Observability collects the three signal types — metrics, logs and traces — from every service and makes them queryable. It is how on-call engineers answer “is the fleet healthy?” and “why did this request fail?” without SSH-ing into a box.
Three pipelines, one collector edge
Every service ships signals to a local OpenTelemetry Collector, which fans them out to the right backend. Splitting metrics, logs and traces keeps each storage tier tuned to its very different access pattern.
Could not build diagram for model "example-iot".
Prometheus scraping and remote write for time-series metrics.
Loki + Vector aggregation for structured logs.
Tempo + OTel collectors for distributed traces.
High-resolution series kept 15d; downsampled rollups longer for capacity planning.