Scaling to a Million Machines with Prometheus

PromCon · 2016 · Berlin · 2016-08-25
prometheusmonitoringdevopsdigitalocean video

PromCon 2016 · Berlin

How DigitalOcean ran a public-facing monitoring platform on Prometheus toward multi-datacenter, high-cardinality load.

Themes

  • Prometheus as the customer-facing metrics plane
  • Sharding, storage pressure, and operational reality past a few thousand machines
  • Lessons that fed later work on Vulcan (compatible timeseries pipeline)