Leveraging Debezium for the Inventory Resource outbox will come with the overhead of deploying and managing it. This spike is to investigate the initial and long-term operational complexities of using the Debezium connector for postgres in commercial and FedRAMP environments.

Some potential ideas to dive in to:

• What exists today? Can we use any of it?
• Postgres configuration and setup
• Schema management
• Performance/Scaling implications
• Replication slot management
  • Unprocessed write ahead logging from inactive slots can fill disks. Think about monitoring lag and slot activity.
• Monitoring and Logging
• Error Handling and Recovery
  • Connector failures due to data issues, config changes, instability...etc.
• Data serialization and transformation
  • How do we get our outbox transformed into something Relations and SP can process
• Versioning
  • Compatibility between postgres, debezium, kafka versions
  • Does this have FedRAMP implications?
• Hosted vs on-prem implications (if any)

Check out this write up that Libor previously created: Using Debezium for Write to SpiceDB first, and sync to RBAC