Facette
Table of Contents

Configuration

The main configuration file is passed using -c argument when launching the facette server. If not specified, its default location is /etc/facette/facette.json.

All the configuration files are stored in JSON format, and must feature the extension .json.

Base Settings

* : Requires write permissions

Endpoint Binding Configuration

Facette back-end can listen on different types of endpoints:

Example:

{
    "bind": "tcp://localhost:12003",
    "base_dir": "/usr/local/share/facette",
    "data_dir": "/var/lib/facette",
    "pid_file": "/var/run/facette/facette.pid",
    "providers_dir": "/etc/facette/providers",
    "read_only": false,
    …
}

Catalog Setup

Configuring the catalog consists in defining data providers. At this point, we strongly recommend reading about the architecture of Facette to understand how the catalog works.

Definining Providers

Defining providers registers data origins in Facette, i.e. where to find sources and their metrics, and how to add them to the catalog. Providers are defined by a mandatory connector description, and optional filters. The configuration is stored in a dedicated file in the directory defined by the providers_dir setting in the main configuration file; you can have as many definition files as your want, for instance if your polling/collect system is sharded on many storage backends.

In case of a syntax error or incorrect configuration setting, the server will continue its initialization and discard the bogus provider (a warning message will be logged).

Here’s what a typical structure of a provider looks like:

$ cat /etc/facette/providers/my_metrics.json
{
	"connector": {
		…
	},

	"filters": [
		…
	],

	"refresh_interval": 300
}

The name of the configuration file will be used as origin of the sources/metrics retrieved from this origin in the catalog, i.e. in the example above “my_metrics”.

Connectors

Facette providers use connectors to know how to bind to origins and to inventory their local sources and metrics. See this page to view the list of available connectors and how to configure them.

Filters

By default the inventory mechanism of Facette providers retrieves all the known sources and metrics from the origin, named according to the origin’s storage structure/format. You may want to rename some of the sources/metrics or even discard some of them to keep the catalog relevant. Enter filters.

Optional Provider Settings

Example: collectd

Here is an actual provider configuration for using files created by the RRDtool collectd plugin, featuring a few filters to rewrite original metric names in a more friendly way and discard entropy-related metrics:

{
  "connector": {
    "type": "rrd",
    "path": "/var/lib/collectd/rrd",
    "pattern": "(?P<source>[^/]+)/(?P<metric>.+).rrd"
  },

  "filters": [
    { "action": "sieve", "target": "metric", "pattern": "/average$" },
    { "action": "rewrite", "target": "metric", "pattern": "/average$", "into": "" },
    { "action": "rewrite", "target": "metric", "pattern": "/", "into": "." },
    { "action": "rewrite", "target": "metric", "pattern": "^cpu-(\\d+)\\.cpu-(.+)\\.value$", "into": "cpu.$1.$2" },
    { "action": "rewrite", "target": "metric", "pattern": "^df-(.+)\\.df_complex-(.+)\\.value", "into": "df.$1.$2" },
    { "action": "rewrite", "target": "metric", "pattern": "^disk-(.+)\\.disk_(.+)", "into": "disk.$1.$2" },
    { "action": "rewrite", "target": "metric", "pattern": "^entropy\\.entropy", "into": "entropy" },
    { "action": "rewrite", "target": "metric", "pattern": "^interface-(.+)\\.if_(.+)\\.(.+)$", "into": "net.$1.$2.$3" },
    { "action": "rewrite", "target": "metric", "pattern": "^irq.irq-(.+)\\.value$", "into": "irq.$1" },
    { "action": "rewrite", "target": "metric", "pattern": "^load\\.load", "into": "load" },
    { "action": "rewrite", "target": "metric", "pattern": "^memory\\.memory-(.+)\\.value$", "into": "memory.$1" },
    { "action": "rewrite", "target": "metric", "pattern": "^processes\\.ps_state-(.+)\\.value$", "into": "proc.state.$1" },
    { "action": "rewrite", "target": "metric", "pattern": "^processes\\.(.+)\\.value$", "into": "proc.$1" },
    { "action": "rewrite", "target": "metric", "pattern": "^swap\\.swap-(.+)\\.value$", "into": "swap.$1" },
    { "action": "rewrite", "target": "metric", "pattern": "^swap\\.swap_io-(.+)\\.value$", "into": "swap.io.$1" },
    { "action": "rewrite", "target": "metric", "pattern": "^users\\.users\\.value", "into": "users.count" }
  ]
}

Example: Graphite

This configuration snippet shows a provider definition for using a Graphite webapp, with a filter to rewrite the underscore-escaped hostname dots back with dots; besides, this provider will self-refresh every hour.

{
  "connector": {
    "type": "graphite",
    "url": "https://graphite.mycorporation.net/",
    "allow_insecure_tls": true
  },

  "filters": [
    { "pattern": "_", "rewrite": ".", "target": "source" }
  ],

  "refresh_interval": 3600
}