Context

At Meta scale, a 10TB event log backing surfaces like Facebook Feed or Instagram Reels can become slow to query and expensive to maintain if it remains a single heap table. Interviewers want to see whether you can design a partitioning strategy that improves scan efficiency without making ingestion and maintenance fragile.

Core question

Explain how you would partition a 10TB events table in PostgreSQL for performance. Your answer should cover:

1. Which partition key you would choose and why
2. Whether you would use range, list, or hash partitioning
3. How query patterns affect the design
4. How indexes should be defined on partitions
5. How you would handle retention, backfills, and skewed tenants or event types
6. What trade-offs or failure modes you would watch for

Scope guidance

Keep the answer practical and PostgreSQL-specific. The interviewer is not looking for generic distributed data lake advice; they want a database design discussion grounded in partition pruning, index strategy, maintenance operations, and operational trade-offs.