Metrics & traces¶

The full catalogue of telemetry Ampora itself emits via OpenTelemetry. Headlines and alerting guidance are on Operations → Self-observability.

Metrics¶

All metrics live under the ampora_* namespace.

OpAMP¶

Metric	Type	Labels	Description
`ampora_opamp_active_sessions`	Gauge	`tenant`, `instance`	Currently open WebSocket sessions
`ampora_opamp_frames_received_total`	Counter	`tenant`, `frame_kind`	Inbound frames
`ampora_opamp_frames_rejected_total`	Counter	`tenant`, `reason`	`oversize`, `malformed`, `cap_mismatch`, `auth_failed`
`ampora_opamp_apply_duration_seconds`	Histogram	`tenant`, `agent_type`	Push → APPLIED wall-clock
`ampora_opamp_bootstrap_redemptions_total`	Counter	`tenant`, `outcome`	First-connect token redemption

Dispatch backplane¶

Metric	Type	Labels	Description
`ampora_dispatch_local_total`	Counter	`tenant`	Same-instance dispatches
`ampora_dispatch_forwarded_total`	Counter	`tenant`, `target_instance`	Cross-instance dispatches
`ampora_dispatch_no_session_total`	Counter	`tenant`	Targeted agent had no live session
`ampora_dispatch_envelope_age_seconds`	Histogram	`tenant`	Time from publish to consume
`ampora_session_ownership_acquires_total`	Counter	`outcome`	Lease acquisition outcome

Live-update bus¶

Metric	Type	Labels	Description
`ampora_live_bus_events_total`	Counter	`tenant`, `event_type`, `direction`	Events published / consumed
`ampora_live_bus_dropped_total`	Counter	`tenant`, `subscriber`	Subscriber overflow drops

Governance¶

Metric	Type	Labels	Description
`ampora_policy_evaluations_total`	Counter	`tenant`, `policy`, `outcome`	`allow` or `deny`
`ampora_policy_denied_total`	Counter	`tenant`, `policy`	Denial-rate per policy
`ampora_policy_evaluation_duration_seconds`	Histogram	`tenant`, `policy`	Per-eval latency
`ampora_audit_archived_total`	Counter	`tenant`	Rows moved hot → archive
`ampora_audit_purged_total`	Counter	`tenant`	Rows purged from archive

Rollouts¶

Metric	Type	Labels	Description
`ampora_rollout_state_transitions_total`	Counter	`tenant`, `from`, `to`	State machine transitions
`ampora_rollout_gate_fired_total`	Counter	`tenant`, `gate`	Per-gate firing rate
`ampora_rollout_targets_succeeded_total`	Counter	`tenant`, `rollout_id`	Per-rollout success counter
`ampora_rollout_targets_failed_total`	Counter	`tenant`, `rollout_id`	Per-rollout failure counter

GitOps¶

Metric	Type	Labels	Description
`ampora_gitops_sync_runs_total`	Counter	`tenant`, `source`, `outcome`	Sweep outcomes
`ampora_gitops_sync_duration_seconds`	Histogram	`tenant`, `source`	Per-sweep duration
`ampora_gitops_files_imported_total`	Counter	`tenant`, `source`	Files producing new versions

Federation¶

Metric	Type	Labels	Description
`ampora_federation_calls_total`	Counter	`tenant`, `peer`, `direction`, `outcome`	Per-peer call counters
`ampora_federation_call_duration_seconds`	Histogram	`tenant`, `peer`, `direction`	Per-call latency

Database¶

Metric	Type	Labels	Description
`ampora_db_query_duration_seconds`	Histogram	`tenant`, `query`	Per-query duration (top queries labelled)
`ampora_db_pool_open_connections`	Gauge	`instance`	Npgsql open connections
`ampora_db_pool_busy_connections`	Gauge	`instance`	Npgsql busy connections

PKI¶

Metric	Type	Labels	Description
`ampora_pki_certs_issued_total`	Counter	`tenant`, `kind`	`agent`, `ocsp`, `federation`
`ampora_pki_certs_revoked_total`	Counter	`tenant`, `kind`
`ampora_pki_signing_duration_seconds`	Histogram	`tenant`, `provider`	Provider-side signing latency
`ampora_pki_crl_publish_duration_seconds`	Histogram	`tenant`	Per-publish duration

Traces¶

Span names are stable; attributes follow OpenTelemetry semantic conventions where applicable.

Span	Notable attributes
`opamp.frame_received`	`agent.id`, `frame.kind`, `frame.size_bytes`
`opamp.frame_sent`	`agent.id`, `frame.kind`
`opamp.dispatch`	`dispatch.target_instance`, `dispatch.outcome`
`opamp.apply_status`	`agent.id`, `remote_config.status`
`live_update.publish`	`event.type`, `tenant`
`live_update.consume`	`event.type`, `subscriber`
`gitops.sync`	`source.id`, `tip.sha`
`federation.outbound_call`	`peer.name`, `endpoint`, `http.status_code`
`policy.evaluate`	`policy.name`, `outcome`
`rollout.tick`	`rollout.id`, `state`

W3C trace context propagates over the dispatch backplane and the live-update bus, so a single trace can link "operator clicked X" to "agent applied Y" across instances.

Logs¶

Structured JSON to stdout. Every line includes:

{
  "timestamp": "2026-04-25T14:30:00.123Z",
  "level": "Information",
  "category": "Ampora.Fleet.Rollouts.RolloutService",
  "message": "Rollout {RolloutId} entered state {State}",
  "trace_id": "...",
  "span_id": "...",
  "tenant_id": "...",
  "RolloutId": "...",
  "State": "InProgress"
}

Log levels follow .NET conventions: Trace < Debug < Information < Warning < Error < Critical. Production deployments default to Information.