Skip to content

test(v0.49): drain two real MDM cases after downtime with bounded worker batches - #18

Merged
grove merged 4 commits into
issue/11-explicit-escalation-timefrom
test/v049-two-case-downtime-drain
Oct 21, 2026
Merged

grove merged 4 commits into
issue/11-explicit-escalation-timefrom
test/v049-two-case-downtime-drain

Conversation

@grove

@grove grove commented Oct 21, 2026

Copy link
Copy Markdown
Contributor

Scope — real two-case overdue backlog after PostgreSQL downtime

Dependent draft based on the current React #14 feature branch at 2ee3f6e. Adds an optional --multi-outage mode to the existing exact-MDM-pinned joint-stack audit. The normal full audit, single-case outage, existing temporal SQL and production paths remain unchanged.

Real MDM fixture, not synthetic routing rows

  • Select an additional real, open MDM occurrence with no previous applied level-2 escalation. Use authenticated mdm_steward.set_case_controls to prepare its level-1 / future-deadline state before the React deadline rule registers.
  • Keep both original MDM case_key/occurrence identities and deadline values, one second apart. Assert both are genuinely future due dates when the database stops.
  • Set pg_react.batch_size=1 and stop the whole PostgreSQL container before both due times. Keep it offline until both deadlines have passed, then restart without modifying any business-source rows or manually running a refresh.
  • Wait for exactly two distinct APPLIED_CONTROL receipts issued by the ordinary managed worker, with per-case next-level-2 control, complete expected MDM case/source populations, exact request and receipt bindings, immutable request/attempt records and no extra actions.
  • Observe subsequent managed heartbeats and a second cold PostgreSQL restart; require unchanged complete durable case, receipt, intent, work, attempt, business and technical-source vectors.
  • Run this additional mode after the unchanged full audit and existing one-case outage in .github/workflows/producer-compatibility.yml; retain separate JSON/SQL logs in the pinned-producer artifact.

Deliberate qualification boundary

This is an executable two-case managed outage/backlog correctness candidate. Configured batch_size=1 and observed two-case drain do not independently establish the configured per-cycle admission count, backlog fairness under mixed churn, supported drain latency, 300-case throughput, resource budgets, clone/restore, signed operating profile or full T49.06/T49.12. No new scheduler, broad runtime rewrite, production policy enabling or dependency pin change.

Exact-head GitHub CI is pending. Keep draft. Do not merge into main, close React #11/#12 or claim an approved release.

@grove
grove marked this pull request as ready for review October 21, 2026 20:19
@grove
grove merged commit 1d44dfd into issue/11-explicit-escalation-time Oct 21, 2026
7 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant