Observability
Metrics, logs and traces — the full stack from cluster to portal.
gh0stcloud runs a complete, CNCF-based observability stack. All telemetry is collected with tenant context, stored long-term, and made accessible through the portal — without tenants needing their own monitoring infrastructure.
Telemetry pipeline
Stack components
| Component | Purpose | Layer |
|---|---|---|
| Prometheus | Scraping workload and system metrics | Runtime |
| Alloy | Log collection and trace forwarding (OpenTelemetry) | Runtime |
| Beyla | eBPF-based auto-instrumentation without code changes | Runtime |
| Mimir | Long-term metrics storage (TSDB, Grafana-compatible) | Management Plane |
| Loki | Log aggregation and querying | Management Plane |
| Tempo | Distributed tracing (OTLP-compatible) | Management Plane |
| Grafana | Dashboards, alerting, SLO monitoring | Management Plane |
| Grafana Operator | GitOps-managed Grafana instances | Management Plane |
| OpenCost | Kubernetes cost attribution per namespace | Management Plane |
Tenant telemetry context
All metrics, logs and traces are enriched with tenant context before they enter the backend. Tenant scope is enforced at the backend level — not just as a dashboard filter. A query from Tenant A structurally cannot return data from Tenant B.
Access model by tier
| Tier | Observability access |
|---|---|
| Starter | Aggregated usage trends and runtime summaries in the portal dashboard |
| Growth | Namespace-granular metrics, log streaming, access to Grafana dashboards |
| Enterprise | Full access to metrics + logs + traces, custom Grafana dashboards, alert configuration |
Retention
Retention windows are configured per signal type and contract tier:
| Signal | Default retention |
|---|---|
| Metrics (Mimir) | 90 days |
| Logs (Loki) | 30 days |
| Traces (Tempo) | 14 days |
Extended retention is available for Enterprise contracts.
Alerting
Grafana AlertManager is pre-configured for platform SLOs. From the Growth tier, tenants can define their own alert rules, triggered via webhook, email or PagerDuty.
No custom monitoring stack needed
Beyla instruments applications automatically via eBPF — without code changes, without a sidecar. Tenants immediately get latency, error rate and throughput metrics for their services, even if they have no OpenTelemetry SDK integrated.
Questions or ready to get started?
Talk to us