Story: Custom MQ tables and in-process scheduler

Goal
Status
Acceptance
Tasks
Decisions
Out of scope
See also

This page documents a story in Sprint 14. It captures the goal, current status, acceptance criteria, and the tasks that compose it.

Goal

Replace the external pgmq + pg_cron extensions with custom in- process tables + scheduler. Tighter integration with RLS, full control over data structures, unified codebase.

Status

Field	Value
State	DONE
Parent sprint	Sprint 14
Now	Completed 2026-03-04.
Waiting on	None.
Next	None.
Last touched	2026-03-04

Continued from: Scheduling subsystem (sprint 13). That story built ores.scheduler on top of pg_cron; this one replaces the external pgmq + pg_cron extensions entirely with custom in-process infrastructure.

Acceptance

pgmq + pg_cron extensions removed.
Custom ores_mq_* tables with scope + queue type + RLS.
In-process scheduler_loop using boost::asio.
TimescaleDB hypertable for job instance history.
Protocol 48.0 (breaking).

Tasks

Task	State	Start	End	Description
Replace pgmq + pg_cron with custom tables and scheduler	DONE	2026-05-20	2026-03-04	Remove pgmq + pg_cron extensions; custom ores_mq_* tables (party / tenant / system scope; task / channel queue types; native RLS); in-process scheduler_loop using boost::asio::steady_timer with sql_action_handler + mq_action_handler; ores_scheduler_job_instances_tbl TimescaleDB hypertable replaces cron.job_run_details; mq_service facade; protocol 48.0 (breaking).

Decisions

In-process scheduler over external pg_cron: we own the action handlers (sql_action_handler + mq_action_handler); RLS-aware by construction.
Custom MQ tables with scope levels: party / tenant / system scoping matches the RLS pattern instead of fighting it.

Out of scope

Inter-process messaging (covered by the NNG → NATS journey later in the sprint).