Problem

During load testing with 2000 concurrent users, the audit_trails table accumulated 995,412 rows (788 MB) in 7 hours, causing:

• PostgreSQL memory exhaustion (99.77% of 4GB limit)
• RPS drop
• Heavy checkpoint I/O (9000+ buffers every 5 minutes)

The AuditTrailService logs every CRUD operation (including reads) unconditionally with no way to disable it.

Solution

Add AUDIT_TRAIL_ENABLED environment variable (default: false) to control audit trail logging.

Files to modify:

• mcpgateway/config.py - Add audit_trail_enabled setting
• mcpgateway/services/audit_trail_service.py - Check setting in log_action()
• .env.example - Document the setting
• docker-compose.yml - Add setting
• docker-compose-performance.yml - Add setting
• charts/mcp-stack/values.yaml - Add to Helm chart
• docs/docs/manage/configuration.md - Document
• docs/docs/manage/logging.md - Document

Usage

# Disabled for load testing / development (default)
AUDIT_TRAIL_ENABLED=false

# Enabled for production compliance (SOC2, HIPAA, GDPR)
AUDIT_TRAIL_ENABLED=true

Performance Impact

Related

• Audit trail is separate from SECURITY_LOGGING_ENABLED (which controls security_events table)
• Audit trail logs: view_server, view_prompt, create_*, update_*, delete_* actions