System

System statistics collector for statsd on linux written in Go .

Usage

  Usage:
    system-stats
      [--statsd-address addr]
      [--memory-interval i]
      [--disk-interval i]
      [--cpu-interval i]
      [--extended]
      [--name name]
    system-stats -h | --help
    system-stats --version

  Options:
    --statsd-address addr   statsd address [default: :8125]
    --memory-interval i     memory reporting interval [default: 10s]
    --disk-interval i       disk reporting interval [default: 1m]
    --cpu-interval i        cpu reporting interval [default: 2s]
    --name name             node name defaulting to hostname [default: hostname]
    --extended              output additional extended metrics
    -h, --help              output help information
    -v, --version           output version

Installation

Via go-get:

$ go get github.com/statsd/system

Via binaries:

Coming soon!

Metrics

Metrics are prefixed with the hostname (or --name), and namespaced by the resource, for example:

api-2.cpu.blocked:7|c
api-2.cpu.running:4|c
api-2.cpu.interrupts:19695796035|c
api-2.cpu.percent:26|g
api-2.cpu.switches:25195265352|c
api-2.cpu.processes:20027|c
api-2.cpu.blocked:7|c
api-2.cpu.running:4|c
api-2.cpu.interrupts:19695796035|c
api-2.cpu.percent:26|g
...

Depending on the statd implementation that you use this may result in different outputs. For example with the armon/statsite implementation this would result in gauges.api-2.memory.free.

The --extended flag enables extended metrics per resource and are listed as extended below.

CPU

• cpu.percent gauge
• cpu.switches counter extended
• cpu.interrupts counter extended
• cpu.blocked counter extended

Memory

Memory values are represented in bytes.

• memory.percent gauge
• memory.used gauge
• memory.active gauge extended
• memory.total gauge extended
• memory.free gauge extended
• memory.swap.percent gauge
• memory.swap.total gauge extended
• memory.swap.free gauge extended

Disk

Disk values are represented in bytes. <volume> is the path the fs is mounted on (/, /data, etc).

• disk.<volume>.percent gauge
• disk.<volume>.free gauge
• disk.<volume>.used gauge

IO

Coming soon!

Daemonization

system-stats(1) doesn't support running as a daemon natively, you'll want to use upstart or similar for this. Add the following example upstart script to /etc/init/system-stats.conf:

respawn

start on runlevel [2345]
stop on runlevel [016]

exec system-stats --statsd-address 10.0.0.214:5000

Then run sudo start system-stats and you're good to go!

Debugging

Run with DEBUG=stats to view the go-debug output:

2014-08-13 22:04:36 INFO - cpu: reporting
22:04:36.098 2s     2s     statsd - vagrant-ubuntu-precise-64.cpu.switches:20384|c
22:04:36.098 4us    4us    statsd - vagrant-ubuntu-precise-64.cpu.processes:0|c
22:04:36.098 3us    3us    statsd - vagrant-ubuntu-precise-64.cpu.running:0|c
22:04:36.098 3us    3us    statsd - vagrant-ubuntu-precise-64.cpu.interrupts:656|c
22:04:36.098 3us    3us    statsd - vagrant-ubuntu-precise-64.cpu.percent:100|g
2014-08-13 22:04:38 INFO - cpu: reporting
22:04:38.098 2s     2s     statsd - vagrant-ubuntu-precise-64.cpu.switches:24074|c
22:04:38.098 23us   13us   statsd - vagrant-ubuntu-precise-64.cpu.processes:0|c
22:04:38.098 15us   8us    statsd - vagrant-ubuntu-precise-64.cpu.running:1|c
22:04:38.098 12us   7us    statsd - vagrant-ubuntu-precise-64.cpu.interrupts:638|c
22:04:38.099 11us   7us    statsd - vagrant-ubuntu-precise-64.cpu.percent:100|g

License

MIT