Canton Observability

Overview

Running a Canton app in production needs visibility into your backend and the Canton nodes. Three pillars: logging, metrics, tracing — plus dashboards/alerting. cn-quickstart ships a ready observability stack (Grafana/Prometheus/Loki) you can model yours on.

Logging

• Structured JSON via Logback (Canton nodes and cn-quickstart use it). Console human-readable logs are fine for dev; switch to JSON in prod for ingestion into ELK / Grafana Loki.
• Correlate with the command-id: put it in the MDC (Mapped Diagnostic Context) so you can trace a single Ledger API command through your backend logs and match it to the Canton node logs that processed it.
• What to log: command submissions (command id, template/choice, party), completions (success/failure + error code), slow PQS queries (SQL + elapsed), auth events (token-validation failures, party resolution).
• Don't log contract payloads in production — they may contain private data that must not leave the validator boundary.