Services and health
Understand service identity, activity summaries, resource attributes, and missing evidence.
Find a service
Open Observability → Services and select a service to open its details. The inventory lists every service discovered from spans, logs, or metrics, even when it has no recent traffic or its older telemetry has expired. Search, environment, and health filters narrow the list; time ranges only affect measurements, not service discovery.
Services are discovered from reported telemetry, not created by adding a name to a dashboard form. Initialize an SDK and send telemetry to establish activity. The inventory shows measurements from the last 24 hours; service details lets you choose another period.
Keep serviceName stable and descriptive, such as orders-api. Use serviceVersion, environment, and optional serviceNamespace to describe deployment context. Only add non-secret resource attributes.
Read the activity summary
Service details shows operations, throughput, error rate, and P95 latency for the selected period. Where server spans are available, operations and throughput represent completed server operations. Otherwise they can represent reported spans; do not assume every span is one incoming HTTP request.
Error rate counts HTTP 5xx responses or spans explicitly marked as failed, using the same failure definition as Requests and request-error alerts. Ordinary 4xx responses do not increase it; use Requests to inspect their separate rate and status-code breakdown.
P95 describes the measured distribution, not the slowest request. For a specific operation, follow the links to Requests or Traces. Resource details show the latest reported environment, namespace, version, region, instrumentation scope, and last telemetry timestamp.
Interpret health cautiously
The Healthy, Degraded, Critical, or Unknown badge reflects telemetry available in the selected range. No measured operations leaves health, rates, and latency unknown, but the service stays listed. Services that only send logs or metrics also remain listed without implying healthy request traffic.
This is not an external reachability test or a guarantee that every dependency is healthy. Use Uptime monitors for HTTP availability checks and alerts for explicit threshold rules.
Troubleshoot a missing or stale service
Confirm the organization matches the token, the configured service name is what you expect, and telemetry is being exported. A service does not disappear from the inventory just because it stopped reporting. An unknown health state is not proof of an outage.
When refresh fails, the dashboard can retain the last available evidence and show its update time. Read that warning: old observations do not become current because the page is open. Missing resource attributes mean the application did not report them, not that OutRay inferred a value.