Observability
Per-GPU telemetry, pod metrics, logs and traces — OpenTelemetry-compatible export.
What it is
Every pod, function and managed Kubernetes node emits metrics, logs and traces into AIARCO Observe. For GPUs you get DCGM-class telemetry: utilisation, memory, power, temperature, ECC errors. Everything is exportable via OTLP to any compatible back-end you operate yourself.
When to use it
- Spot under-utilised GPUs across training runs.
- Send AIARCO telemetry to your own observability stack via OTLP.
- Build dashboards for cost per request, p50/p99 latency, error rate.
Quickstart
asc observe metrics query 'gpu_utilization{pod="pod_abc"}' --from -1h
asc observe logs tail --pod pod_abcrows = client.observe.metrics.query('gpu_utilization{pod="pod_abc"}', from_='-1h')const rows = await client.observe.metrics.query('gpu_utilization{pod="pod_abc"}', { from: '-1h' });Limits & quotas
| Limit | Default | Burst | Notes |
|---|---|---|---|
| Metric retention | 30 days hot / 13 months rollup | — | |
| Log retention | 14 days | — | Configurable up to 90 days |
| OTLP export | Free | — | Endpoint per region |
Pricing
See pricing. Pay-as-you-go, billed monthly via Stripe.
API surface
GET /v1/observe/metricsGET /v1/observe/logs
Required scope(s): observe:read. See Scopes.
Security
All access is authenticated and scoped. See Auth & scopes and Network controls.