Building a Scalable Observability Platform for Edge Compute Nodes
summary
Modernized vessel edge from Docker to k3s single‑ or multi‑node clusters and runs on‑board Prometheus collecting at 1‑minute intervals with ~500 MB tuned footprint, shipping only selected metrics via a custom identifier → Kafka ingest to reduce bandwidth. Cloud stack is Prometheus+Thanos with downsampling and a hash‑ring for per‑tenant storage, and logging uses Loki+Fluent Bit with protobuff compression, Falco alerts, OAuth2 proxy + Keycloak JWTs, 15‑day local retention and retry/caching for intermittent satellite links.
AI-written summary · about
chapters (7)
- 0:00Edge compute and observability
- 7:25Monitoring edge nodes challenges
- 13:17Log collection challenges at edge
- 17:33Security for edge observability
- 20:01Multi-tenancy in observability storage
- 27:49Lessons learned for edge observability
- 39:27Q&A on edge data
related talks
24mObservability at Scale: How Open Systems Collects Telemetry from Over 10'000 Edge Devices Worldwide
35mFrom New Cluster to Insight. Deploying Monitoring and Logging to Kubernetes
25mMulti-Cluster Observability: What you need to survive with a fleet
24mTaming Observability at Scale in a Multi-Cluster Kubernetes Platform at Bloom...
28mMonitoring Stack was yesterday – Observability Platform at scale!
26mPrometheus That Scales
31mObservability: Dust Off Prometheus with VictoriaMetrics
01h 02mDesigning an open-source observability stack for Kubernetes