Observability — overview

Observability in SOAP-2-REST by Specaria is not a separate product you license and run alongside the platform — it is built into the same no-code UI you manage mappings from. Every conversion call, every configuration change, and every backend test produces telemetry, and all of it is queryable in-app.

This section covers the four observability surfaces:

• Dashboards & KPIs — the operations dashboard: KPI tiles, timeline charts, and per-service rollups.
• Metrics & SLO reports — aggregated metrics over fixed and custom windows, plus SLO reports with availability/latency targets and breach detection.
• Logs & tracing — structured runtime logs, the live traffic log with full payloads/headers, contract-drift detection, async-callback tracking, and debug traces.
• Audit trail — who changed what, when, with before/after state.

All of these are queryable through the Admin (control-plane) API under /admin/v1/* and surfaced in the Observability area of the UI.

---

The one architectural rule: read aggregations, not raw logs

Every time-bucketed metric surface in the product — dashboards, KPI tiles, timeline charts, SLO reports, per-service rollups — reads from pre-computed aggregation tables, never from raw logs at query time. This applies to all traffic sources, not just one.

This is why:

• Dashboards stay sub-second even across large estates with tens of millions of historical rows. A reference estate of ~175 discovered services renders correct 24-hour / 7-day / 1-month per-service counts in sub-second time, because the counts are pre-aggregated.
• Numbers are consistent across surfaces — the dashboard, the SLO report, and the per-service timeline all read the same aggregations, so they agree.

Raw logs are reserved for exactly three jobs:

1. Ingest-health signals — "how long since the last row arrived?" / zero-row-window detection, per source.
2. Drill-down on a specific correlation ID — after you click into a single event, the product reads the raw row for that one correlation ID.
3. Curator workers — background jobs that derive downstream tables (learned examples, aggregations) offline.

When you query a metric and it loads instantly, that is the aggregation read path working as designed.

Traffic sources

Telemetry carries an explicit source dimension so you can filter or sum across origins:

• runtime — traffic the platform's own conversion runtime handled (PUBLISH / CONSUME).
• f5 (and other connectors) — traffic discovered from gateway/load-balancer feeds.

Default metric views sum across all sources; the source dimension lets you narrow to one.

---

Correlation IDs tie it together

A single correlation ID travels through the whole request lifecycle — inbound call, backend call, mapping, structured log, debug trace, and async callback. It is the join key across every surface in this section: a consumer reports an error at 14:32, you search the logs by correlation ID, and you land on the exact call, its full payloads, and its trace. See correlation IDs.

---

Retention

Telemetry and audit data are retained for 60 days by default, and the window is configurable to your policy. See Data retention for the per-table breakdown.

---

Related pages

• Dashboards & KPIs
• Metrics & SLO reports
• Logs & tracing
• Audit trail
• Admin API — observability endpoints

---

All Specaria SOAP to REST docs