Noderig collect OS metrics and expose them through a Sensision HTTP endpoint. Each collector is easily configurable, thanks to a simple level cursor

Noderig metrics:

• CPU
• Memory
• Load
• Disk
• Net
• External collectors

Noderig is currently under development. Feel free to comment or contribute!

Noderig is pretty easy to build.

• Clone the repository
• Install glide, follow instructions here https://glide.sh/
• Download dependencies glide install
• Build and run go run noderig.go

noderig [flags]

Flags:
      --config string       config file to use
  -l  --listen string       listen address (default "127.0.0.1:9100")
  -v  --verbose             verbose output
      --period uint         default collection period (default 1000)
      --cpu uint8           cpu metrics level (default 1)
      --disk uint8          disk metrics level (default 1)
      --mem uint8           memory metrics level (default 1)
      --net uint8           network metrics level (default 1)
      --load uint8          load metrics level (default 1)
  -c  --collectors string   external collectors directory (default "./collectors")
  -k  --keep-for uint       keep collectors data for the given number of fetch (default 3)
      --net-opts.interfaces give a filtering list of network interfaces to collect metrics on
      --disk-opts.names     give a filtering list of disks names to collect metrics on

Noderig have some built-in collectors.

With Noderig you can define set-up custom collectors as defined in http://bosun.org/scollector/external-collectors. To be enable you need to define a collectors folder using the noderig parameter "collectors". This fold need to have a strict arborescence: a number folder and then the exectutable collectors.

For example to define a script shell collectors reach the noderig collectors file:

cd ~/collectors
mkdir 10

Then inside the 10 folder write the following executable test.sh shell script.

#!/bin/sh

now=$(date +%s)

echo my.metric $now 42

And execute noderig:

./build/noderig --collectors ~/collectors

To conclude you can tun noderig custom collectors with the following configuration parameters:

keep-metrics: true # To always keep in Noderig the last metrics values
keep-for: 3 # Keep-for returned the number values to keep

The keep-for parameter with keep-metrics at true keep the last N values otherwise it keep each values for n calls to the noderig metrics endpoint.

Noderig can read a simple default config file.

Configuration is load and override in the following order:

• /etc/noderig/config.yaml
• ~/noderig/config.yaml
• ./config.yaml
• config filepath from command line

Config is composed of three main parts and some config fields:

Noderig have some built-in collectors. They could be configured by a log level. You can also defined custom collectors, in an scollector way. (see: http://bosun.org/scollector/external-collectors) To configure a custom collectors in noderig reach custom collectors.

cpu: 1  # CPU collector level     (Optional, default: 1)
mem: 1  # Memory collector level  (Optional, default: 1)
load: 1 # Load collector level    (Optional, default: 1)
disk: 1 # Disk collector level    (Optional, default: 1)
net: 1  # Network collector level (Optional, default: 1)

Some collectors have additionals modules. Add module to <collector>-mods list to enable them.

cpu-mods:
  - temperature

Some collectors can accept optional parameters.

net-opts:
  interfaces:            # Give a filtering list of interfaces for which you want metrics
    - eth0
    - eth1

Net-opts, interfaces field support now regular expression to white-list interface based on golang MatchString implementation. However to use a regular expression you need to prefix the string value by a ~. To whitelist all eth interfaces, you can set:

net-opts:
  interfaces:            # Give a filtering list of interfaces for which you want metrics
    - ~eth*

disk-opts:
  names:            # Give a filtering list of disks for which you want metrics
    - sda1
    - sda3

Disk-opts, names field support now regular expression to white-list disks names based on golang MatchString implementation. However to use a regular expression you need to prefix the string value by a ~. To whitelist all disk names, you can set:

disk-opts:
  names:            # Give a filtering list of disks names for which you want metrics
    - ~disk*

Noderig can be customized through some parameters.

period: 1000             # Duration within all the sources should be scraped in ms (Optional, default: 1000)
listen: none             # Listen address, set to none to disable http endpoint    (Optional, default: 127.0.0.1:9100)
collectors: /opt/noderig # Custom collectors directory                             (Optional, default: none)

To force default labels to each metrics in Noderig, you can set up a configuration key called labels. It expects a label string map as defined below:

labels: { 
  host: "srv001", 
  dc: "uk1", 
  type: "web_server", 
}

1484828198557102// os.cpu{} 2.5202020226869237
1484828198560976// os.mem{} 24.328345730457112
1484828198560976// os.swap{} 0
1484828198557435// os.load1{} 0.63
1484828198561366// os.net.bytes{direction=in} 858
1484828198561366// os.net.bytes{direction=out} 778
1484828197570759// os.disk.fs{disk=/dev/sda1} 4.967614357908193

To use Noderig and expose a Prometheus native format, just set the following two configuration lines in the config file:

format: "prometheus"  # Expose a Prometheus format in Noderig as: https://prometheus.io/docs/instrumenting/exposition_formats/
separator: "_"        # Metrics classnames separator, '_' is the default one for Prom, but you can use any other supported by your storage backend

Instructions on how to contribute to Noderig are available on the Contributing page.

• Twitter: @notd33d33