Explain the bad prepared plan used for the largest tenant
The same prepared inbox statement serves tiny tenants and the dominant tenant. A plan that works for one distribution performs poorly for the other.
- Focused work estimate
- 5h 30m + prerequisites
- Priority in the scenario
- Medium
- Engineering practice
- Query planning · Data skew
Estimated field mix
- Database engineering60%
- Performance engineering40%
Field percentages are editorial estimates of the ticket's engineering focus. They total 100%; they are not measured time, proficiency scores, or ownership evidence.
Review it, then add it to your workspace.
The board opens an editable draft; nothing is saved until you confirm it. Sign-in and workspace permissions apply, and Demo boards remain ephemeral.
Project context
A fictional maintenance service lists open work orders beside years of closed history. A query that was cheap in a small demo now scans far more rows than it returns. Recreate the schema and synthetic workload locally; no customer database or supplied fixture is assumed.
Setup prerequisites
- Create an isolated local database with tenants, work orders and assignment history.
- Seed 200,000 synthetic work orders with documented skew and a repeatable random seed; record PostgreSQL version, settings and resource limits.
Preceding work
Complete these dependencies, or supply their agreed outputs before taking this ticket.
Acceptance criteria
- Reproduce the plan difference with documented preparation and execution sequences.
- Compare statistics correction, query restructuring and bounded per-query planning choices; select a justified option.
- Keep tenant scope and result ordering unchanged, and document planning overhead as part of the decision.
Implementation constraints
- Do not change a global planner setting as an unexplained workaround; record the exact PostgreSQL version because planning behavior matters.
Verification to include
- Alternate large and small tenant executions across repeated sessions and retain planning plus execution measurements.
- Rebuild statistics after a distribution change and verify the selected solution remains correct or clearly flags an invalid baseline.
Deliverables
- Prepared-plan reproduction and a supported mitigation with regression cases
Rollout and recovery
Limit any configuration change to the affected query path; remove it and restore the baseline query if small-tenant latency breaches its budget.
Value of the work
For the engineer: Practice reading execution plans, balancing read and write cost, and validating performance without weakening tenant boundaries.
For the team: Create reproducible query diagnostics and reversible index or query proposals for an operational application.
Evidence boundaries
Outcome Evidence: Tests, patches, and runbooks are requested deliverables. They become Outcome Evidence only through a qualified Mission and immutable Evidence IDs.
Ownership Evidence: Independent adaptation must be observed under a declared verification policy and cite immutable Evidence IDs. Completing a planning ticket establishes no Ownership Evidence.