This repository holds the go-newrelic-plugin which combines collectors into one binary and makes them available to the newrelic-infra agent.
There are two parts to the architecture of the plugin.
Commands live under the top level folder cmd These files are all apart of the same package cmd each command should have a corresponding collector(more on that later). We are using a package called cobra to parse the commands and flags. This allows us to bundle all the collectors into one binary and also gives us an awesome help command.
-$ go run main.go help
A set of plugins to integrate custom checks into the newrelic infrastructure
Usage:
go-newrelic-plugin [command]
Available Commands:
couchbase execute a couchbase collection
help Help about any command
haproxy execute a haproxy collection
jenkins execute a jenkins collection
kraken execute a kraken collection
mongo execute a mongo collection
nginx execute an nginx collection
rabbitmq execute a rabbitmq collection
redis execute a redis collection
saucelabs execute a saucelabs collection
version Print the version of go-newrelic-plugin
Flags:
--pretty-print pretty print output
--verbose verbose output
All of the commands besides root.go follow the same basic pattern. Import your collector and call the Run function. You can model your command function off of the skel.go command. Just make sure you update the Use and Short keys. Use is the name of the command and it should match the name of your collector. Short is a description of your collector. Both of these will show up in the help command output.
Collectors are designed to collect the stats for a given technology and report back to the newrelic infrastructure app. In general, collector development is where contributors will be spending their time.
Each collector is its own package. Take a look at the skel package The entry point to this package is Run(log *logrus.Logger, prettyPrint bool, version string)
Your collector's Run method will be called everytime New Relic requests stats.
Once your function is created, you can begin development of the logic for collecting and reporting stats of your specific technology.
New Relic will pass in environment variables that you configure through a yaml config file. See ./skel/skel.yaml for an example
name: 'Name of your collector'
description: 'Short Description of your collector'
protocol_version: 'New Relics collection protocol'
os: 'Os this collector supports'
source:
- command:
- 'Location of your binary not really import for collector development'
prefix: 'Application prefix for newrelic'
interval: 'How often the agent checks for stats'
env:
KEY: "VALUE"
The important thing to note with the config is the env section. All of your config values should go here.
Important: In order to test your application you should export the variables you setup in your ~/.profile of ~/.bash_profile
Your collector should be named after the technology you are gathering metrics for. If you were developing nginx, you collector would live in a file called nginx.go and live in a folder nginx
Your collectors module should export a function called Run and accepts 3 parameters Run(log *logrus.Logger, prettyPrint bool, version string)
In the event that your collector has an error in retrieving stats and you are unable to report stats back, you should os.Exit(-1) or anything but zero to tell the newrelic agent their was an issue and to disregard any reported stats.
Here you will find the infrastructure Plugins and Agents SDK Draft This document outlines the extensibility mechanism built into the New Relic Infrastructure (NRI) Agent that allows you to add new sources of data to your Infrastructure account.
- Create a directory and go file, named after your collector. You can copy ./skel as a starting point
- Follow the function naming standards as defined above
- Add your metrics to the plugin data interface
- Update the following README sections
- Available Collectors
- Submit a PR