Scaling Policies

Scaler’s auto-scaling is controlled by the scheduler policy engine. Pick the engine with policy_engine_type, then configure it with policy_content.

Available policy engines:

Policy Engine	Description	Best For
simple	General-purpose engine for most deployments.	One manager or a small manager set Homogeneous workloads: even_load + vanilla Capability-aware workloads: capability + capability
waterfall_v1	Priority-based multi-manager scaling.	Strict tier preference across managers Cost tiers (local/on-prem first, cloud overflow second) Latency tiers where preferred resources fill first Deterministic scale-up and scale-down order