Skip to main content

Data, security, and observability

This page groups the core platform services that support WardMitra reliability and governance.

Primary relational database​

Use Amazon RDS for PostgreSQL 15 as the primary operational database target.

Recommended baseline:

  • Multi-AZ for production
  • automated backups enabled
  • point-in-time recovery enabled
  • PgBouncer for connection pooling

Legacy data migration​

The current Aviel DB usage needs a dedicated migration assessment. Deliverable 1 should assume:

  • schema review before final cutover design
  • separate migration rehearsal in non-production
  • rollback criteria defined before production migration

Document or non-relational store​

Where document-style storage is still required, evaluate:

  • MongoDB Atlas, if SPWHI wants a managed external service
  • Amazon DocumentDB, if keeping more inside AWS is preferred

This decision can stay open in Deliverable 1, but the architecture should reserve room for it.

Object storage​

Use S3 for:

  • complaint photos and uploads
  • exported reports
  • model artifacts if AI workloads are introduced
  • static web assets

Security and secrets baseline​

Detailed policy work belongs to Deliverable 3, but the architecture should already assume the following baseline:

  • IRSA for workload-level AWS access
  • SSM Parameter Store for most config values
  • Secrets Manager only where managed rotation is actually useful, especially database credentials
  • least-privilege IAM roles
  • WAF in front of the public ingress path
  • CloudTrail and audit logging enabled

Observability baseline​

Detailed monitoring playbooks will come later, but the target architecture should budget for:

  • metrics with Prometheus and Grafana
  • log aggregation with Loki or CloudWatch-backed equivalents
  • alerting for application and infrastructure failure paths
  • audit visibility for CI/CD and AWS control-plane changes

At minimum, the team should be able to answer:

  • is the API healthy?
  • are pods restarting or failing?
  • is the database under pressure?
  • did a recent deployment cause the issue?