Monitoring as a First Step to a New Service
summary
Argument: monitor first, before there is data and before deployment. Prototype uses console and its DNS HTTP key value interface; initial worries about memory and process interference prove unfounded in staging. Two metrics matter: do we have a leader and leadership transitions. Production shows many transitions until node count rises (staging → prod has 3.5× nodes, five nodes stabilizes). Building config on one node and distributing via the KV store removes pauses and cluster-wide failures; monitoring guides those concrete changes.
AI-written summary · about
chapters (10)
- 0:00Start
- 0:23AT THE END OF A PROJECT MONITOR LAST
- 0:43SETUP YOUR TOOLS
- 1:13HOW DO YOU KNOW
- 1:43CONSUL
- 1:58FEAR IS THE MIND KILLER
- 2:13STAGING
- 2:28YOUR FIRST 3 GRAPHS
- 4:29EXPERIMENT
- 4:44SUSPICIOUS SHAPES
related talks
36mIt's not in Production Unless it's Monitored
39mThe Story of One Monitoring
35mMonitoring Microservices
28mProduction Testing through Monitoring
28mMonitoring Stack was yesterday – Observability Platform at scale!
09mMonitoring? Not OPS problem
43mDistributed Monitoring: How to Understand the Chaos
10mMonitoring Microservices: A Challenge