Scaling
Measure first, then change the service compute tier
Replov currently provides vertical scaling through service plan changes. Automatic replica scaling is not part of the production core.
1. Inspect the service
Use service logs, health information, CPU and memory metrics, and plan limits to identify the actual bottleneck.
2. Change the compute tier
- 1. Open the service details page and go to its plan settings.
- 2. Select an active compute plan and billing type.
- 3. Confirm the price and apply the plan change.
- 4. Watch service state and metrics after Docker applies the new CPU and memory limits.
Plan changes can restart a service
A plan change updates Docker service constraints and may briefly interrupt traffic. Schedule production changes appropriately and verify the application afterward.
3. Scale the application safely
Move sessions and shared state to Redis or another datastore before adding horizontal replicas in a future release.
Keep workers idempotent, use application-level queue retry rules, and use database locks or unique constraints for operations that must run once.
Use Vitibase connection pooling for PostgreSQL-backed applications and size the pool for the selected compute tier.