serverfarmer / sf-monitoring-newrelic

Old, deprecated extension to Server Farmer.

Home Page:http://serverfarmer.org/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Deprecated

This repository contains old, deprecated extension to Server Farmer.

History of New Relic support in Server Farmer

2012-2015

New Relic was used by Server Farmer authors since around 2012-13 - in fact, we were its early adopters in Allegro.pl (of paid version, at larger scale). We even got a few T-shirts from them :)

We were so satisfied of their services, that we also wanted to implement them (this time in free version) as an automated service discovery, monitoring and alerting platform for our side customers.

Initial implementations for particular customers were made in 2013 - and the final, universal implementation in Server Farmer, meeting its code quality and separation from customer-specific stuff, was finished around 2014-15.

We also created a few New Relic Plugins, integrated with Server Farmer:

2017-2018

In 2017 we had New Relic Servers (now referred as the "old" platform), with basic alerting, enabled on over 1000 hosts:

  • for many various customers
  • using many separate New Relic accounts (our own + per customer)
  • with completely different alerting policies and integrations for each customer
  • for Linux and Windows
  • 100% free (with 1 exception: 1 customer paid for 2 hosts for 1 month, to solve a particular problem with MySQL)

Unfortunately, in 2017, after gaining a sufficiently large group of paid clients, New Relic started to cut down their really-free services and replacing them with either 30-day free trials, or (later) services with limited free volume. It was first announced here, and our protests were completely ignored:

https://discuss.newrelic.com/t/important-upcoming-changes-for-new-relic-servers-and-legacy-alerting-features/49474

Our reaction

We decided to replace New Relic Servers with combination of 3 separate services:

  1. Heartbeat subproject (implementation finished in 2018) - for service discovery, and heartbeat service itself. So each server reports to Heartbeat every 2 minutes, which services were detected to be alive.

  2. Any internal or external monitoring service (possibly separate for each customer) - to query Heartbeat server and send email/push/Slack/Teams alerts about dead services. We choosed UptimeRobot, since back then, it was almost 10x cheaper than competing StatusCake and Pingdom (at least for bigger plans), while having enough quality.

  3. Cacti for recording metrics, where it was really required.

We also decided to drop most raw metrics and convert them to simple boolean alive/dead statuses on-the-fly in Heartbeat - eg. a hard drive either meets basic quality requirements, or not (see the details).

At that time, we didn't touch New Relic Plugins.

2019-2021

In early 2021, New Relic made another announcement, about New Relic Plugins:

https://discuss.newrelic.com/t/new-relic-plugin-eol-wednesday-june-16th-2021/127267

Fortunately, we managed to migrate all remaining monitoring services from New Relic to other platforms yet in 2019, so it didn't affect our customers at all.

2022

In 2022, as a part of the new roadmap for Server Farmer, we plan to finally remove several features deprecated in past years, including anything related to New Relic.

About

Old, deprecated extension to Server Farmer.

http://serverfarmer.org/

License:MIT License


Languages

Language:Shell 100.0%