OpenTelemetry Collector Deployment Patterns
Talkat KubeCon + CloudNativeCon North America 2021 · on Wednesday, 2021-10-13· at 15:25· Room 403 AB + Online·
summary
Covers concrete OpenTelemetry Collector setups: agent and collector sidecar and daemonset patterns with per-pod tuning, a normalizer that unifies labels between Prometheus and Jaeger, a load‑balancing layer that hashes trace id to guarantee all spans for a trace hit the same backend (example uses three replicas), a multi‑cluster control‑plane relay, multi‑tenant routing and a per‑signal split for metrics vs traces. Live Kubernetes demos run and configs plus contrib images sit in the repo; typical collector sidecar memory is ~5–10 MB.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 0:55OpenTelemetry
- 7:02Patterns
- 15:10Load Balancing
- 18:09MultiCluster
- 19:50Multitenant
- 21:16Per Signal
- 23:12Summary
related talks
28mImplementing Observability in Kubernetes
34mKubernetes Observability with OpenTelemetry
23mControlling Data Overhead with the OpenTelemetry Collector
35mOpenTelemetry for Busy People
49mAdding OpenTelemetry to Production Apps Lessons Learned
16mUsing the OpenTelemetry Collector to Empower End Users
22mOpenTelemetry – An Observability Framework for Cloud-Native Software
25mOpenTelemetry for Platform Engineers: Paving the Road To Self-Service Observability