git.stella-ops.org/docs/db/reports/scheduler-parity-20251214.md

Scheduler Parity Report · 2025-12-14

Scope

• Backfill and parity verification for Scheduler (Sprint 3402 · PG-T2.9–T2.11).
• Compare MongoDB source vs PostgreSQL target for job/trigger/lease history.

Inputs

• Mongo snapshot:
• Postgres target:
• Backfill tool: Scheduler.Backfill (version/hash)
• Config: Persistence:Scheduler=Postgres after backfill? yes/no

Methods

• Backfill steps:
  1. Restore Mongo snapshot (if applicable)
  2. Run Scheduler.Backfill CLI
  3. Capture logs and row counts per table
• Parity checks:
  • Table counts: jobs, triggers, leases, job_history, metrics
  • Trigger next_fire_at sampling (top 100 by tenant)
  • Determinism checks: order by next_fire_at, tenant_id, id
  • Advisory-lock contention smoke: Acquire/Release sequence on Postgres
• Optional clean-start path: if start-clean approved, document rationale and skip Mongo counts.

Results

• Counts Mongo/Postgres:
  • Jobs: /
  • Triggers: /
  • Leases: /
  • Job history: /
  • Metrics: /
• Determinism sample: pass/fail; details
• Lock smoke: pass/fail; details
• Issues found:

Verdict

• Parity status: PASS / FAIL / START-CLEAN (approved)
• Cutover readiness: YES/NO

Next Actions

• If PASS: mark PG-T2.9–T2.11 DONE and flip Scheduler to Postgres-only.
• If FAIL: log defects + owners; rerun after fixes.
• If START-CLEAN: ensure configs set to Postgres-only and document empty baseline.