Development Guide

• SHOULD read guidelines first - Check development-guide.md before making schema changes
• SHOULD use preferred language patterns - Check language.md
• SHOULD Execute queries directly - PREFER psql via scripts/psql-connect.sh for ad-hoc queries
• REQUIRED: Follow DDL Guidelines - Refer to DDL Rules
• SHALL repeatedly generate fresh tokens - Refer to Connection Limits
• ALWAYS use ASYNC indexes - CREATE INDEX ASYNC is mandatory
• MUST Serialize arrays/JSON as TEXT - Store arrays/JSON as TEXT (comma separated, JSON.stringify)
• ALWAYS Batch within row limit - maintain transaction limits (defaults: 3,000 rows, 10 MiB, 5 minutes — verify via the AWS MCP Server’s aws___search_documentation if available, or check the DSQL documentation (opens in a new tab) directly: aurora dsql transaction limits)
• REQUIRED: Sanitize SQL inputs with allowlists, regex, and quote escaping - See Input Validation
• MUST follow correct Application Layer Patterns - when multi-tenant isolation or application referential integrity are required; refer to Application Layer Patterns
• REQUIRED use DELETE for truncation - DELETE is the only supported operation for truncation
• SHOULD test any migrations - Verify DDL on dev clusters before production
• Plan for Horizontal Scale - DSQL is designed to optimize for massive scales without latency drops; refer to Horizontal Scaling
• SHOULD use connection pooling in production applications - Refer to Connection Pooling
• SHOULD debug with the troubleshooting guide: - Always refer to the resources and guidelines in troubleshooting.md
• ALWAYS use scoped roles for applications - Create database roles with dsql:DbConnect; refer to Access Control

Detailed References

• authentication-guide.md — IAM auth, token management, secrets, SSL/TLS, connection pooling, audit logging, access control
• connectivity-tools.md — Database drivers, ORMs, adapters, and data loading tools
• scaling-guide.md — Horizontal scaling strategy, batch optimization, hot key avoidance, identifier types

Operational Rules

Query Execution

For Ad-Hoc Queries and Data Exploration:

• MUST ALWAYS Execute via psql (use scripts/psql-connect.sh --command for single statements, --script for multi-statement files) or your driver’s read path
• SHOULD Return results immediately

Writing Scripts REQUIRES at least 1 of:

• Permanent migrations in database
• Reusable utilities
• EXPLICIT user request

Schema Design Rules

• MUST use simple PostgreSQL types: VARCHAR, TEXT, INTEGER, BOOLEAN, TIMESTAMP
• MUST store arrays as TEXT (comma-separated is recommended)
• MUST store JSON objects as TEXT (JSON.stringify)
• ALWAYS include tenant_id in tables for multi-tenant isolation
• SHOULD create async indexes for tenant_id and common query patterns

Schema (DDL) Rules

• REQUIRED: at most one DDL statement per operation
• ALWAYS separate schema (DDL) and data (DML) changes
• MUST use CREATE INDEX ASYNC: No synchronous creation (defaults: max 24 indexes per table, 8 columns per index — verify via the AWS MCP Server’s aws___search_documentation if available, or check the DSQL documentation (opens in a new tab): aurora dsql index limits)
  • MAXIMUM: 24 indexes per table
  • MAXIMUM: 8 columns per index
• Asynchronous Execution: DDL ALWAYS runs asynchronously
• To add a column with DEFAULT or NOT NULL:
  1. MUST issue ADD COLUMN specifying only the column name and data type
  2. MUST then issue UPDATE to populate existing rows
  3. MAY then issue ALTER COLUMN to apply the constraint
• MUST issue a separate ALTER TABLE statement for each column modification.

Transaction Rules

Defaults below; verify against the live limits via the AWS MCP Server’s aws___search_documentation if available (aurora dsql transaction limits), or read the DSQL documentation (opens in a new tab) directly:

• SHOULD modify at most 3,000 rows per transaction
• SHOULD have maximum 10 MiB data size per write transaction
• SHOULD expect 5-minute transaction duration
• ALWAYS expect repeatable read isolation

Application-Layer Patterns

MANDATORY for Application Referential Integrity: If foreign key constraints (application referential integrity) are required, implement the following pattern instead:

• MUST validate parent references before INSERT
• MUST check for dependents before DELETE
• MUST implement cascade logic in application code
• MUST handle orphaned records in application layer

MANDATORY for Multi-Tenant Isolation:

• tenantId is ALWAYS first parameter in repository methods
• ALL queries include WHERE tenant_id = ?
• ALWAYS validate tenant ownership before operations
• ALWAYS reject cross-tenant data access

Migration Patterns

• REQUIRED: One DDL statement per migration step
• SHOULD Use IF NOT EXISTS for idempotency
• SHOULD Add column first, then UPDATE with defaults
• REQUIRED: Each DDL executes separately

Quick Reference

Schema Operations

Supported Data Types

Supported Key

Join on any keys; DSQL enforces PRIMARY KEY, UNIQUE, NOT NULL, and CHECK constraints at the database level. Foreign-key referential integrity must be enforced in the application layer (see Application-Layer Patterns above).

Transaction Requirements

Defaults below; verify against the live limits via the AWS MCP Server’s aws___search_documentation if available (aurora dsql transaction limits), or read the DSQL documentation (opens in a new tab) directly: