want to reuse in the dashboard, you can load a external yaml file from in dash.yaml. The path is relative to the templatedir and it does not support recursive includes.

Examples are a list of server colors, timezones, etc. In dash.yaml:

include_properties: 
 - common.yml
 - black-theme.yml

Example common.yml:

graph_properties: 
    timezone:         Europe/London
    hide_legend:      false

Example black-theme.yml:

graph_properties: 
    background_color: white
    foreground_color: black
    vertical_mark_color: "#330000"

I wrote Leonardo to be able to add features that I find useful in a production environment :

You can use regexps in the search box to display multiple Dashboards at the same time. That way you can compare dashboards with each other, which is very useful for instance when comparing several server metrics :

Using the + / - buttons on a dashboard, you can zoom in or out. The number of columns will adapt automatically :

The logging is configurable via options in the config/leonardo.yaml. There are options for logging to stdout, files, syslog, and email.

Examples:

• Logging to stdout

logging:
  - stdout:
      level: debug

• Logging to a file

logging:
  - file:
     filename: /path/to/log/file
     level: info

• Logging to syslog

logging:
  - syslog:
      level: debug

• Logging to email

logging:
  - email:
     mailhost: example.com
     fromaddr: server-error@example.com
     toaddrs:
       - admin1@example.com
       - admin2@example.com
     subject: example message
     level: error

Add "theme: dark" to options in leonardo.yaml to switch Leonardo to dark mode :

options:
	theme: dark