Prometheus by design implements a pull-based approach for collecting metrics. For most of us, this is a departure from the push-based monitoring systems we’ve been using for the past decade.
Over the past six months, Fastly has migrated away from its legacy push-based monitoring system and has deployed Prometheus as its primary system for infrastructure and application monitoring. The new Prometheus approach posed some unique challenges to the company’s architecture and thinking, and its migration hasn’t been completely smooth sailing. Deploying Prometheus across a globe-spanning network serving over 10% of the world’s internet traffic has raised its fair share of technical challenges.
The journey taught the company a number of lessons that it is evolving into new patterns for application and infrastructure instrumentation. Marcus Barczak explains how Fastly navigated the path to Prometheus from its pull-based systems, shares the patterns the company has adopted and the mistakes made along the way, and discusses how the new system has been received by Fastly’s teams.
Marcus is a Senior Principal Engineer at Fastly where he works on the Platform Engineering team. Having first cut his teeth on MRTG back in the day through to exploring new ways of drawing insight out millions of metrics at Etsy. Marcus loves helping people better understand how their software runs wild in production.
©2018, O’Reilly UK Ltd • (800) 889-8969 or (707) 827-7019 • Monday-Friday 7:30am-5pm PT • All trademarks and registered trademarks appearing on oreilly.com are the property of their respective owners. • firstname.lastname@example.org