Operate¶

Pages for the people running Underwrite in production — SREs, operators, on-call engineers.

Multi-stage Docker image, multi-arch builds, container hardening, and the production environment variables.

Day-2 operations — restarts, log capture, configuration reloads, key rotation, and incident triage.

Prometheus metrics on /metrics, OpenTelemetry tracing (console / OTLP), structured log output, and what to alert on.

/healthz, /readyz, and /v1/health — what each probe checks and how to wire them into Kubernetes.

Inspecting the dead-letter queue, replaying events, and the rules of thumb for poisoning recovery.

Throughput baselines, latency budgets, and the knobs that move each.

The SQLite store, WAL journaling, busy timeout, schema migrations, and durable DLQ.

Transactional schema migrations and how to write one.

Reading the audit ledger, dumping traces, and the most common runtime traps.

Tag-driven publishing, reproducible wheels, GitHub Actions matrix, and the on-call checklist.

Keep-a-Changelog format, release categorization, and how to move items from Unreleased to a tagged section.

Building the multi-arch image, running it locally, and the docker-compose profiles.