License to Observe - Why observability solutions need agents
summary
Agents sit between apps and storage to translate protocols (OTLP, Prometheus, syslog), enrich and relabel data, do sampling and turn logs into metrics so apps never need redeploys to change telemetry. Deploy patterns include sidecar per app for sensitive endpoints, node-level daemonsets for logs, or clustered collectors (Grafana Alloy supports built-in clustering) to avoid duplicate scrapes; opAMP enables central config. Recommendation: pull metrics, log to files/stdout, and instrument traces first.
AI-written summary · about
chapters (13)
- 0:00Observability agents intro
- 1:44Why agents are needed
- 5:14Agents and collectors roles
- 6:39OpenTelemetry standards
- 8:01Agent data processing
- 11:08Push vs pull metrics
- 13:58Sampling strategies
- 14:40Kubernetes deployment patterns
- 17:27Scaling agents and load balancing
- 20:35Observability setup best practices
- 22:41Chaining collectors
- 24:25Centralized config with OpAMP
- 25:47Q&A on Java
related talks
37mObservability For You And Me With Opentelemetry
46mObservability in the real world: Improve systems w/iterative approach
48mAPI Observability by Design: Building Systems That Tell Their Own Stories
36mObservability from 0 to 100
17mHow Observability-First Development Lets You Ship Agents in Weeks...
29mObservability; a gentle introduction
33mWhat Did My Agent Do? Observability and Accountability for AI Agents
33mWhat Developers Should Know About Observability