Real-time-first metrics
summary
Argues that pushing netdata-style per-second metrics to the machine nearest the process speeds debugging and decision-making, with examples: custom stats pushed over the statsd protocol from app containers, a bash distributed-graph script that parallelizes metric collection, and forwarding summaries to CloudWatch 1–5 minute granularity for long-term alerts. Shows a real case of Lambda/Elasticsearch dropping messages under load (queue capacity ~50) and an alert that predicts disk full in six hours, and demonstrates wrappers that capture stdout/stderr and work in production.
AI-written summary · about
chapters (5)
- 0:00Start
- 6:09Decision-Making and Debugging
- 16:48Debugging
- 26:12Demo
- 40:50Add some Custom Metrics to the Application
related talks
24mNew approach to metrics observability for the cloud era
36mMetrics - What's your code actually doing?
33mNetdata: Open Source, Distributed Observability Pipeline – Journey and Challenges
37mYou Can't Optimize What You Can't Measure
44mActionable insight with derived second order monitoring data
10mTurning data into leverage
44mNetdata: Open Source, Distributed Observability Pipeline – Journey and Challenge
41mKnow your data: The stats behind your alerts