Ephemeral Storage Optimization Available in v1.28.15+
ScaleOps extends workload rightsizing to ephemeral storage (disk) in addition to CPU and memory. It monitors actual ephemeral storage usage per container, calculates recommendations based on historical data, and applies them at pod creation—reducing over-provisioning and avoiding pod evictions from disk pressure.
Ephemeral storage optimization is enabled by default for built-in policies and disabled by default for custom policies. You can disable it cluster-wide for built-in policies via Helm: ephemeralStorageOptimization.enabled=false. When disabled, the ephemeral storage settings remain visible in the product.
How It Works
ScaleOps tracks ephemeral storage usage and recommends request values the same way it does for CPU and memory. When automation is enabled, these values are applied to new pods at creation time. You can enable or disable ephemeral storage optimization per policy, independently of CPU and memory.
If a pod is evicted as a result of ephemeral storage issues, ScaleOps detects the eviction and adjusts the recommendation so it does not happen again. See Disk Eviction Auto-Healing for details.
Resource Over Time
The Ephemeral storage over time chart appears alongside CPU and Memory in the Analysis and cluster troubleshooting views, showing usage, request, optimized request, and allocatable. Use the “Show ephemeral storage” checkbox above the charts to toggle visibility.
Configuration
Ephemeral storage optimization is configured in the policy and optionally via Helm:
- Policy UI: Ephemeral Storage under Request Configuration
- GitOps: Policies
- Helm: Set
ephemeralStorageOptimization.enabled=falseto disable for built-in policies. See Helm Configuration Reference.