Harden the expanded server with auth, observability, and integration tests #62

New issue

Open

opened 2026-07-26 02:35:16 +00:00 by lost-rob0t · 0 comments

lost-rob0t commented

2026-07-26 02:35:16 +00:00

(Migrated from github.com)

Goal

Make the expanded server operable and safe under real ingestion and investigation workloads.

Scope

request IDs propagated HTTP → RabbitMQ → actors → CouchDB events
structured logs and metrics for routes, consumers, actors, queues, view queries, validation, and dead letters
readiness/liveness endpoints that distinguish CouchDB, RabbitMQ, actor-system, and HTTP health
rate limits, payload limits, timeouts, and cancellation
permission model for document read/write, target submit/control, stream subscribe, and view query
secret loading/rotation; redact credentials and sensitive headers from logs
integration environment with RabbitMQ + CouchDB and deterministic fixtures
contract, load, reconnect, failure-injection, and upgrade tests
operator documentation for backup, migration, queue recovery, and stuck delivery diagnosis

Acceptance criteria

one request/target can be traced across all services
health endpoints accurately report degraded dependencies
CI proves API, broker, database, and actor behavior together
overload produces bounded failures instead of unbounded memory/queue growth

Build alongside #58, #59, #60, and #61 rather than as a final cleanup pass.

## Goal Make the expanded server operable and safe under real ingestion and investigation workloads. ## Scope - request IDs propagated HTTP → RabbitMQ → actors → CouchDB events - structured logs and metrics for routes, consumers, actors, queues, view queries, validation, and dead letters - readiness/liveness endpoints that distinguish CouchDB, RabbitMQ, actor-system, and HTTP health - rate limits, payload limits, timeouts, and cancellation - permission model for document read/write, target submit/control, stream subscribe, and view query - secret loading/rotation; redact credentials and sensitive headers from logs - integration environment with RabbitMQ + CouchDB and deterministic fixtures - contract, load, reconnect, failure-injection, and upgrade tests - operator documentation for backup, migration, queue recovery, and stuck delivery diagnosis ## Acceptance criteria - one request/target can be traced across all services - health endpoints accurately report degraded dependencies - CI proves API, broker, database, and actor behavior together - overload produces bounded failures instead of unbounded memory/queue growth Build alongside #58, #59, #60, and #61 rather than as a final cleanup pass.