Earlier this week I was at the Monitorama conference in Portland. This was the third Monitorama but the first that I'd attended, and if you're interested in operations and monitoring issues I can't recommend it highly enough, the signal to noise ratio was very nearly 100%. Biggest takeaway is to never miss this conference again.
This is a quick list of things I noted at the conference, and is basically meant as a reminder to myself to write about the topics in more detail when I get the chance.
Observations:
- There are people still running Nagios at scale. There is no one who likes it.
- MongoDB is teh eebul.
- Got a new monitoring utility you’re writing? The new hot TLD to register it under is .io
- Attendance was predominantly male, but there is hope for the industry, we have role models like Ashe Dryden, Selena Deckelmann, Bridget Kromhout, Katherine Daniels and Jen Andre.
Gold Nuggets:
- Metrics 2.0
- Sensu is the new event engine of choice for monitoring.
- Separate your eventing from your front-end interface and notifications system - use Flapjack
- If you’re not writing monitoring and analytics tools in Go, what’s wrong with you?
- Graphite is where everyone stores metrics, mention OpenTSDB and you get confused looks.
- Everything you know about anomaly detection is wrong.
- Metrics streams and audio streams are the same, you can use DSP algorithms on your metrics.
- Want a NoSQL database to back your monitoring tool? Use Redis.
- Threat detection can be boiled down to a yes/no question - are you being attacked by the Mossad?
- There are two keys to maintain stability and prevent security incidents:
- You can learn everything you need to know about failed deployments from the Vasa case study - http://www.slideshare.net/petecheslock/17th-century-shipbuilding-and-your-failed-software-project-34363385
- Feedback - It's harder than you think