Docs
Kiwi Fabric
Version Current

Configuration reference

Observability and telemetry

Request logs, latency and per-user reports, plus OpenTelemetry logs, traces, metrics and health endpoints.

Fabric Current 2 min read Updated

Request telemetry#

Every query and saga is recorded with the request type, status, user id, duration, row count, a SQL hash and a timestamp. Source breakdowns add, per source: rows fetched, fetch time, whether predicate and projection pushdown happened, an estimate of bytes transferred, cache hits and the exact query sent to the source.

bash
kiwi fabric telemetry                                  # counts, failure rate and average latency
kiwi fabric telemetry-log --type Saga --status Failed  # browse the request log
kiwi fabric telemetry-latency --bucket hour            # average and p95 latency over time
kiwi fabric telemetry-users                            # per-user counts, failure rates, latency
kiwi fabric telemetry-request --id <request-guid>      # one request, with per-source breakdown

--status accepts Success, Failed, Compensated and DeadLettered. Time windows use ISO 8601 --from and --to.

The Web UI request-detail view shows a waterfall timeline (total, federation, and one bar per source fetch) and, for sagas, the step trace.

The Query Log page listing requests with time, type, user, status, duration and row count, with filters for user, type, status and date range
The Query Log: every request, filterable by user, type, status and date.
The Request Detail page with a summary and an execution timeline showing the time spent fetching from each source and in federation
A request in detail: summary and execution timeline.

OpenTelemetry#

Logs, distributed traces and metrics share one OpenTelemetry pipeline. Output goes to the console by default; an OTLP exporter or a first-class Azure Monitor exporter can be enabled by configuration, with no code change. Every query and every handler fetch is a traced span with its duration and row counts.

Health endpoints#

Liveness and readiness endpoints are provided for orchestration platforms.

Table usage and compensation summary#

The REST API also exposes GET /api/telemetry/table-usage and GET /api/telemetry/compensation-summary, which feed the Web UI dashboards. See the REST API reference.