Database Infrastructure

I'm preparing database infrastructure code for a **Database Infrastructure** audit. Please help me collect the relevant files.

## Database context (fill in)
- Database engine & version: [e.g. PostgreSQL 16, MySQL 8.4, MongoDB 7, DynamoDB, CockroachDB]
- Hosting: [e.g. AWS RDS, Cloud SQL, PlanetScale, Supabase, self-hosted on EC2, local SQLite]
- ORM / driver: [e.g. Prisma, Drizzle, TypeORM, SQLAlchemy, ActiveRecord, raw pg/mysql2 driver]
- Size: [e.g. "20 tables, largest has 5M rows", "100GB total", "small but growing fast"]
- Traffic: [e.g. "200 queries/sec peak", "mostly reads", "write-heavy event ingestion"]
- Replication: [e.g. "primary + 2 read replicas", "single instance", "multi-region"]
- Known concerns: [e.g. "slow queries on orders table", "no backups configured", "connection pool exhaustion"]

## Files to gather

### 1. Schema (complete and current)
- Full schema definition: CREATE TABLE, CREATE INDEX, CREATE TYPE, CREATE FUNCTION
- OR the ORM schema: Prisma schema, Django models, SQLAlchemy models, ActiveRecord migrations
- Include ALL constraints: PRIMARY KEY, FOREIGN KEY, UNIQUE, CHECK, NOT NULL, DEFAULT
- Include ALL indexes: B-tree, GIN, GiST, partial indexes, expression indexes
- Enum and custom type definitions
- Views and materialised views

### 2. Migration files
- All migration files in chronological order (or at least the last 15-20)
- Any data migration scripts (not just schema changes)
- Migration runner configuration (how migrations are executed in production)
- Any rollback scripts or down migrations

### 3. Connection and pool configuration
- Database client setup: connection string pattern (host, port, database, SSL mode)
- Connection pool settings: min, max, idle timeout, connection timeout, statement timeout
- Read replica routing: how are reads directed to replicas?
- Connection retry and reconnection logic
- Any PgBouncer, ProxySQL, or RDS Proxy configuration

### 4. Query patterns and performance
- The 5-10 most frequently run queries (from APM or pg_stat_statements)
- Any known slow queries and their EXPLAIN ANALYZE output
- N+1 query patterns or batch loading (DataLoader, includes, prefetch_related)
- Full-text search setup (tsvector, GIN indexes, Elasticsearch integration)
- Any query result caching layer

### 5. Backup and recovery
- Automated backup configuration: frequency, retention period, storage location
- Point-in-time recovery (PITR) configuration
- Backup verification: how are backups tested? When was the last restore test?
- Any manual backup scripts (pg_dump, mysqldump)
- Cross-region backup replication

### 6. Monitoring and maintenance
- Slow query log configuration and threshold
- Database monitoring: pg_stat_statements, performance_schema, MongoDB profiler
- Automated maintenance: VACUUM, ANALYZE, index rebuild schedules
- Storage monitoring and growth projections
- Alerting on connection count, replication lag, disk usage, long-running queries

### 7. Security
- Database user roles and permissions (who has superuser? who has read-only?)
- Row-level security (RLS) policies if used
- Encryption at rest configuration
- SSL/TLS for connections (sslmode=require, certificate configuration)
- Audit logging for data access

## Formatting rules

Format each file:
```
--- prisma/schema.prisma (or schema.sql) ---
--- migrations/20240315_add_orders_table.sql ---
--- lib/db.ts (connection/pool config) ---
--- Slow query examples (EXPLAIN ANALYZE output) ---
--- infrastructure/rds.tf (or equivalent) ---
```

## Don't forget
- [ ] Include the FULL schema, not just the tables you think are relevant — the audit catches missing indexes and constraints you didn't consider
- [ ] Include connection pool settings: wrong pool sizes cause cascading failures under load
- [ ] Show how migrations are run in production (manually? CI? automatic on deploy?)
- [ ] Include EXPLAIN ANALYZE for any queries you suspect are slow
- [ ] Note approximate row counts for each table — indexing advice depends on data size
- [ ] Include any seed or fixture data that reveals expected data patterns
- [ ] Note if you use connection pooling proxies (PgBouncer, ProxySQL) and their config

Keep total under 30,000 characters.