Questions:
1. Worker thread distribution: Are worker threads created per application instance, or is there a global pool shared across all instances?
2. Multi-instance coordination: How do workers across multiple application instances coordinate to avoid processing the same outbox message?
3. Ordered outbox guarantees: For ordered outboxes, how is sequential processing per tenant enforced when workers span multiple application instances?
4. Resource planning: With thousands of tenants in a multi-instance deployment, what's the recommended approach to sizing worker pools to balance throughput and resource consumption?
Context:
We're running a multi-tenant application with horizontal scaling and need to understand the threading model and coordination mechanisms to properly plan capacity and ensure correct outbox processing behaviour.
We also want to know details pertaining to exponential backoff for retries.
Thanks,
-Vinay.
Bluesky
