Observability
Live resource metrics, network traffic, distributed traces, and CPU/memory snapshots in one place.
The Observability page brings your deeper telemetry together across seven tabs: metrics, network traffic, distributed traces, per-container/pod resource snapshots, application logs, latency probes, and API request tracking.
What you'll see
- Metrics: a live snapshot of CPU% and memory for every Docker container and Kubernetes pod.
- Traffic: network RX/TX bytes and rate, per resource.
- Traces: distributed request flows across services, with per-span timing.
- Profiles: the same live CPU/memory snapshot as Metrics, framed against each container or pod's configured limit, to spot what's closest to its ceiling.
- OpenTelemetry: your application's OTLP-ingested logs, filterable by severity and free-text search. See Logs.
- Latency: round-trip probe latency for agents, nodes, containers, and any custom target. See Latency.
- API Requests: throughput, error rate, and latency percentiles for your reported API traffic, including a p95-latency trend and an error/latency correlation report by route. See API Requests.
All seven tabs refresh automatically in the background; Metrics, Traffic, and Profiles have always done this, and Traces, OpenTelemetry, Latency, and API Requests now do too.
Kubernetes pod CPU/memory (Metrics and Profiles) works once [metrics-server](https://github.com/kubernetes-sigs/metrics-server) is installed on the cluster; without it, those rows read zero. Pod network throughput reaches the Traffic tab a different way, pulled from each node's Kubelet `/stats/summary` endpoint, so it works even without metrics-server. The Metrics tab's own Net RX/TX and Disk R/W columns are Docker-only today though: a Kubernetes pod row there shows CPU% and Mem% but not those two columns.
Getting data in
- Metrics, traffic, and profiles flow automatically from the agent once it's connected.
- Distributed traces require a tracing backend. Add a Jaeger or Zipkin integration (see Integrations), or send spans via OpenTelemetry (OTLP). Until a tracing source is connected, the traces panel shows an empty state.
Metrics, Traffic, Profiles, OpenTelemetry, Latency, and API Requests are available on every plan. Traces (distributed tracing) is Pro and Enterprise only; on Free, the trace-backend calls are rejected and the tab just shows a failure to reach the backend.